Documente Academic
Documente Profesional
Documente Cultură
MODELLING
OF WELDING
Mathematical Modelling of Weld Phenomena Edited by H. Cerjak and K. E. Easterling Book 533 ISBN 0 90171616 2. 1992 Mathematical Modelling of Weld Phenomena 2 Edited by H. Cerjak and H. K. D. H. Bhadeshia Book 594 ISBN 0 901716 63 4. 1994 Mathematical Modelling of Weld Phenomena 3 Edited by H. Cerjak and H. K. D. H. Bhadeshia Book 650 ISBN 1 86125 010 X
MATERIALS MODELLING
SERIES
Editor: H. K. D. H. Bhadeshia
The University of Cambridge Department of Materials Science and Metallurgy
Book 677 First published in 1997 by The Institute of Materials 1 Carlton House Terrace London SWIY 5DB First edition (Book 557) Published in 1994 The Institute of Materials 1997 All rights reserved ISBN 1 86125 036 3
Originally typeset by PicA Publishing Services Additional typesetting and corrections by Fakenham Photosetting Ltd Printed and bound in the UK at The University Press, Cambridge
TO TORHILD, TORBJ0RN
AND HAvARD
without your support, this book would never have been finished.
Contents
xiii xiv
1
1 1 2 4 7 10 14 '18 24 24 26 26 28 31 41 45 48 49 56 59 61 72 75 77 77 80 91 95 96 96 99 100 103 105 110 111 112
"
0-
CONTENTS
vii 116
116 116 117 120 120 121 122 123 123 124 124 128 128 128 130 131 131 134 138 140 140 141 141 142 142 143 146 148 148 149 149 152 156 160 166 166 169 170 171 172 173 174 176 177 179 180 182
viii
CONTENTS
2.11.2 Growth and separation of oxide inclusions 2.11.2.1 Buoyancy (Stokes flotation) 2.11.2.2 Fluid flow pattern 2.11.2.3 Separation model........................................................................................................ 2.11.3 Predictions of retained oxygen in the weld metal 2.11.3.1 Thermodynamic model 2.11.3.2 Implications of model 2.12 Non-Metallic Inclusions in Steel Weld Metals 2.12.1 Volume fraction of inclusions 2.12.2 Size distribution of inclusions 2.12.2.1 Effect of heat input 2.12.2.2 Coarsening mechanism 2.12.2.3 Proposed deoxidation model 2.12.3 Constituent elements and phases in inclusions 2.12.3.1 Aluminium, silicon and manganese contents 2.12.3.2 Copper and sulphur contents 2.12.3.3 Titanium and nitrogen contents 2.12.3.4 Constituent phases 2.12.4 Prediction of inclusion composition 2.12.4.1 C-Mn steel weld metals 2.12.4.2 Low-alloy steel weld metals References Appendix 2.1: Nomenclature Appendix 2.2: Derivation of equation (2-60)
184 185 186 188 190 190 192 192 193 195 196 196 201 202 202 202 203 204 204 204 206 212 215 219
Chapter
3: Solidification
Behaviour
221
221 221 222 225 226 228 228 229 230 230 234 239 240 242 251 251 254 254 256 260 260 261
3.1 Introduction 3.2 Structural Zones in Castings and Welds 3.3 Epitaxial Solidification 3.3.1 Energy barrier to nucleation 3.3.2 Implications of epitaxial solidification 3.4 Weld Pool Shape and Columnar Grain Structures 3.4.1 Weld pool geometry 3.4.2 Columnar grain morphology 3.4.3 Growth rate of columnar grains 3.4.3.1 Nominal crystal growth rate 3.4.3.2 Local crystal growth rate 3.4.4 Reorientation of columnar grains 3.4.4.1 Bowing of crystals 3.4.4.2 Renucleation of crystals 3.5 Solidification Microstructures 3.5.1 Substructure characteristics 3.5.2 Stability of the solidification front 3.5.2.1 Interface stability criterion 3.5.2.2 Factors affecting the interface stability 3.5.3 Dendrite morphology 3.5.3.1 Dendrite tip radius 3.5.3.2 Primary dendrite arm spacing
CONTENTS
ix
264 268 268 --272 272 272 278 279 279 281 284 286 286 286 290 290 290 290 292 292 293 296
3.5.3.3 Secondary dendrite arm spacing 3.6 Equiaxed Dendritic Growth 3.6.1 Columnar to equiaxed transition 3.6.2 Nucleation mechanisms 3.7 Solute Redistribution 3.7.1 Microsegregation 3.7.2 Macrosegregation 3.7.3 Gas porosity 3.7.3.1 Nucleation of gas bubbles 3.7.3.2 Growth and detachment of gas bubbles 3.7.3.3 Separation of gas bubbles 3.7.4 Removal of microsegregations during cooling . 3.7.4.1 Diffusion model 3.7.4.2 Application to continuous cooling 3.8 Peritectic Solidification 3.8.1 Primary precipitation of the 'Yp -phase 3.8.2 Transformation behaviour of low-alloy steel weld metals 3.8.2.1 Primary precipitation of delta ferrite 3.8.2.2 Primary precipitation of austenite 3.8.2.3 Primary precipitation of both delta ferrite and austenite References Appendix 3.1: Nomenclature
301
301 301 301 303 304 304 308 314 314 314 315 315 316 316 319 322 325 326 328 329 332 334 334
;..-
CONTENTS
337
337 337 337 339 340 341 342 343 343 345 345 345 347 348 351 356 360 360 360 361 361 363 364 364 367 370 372 375 380 380 382 382 384
5.1 Introduction 5.2 Factors Affecting the Grain Boundary Mobility 5.2.1 Characterisation of grain structures 5.2.2 Driving pressure for grain growth 5.2.3 Drag from impurity elements in solid solution 5.2.4 Drag from a random particle distribution 5.2.5 Combined effect of impurities and particles 5.3 Analytical Modelling of Normal Grain Growth 5.3.1 Limiting grain size 5.3.2 Grain boundary mobility 5.3.3 Grain growth mechanisms 5.3.3.1 Generic grain growth model 5.3.3.2 Grain growth in the absence of pinning precipitates 5.3.3.3 Grain growth in the presence of stable precipitates 5.3.3.4 Grain growth in the presence of growing precipitates 5.3.3.5 Grain growth in the presence of dissolving precipitates 5.4 Grain Growth Diagrams for Steel Welding 5.4.1 Construction of diagrams 5.4.1.1 Heat flow models 5.4.1.2 Grain growth model 5.4.1.3 Calibration procedure 5.4.1.4 Axes and features of diagrams 5.4.2 Case studies 5.4.2.1 Titanium-microalloyed steels 5.4.2.2 Niobium-microalloyed steels 5.4.2.3 C-Mn steel weld metals 5.4.2.4 Cr-Mo low-alloy steels 5.4.2.5 Type 316 austenitic stainless steels 5.5 Computer Simulation of Grain Growth 5.5.1 Grain growth in the presence of a temperature gradient 5.5.2 Free surface effects References Appendix 5.1: Nomenclature
387
387 387 387 389 389 390 392 396 396 397 400 400
CONTENTS
xi
402 402 403 404 404 405 406 406 406 408 408 408 422 427 428 428 429 430 432 444 444 444 447 448 448 448 453 456 456 458 459 459 461 464 467 471 475
6.2.4.2 Site saturation 6.2.5 Non-isothermal transformations 6.2.5.1 The principles of additivity 6.2.5.2 Isokinetic reactions 6.2.5.3 Additivity in relation to the Avrami equation 6.2.5.4 Non-additive reactions 6.3 High Strength Low-Alloy Steels 6.3.1 Classification of microstructures 6.3.2 Currently used nomenclature 6.3.3 Grain boundary ferrite 6.3.3.1 Crystallography of grain boundary ferrite 6.3.3.2 Nucleation of grain boundary ferrite 6.3.3.3 Growth of grain boundary ferrite 6.3.4 Widmanstatten ferrite 6.3.5 Acicular ferrite in steel weld deposits 6.3.5.1 Crystallography of acicular ferrite 6.3.5.2 Texture components of acicular ferrite 6.3.5.3 Nature of acicular ferrite 6.3.5.4 Nucleation and growth of acicular ferrite 6.3.6 Acicular ferrite in wrought steels 6.3.7 Bainite 6.3.7.1 Upper bainite 6.3.7.2 Lower bainite 6.3.8 Martensite 6.3.8.1 Lath martensite 6.3.8.2 Plate (twinned) martensite 6.4 Austenitic Stainless Steels 6.4.1 Kinetics of chromium carbide formation 6.4.2 Area of weld decay 6.5 Al-Mg-Si Alloys 6.5.1 Quench-sensitivity in relation to welding 6.5.1.1 Conditions for ~'(Mg2Si) precipitation during cooling 6.5.1.2 Strength recovery during natural ageing 6.5.2 Subgrain evolution during continuous drive friction welding References Appendix 6.1: Nomenclature Appendix 6.2: Additivity in relation to the Avrami Equation
477
477 477 477 478 480 485 486 491 494
xii
CONTENTS
7.2.2.1 HAZ hardness and strength level 7.2.2.2 Tempering of the heat affected zone 7.2.2.3 HAZ toughness 7.2.3 Hydrogen cracking 7.2.3.1 Mechanisms of hydrogen cracking 7.2.3.2 Solubility of hydrogen in steel 7.2.3.3 Diffusivity of hydrogen in steel 7.2.3.4 Diffusion of hydrogen in welds 7.2.3.5 Factors affecting the HAZ cracking resistance 7.2.4 H2S stress corrosion cracking 7.2.4.1 Threshold stress for cracking 7.2.4.2 Prediction of HAZ cracking resistance 7.3 Stainless Steel Weldments 7.3.1 HAZ corrosion resistance 7.3.2 HAZ strength level 7.3.3 HAZ toughness 7.3.4 Solidification cracking 7.4 Aluminium Weldments 7.4.1 Solidification cracking 7.4.2 Hot cracking 7.4.2.1 Constitutionalliquation in binary AI-Si alloys .7 .4.2.2 Constitutionalliquation in ternary AI-Mg-Si alloys 7.4.2.3 Factors affecting the hot cracking susceptibility 7.4.3 HAZ microstructure and strength evolution during fusion welding 7.4.3.1 Effects of reheating on weld properties 7.4.3.2 Strengthening mechanisms in AI-Mg-Si alloys 7.4.3.3 Constitutive equations 7.4.3.4 Predictions of HAZ hardness and strength distribution 7.4.4 HAZ microstructure and strength 'evolution during friction welding 7.4.4.1 Heat generation in friction welding 7.4.4.2 Response of AI-Mg-Si alloys and AI-SiC MMCs to friction welding 7.4.4.3 Constitutive equations : 7.4.4.4 Coupling of models 7.4.4.5 Prediction of the HAZ hardness distribution References Appendix 7.1: Nomenclature
<,
495 500 502 509 509 513 514 514 518 524 524 525 527 527 529 530 532 536 536 540 541 542 544 547 547 548 548 550 556 556 557 558 558 560 564 567
Besides correcting some minor linguistic and print errors, I have in the second edition included a collection of different exercise problems which have been used in the training of students at NTNU. They illustrate how the models described in the previous chapters can be used to solve practical problems of more interdisciplinary nature. 'Each of them contains a 'problem description' and some background information on materials and welding conditions. The exercises are designed to illuminate the microstructural connections throughout theweld thermal cycle and show how the properties achieved depend on the operating conditions applied. Solutions to the problems are also presented. These are not complete or exhaustive, but are just meant as an aid to the reader to develop the ideas further. Trondheim, 28 October, 1996 (/)ystein Grong
The purpose of this textbook is to present a broad overview on the fundamentals of welding metallurgy to graduate students, investigators and engineers who already have a good background in physical metallurgy and materials science. However, in contrast to previous textbooks covering the same field, the present book takes a more direct theoretical approach to welding metallurgy based on a synthesis of knowledge from diverse disciplines. The motivation for this work has largely been provided by the need for improved physical models for process optimalisation and microstructure control in the light of the recent advances that have taken lplace within the field of materials processing and alloy design. The present textbook describes a novel approach to the modelling of dynamic processes in welding metallurgy, not previously dealt with. In particular, attempts have been made to rationalise chemical, structural and mechanical changes in weldments in terms of models based on well established concepts from ladle refining, casting, rolling and heat treatment of steels and aluminium alloys. The judicious construction of the constitutive equations makes full use of both dimensionless parameters and calibration techniques to eliminate poorly known kinetic constants. Many of the models presented are thus generic in the sense that they can be generalised to a wide range of materials and processing. To help the reader understand and apply the subjects and models treated, numerous example problems, exercise problems and case studies have been worked out and integrated in the text. These are meant to illustrate the basic physical principles that underline the experimental observations and to provide a way of developing the ideas further. Over the years, I have benefited from interaction and collaboration with numerous people within the scientific community. In particular, I would like to acknowledge the contribution from my father Professor Tor Grong who is partly responsible for my professional upbringing and development as a metallurgist through his positive influence on and interest in my research work. Secondly, I am very grateful to the late Professor Nils Christensen who first introduced me to the fascinating field of welding metallurgy and later taught me the basic principles of scientific work and reasoning. I will also take this opportunity to thank all my friends and colleagues at the Norwegian Institute of Technology (Norway), The Colorado School of Mines (USA), the University of Cambridge (England), and the Universitat der Bundeswehr Hamburg (Germany) whom I have worked with over the past decade. Of this group of people, I would particularly like to mention two names, i.e. our department secretary Mrs. Reidun 0stbye who has helped me to convert my original manuscript into a readable text and Mr. Roald Skjarve who is responsible for all line-drawings in this textbook. Their contributions are gratefully acknowledged. Trondheim, 1December, 1993 (lJystein Grong
1.1 Introduction
Welding metallurgy is concerned with the application of well-known metallurgical principles for assessment of chemical and physical reactions occurring during welding. On purely practical grounds it is nevertheless convenient to consider welding metallurgy as a profession of its own because of the characteristic non-isothermal nature of the process. In welding the reactions are forced to take place within seconds in a small volume of metal where the thermal conditions are highly different from those prevailing in production, refining and fabrication of metals and alloys. For example, steel welding is characterised by: High peak temperatures, up to several thousand C. High temperature gradients, locally of the order of 103oC mm'. Rapid temperature fluctuations, locally of the order of 103oC s'. It follows that a quantitative analysis of metallurgical reactions in welding requires detailed information about the weld thermal history. From a practical point of view the analytical approach to the solution of heat flow problems in welding is preferable, since this makes it possible to derive relatively simple equations which provide the required background for an understanding of the temperature-time pattern. However, because of the complexity of the heat flow phenomena, it is always necessary to check the validity of such predictions against more reliable data obtained from numerical calculations and in situ thermocouple measurements. Although the analytical models suffer from a number of simplifying assumptions, it is obvious that these solutions in many cases are sufficiently accurate to provide at least a qualitative description of the weld thermal programme. An important aspect of the present treatment is the use of different dimensionless groups for a general outline of the temperature distribution in welding. Although this practice involves several problems, it is a convenient way to reduce the total number of variables to an acceptable level and hence, condense general information about the weld thermal programme into two-dimensional (2-D) maps or diagrams. Consequently, readers who are unfamiliar with the concept should accept the challenge and try to overcome the barrier associated with the use of such dimensionless groups in heat flow analyses.
Since heat losses from free surfaces by radiation and convection are usually negligible in welding, the temperature distribution can generally be obtained from the fundamental differential equations for heat conduction in solids. For uniaxial heat conduction, the governing equation can be written as.'
(1-1) where T is the temperature, t is the time, x is the heat flow direction, and a is the thermal diffusivity. The thermal diffusivity is related to the thermal conductivity A. and the volume heat capacity pc through the following equation:
a = A./pc
(1-2)
1
aT = a[a T2 + a T]
2 2
at
ax
ay2
(1-3)
and
(1-4)
The above equations must clearly be satisfied by all solutions of heat conduction problems, but for a given set of initial and boundary conditions there will be one and only one solution.
10
Carbon steel 8
500
Temperature,OC
1000
...
To
1500 =
Table 1.1 Physical properties for some metals and alloys. Data from Refs 2-6.
X. Material Carbon Steels Low Alloy Steels High Alloy Steels Titanium Alloys Aluminium (> 990/0 AI) AI-Mg-Si Alloys AI-Mg Alloys
t
a
(rnm? s ')
(W mrrr+C:') 0.040
(OC) 1520
r:
0.025
0.005
1520
7.50 7.40
2.0
0.020
0.005
1500
2.0 1.4
0.030
10
0.003
1650
4.89 1.73
0.230
85
0.0027
660
0.8
0.167
62
0.0027
652
1.71
0.8
0.149
55
0.0027
650
1.70
0.8
4
(a)
0.08..--------r--------r-----.,..-----.,
o o
~ 0.04
E
0.06
3:
~
_A.!!0t.s~e~
0.02'
200
400
600
Temperature,OC ----
800
1000
(b)
t
o o
E
EO.02
~
0.01
200
400
Temperature,
c -----
600
800
1000
Fig. 1.2. Factors affecting the thermal conductivity ~ of steels; (a) Temperature composition, (b) Heat treatment procedure. Data from Refs. 2-4.
= for t = 0 and x = 0 (alternatively r = 0 or R = 0) = 0 for t = 0 and x:j:. 0 (alternatively r> 0 or R > 0) = 0 for 0 < t < when x = (alternatively r = or R =
00 00 00 00
00).
It is easy to verify that the following solutions satisfy both the basic differential heat flow equations (1-1), (1-3) and (1-4) and the initial and boundary conditions listed above: (i) Plane source in a long rod (Fig. 1.3a):
T - To = QIA pe( 4nat)
112 exp( -x 2
14at)
(1-5)
where Q is the net heat input (energy) released at time t rod. (ii) Line source in a wide plate (Fig. 1.3b):
T-To= Qld exp(-r2/4at) pe( 4nat)
= 0, and A
(1-6)
where d is the plate thickness. (iii) Point source in a heavy slab (Fig. 1.3c):
T - To
Q
pe( 4nat)
3/2
exp( -R
/4at)
(1-7)
Equations (1-5), (1-6) and (1-7) provide the required basis for a comprehensive theoretical treatment of heat flow phenomena in welding. These solutions can either be applied directly or be used in an integral or differential form. In the next sections a few examples will be given to illustrate the direct application of the instantaneous heat source concept to problems related to welding. (a)
6 (b)
o
z
(c)
Fig. 1.3. Schematic representation of instantaneous heat source models (continued); (b) Line source in a wide plate, (c) Point source in a heavy slab.
exp( -R
14at)
(1-8)
In order to obtain a general survey of the thermal programme, it is convenient to write equation (1-8) in a dimensionless form. The following parameters are defined for this purpose: - Dimensionless temperature:
8= (T-To) (Tc-To)
(1-9)
where
ti
(1-11)
where qo is the net arc power (equal to Qlti), and (Hc-Ho) is the heat content per unit volume at the reference temperature. Dimensionless radius vector: (1-12) By substituting these parameters into equation (1-8), we obtain:
(1-13)
Isotherms
Fig. 1.4. Instantaneous point source model for assessment of temperatures in arc strikes.
.......
I c
'1"""'4
0.8 0.6
<D
't}
0.4
0.2
0
0.4
0.6 0.8 1
2
'tt--
4 6 .....
10
Equation (1-13) has been solved numerically for different values of 0"1 and t.. The results are presented graphically in Fig. 1.5. Due to the inherent assumption of instantaneous release of heat in a point, it is not possible to use equation (1-13) down to very small values of 0" 1 and r l' However, at some distance from the heat source and after a time not much shorter than the real (assumed) time of heating, the calculated temperature-time pattern will be reasonably correct. Note that the heavy broken line in Fig. 1.5 represents the locus of the peak temperatures. This locus is obtained by setting aln(S/nl)/a'tl = 0:
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
from which
(1-14)
Consider a small weld crater formed in an arc strike on a thick plate of low alloy steel. Calculate the cooling time from 800 to 500C (~t8/5)' and the total width of the fully transformed region adjacent to the fusion boundary. The operational conditions are as follows:
where 1') is the arc efficiency factor. Relevant thermal data for low alloy steel are given in Table 1.1.
Solution
In the present case it is convenient to use the melting point of the steel as a reference temperature (i.e. 9 = em = 1 when T; = Tm). The corresponding values of nl and e (at 800 and 500C, respectively) are:
n =
I
0.75x80x35 4x7.5x(51t)3/2(0.1)1I2
=356 .
800
= (800 - 20)
0.52
Cooling time L1.tS/5 Since the cooling curves in Fig. 1.5 are virtually parallel at temperatures below 800 C, ~t8/5 will be independent of 0"1 and similar to that calculated for the centre-line (UI = 0). By rearranging equation (1-13) we get:
10
METALLURGICAL
MODELLING
OF WELDING
L1'tl
[( ~ 8 J
500
2/3
- ~ 8800
J2/3]
= (3.56) 0.32
213
_(3.56) 0.52
2/3]
=1.38
and
Total width offully transformed region Zone widths can generally be calculated from equation (1-14), as illustrated in Fig. 1.6. Taking the AC3-temperature equal to 890C for this particular steel, we obtain:
L1O'lm
(2e / 3)
1/2
[(
1I3 (J1I3]
e890 J
- ~
em
and M1m
=L10'Im~4ati
=0.23~4x5xO.1mm=0.32mm
Alternatively, the same information could have been read from Fig. 1.5. Although it is difficult to check the accuracy of these predictions, the calculated values for dt8/5 and M 1m are considered reasonably correct. Thus, because the cooling rate is very large, in arc strikes a hard martensitic microstructure would be expected to form within the transformed parts of the HAZ, in agreement with general experience.
The model considers a line source which penetrates two overlapping plates of similar thermal properties, as illustrated in Fig. 1.7. The heat is assumed to be released instantaneously at time
11
~----r--'--~----Iro...
~~.--...,...J
-f
d
t = O. If transfer of heat into the electrodes is neglected, the temperature distribution is given by equation (1-6). ' This equation can be written in a dimensionless form by introducing the following group of parameters: Dimensionless time: (1-15)
where
th
is the heating time (i.e. the duration of the pulse). Dimensionless operating parameter:
s; / dt
where d, is the total thickness of the joint. Dimensionless radius vector:
( 1-16)
( 1-18)
where
e denotes
12 1.4
1.2
(J)
I
N
1.0
0.8
0.6
0.4
0.2
0.4
0.6 0.8
10
20
Figure 1.8 shows a graphical representation of equation (1-18) for a limited range of (T2 and A closer inspection of the graph reveals that the temperature-time pattern in spot welding is similar to that observed during arc ignition (see Fig. 1.5). The locus of the peak temperatures in Fig. 1.8 is obtained by setting dln(Oln2)/d't2 = O.
't2'
Example (1.2)
Consider spot welding of 2 mm plates of low alloy steel under the following operational conditions:
1= 8kA,
th
= 0.3s,
'l1
= 0.5,
To
= 20C
13
Calculate the cooling time from 800 to 500C (L\tS/5) in the centre of the weld, and the cooling rate (C.R.) at the onset of the austenite to ferrite transformation. Assume in these calculations that the total voltage drop between the electrodes is 1.6 V. The Ms-temperature of the steel is taken equal to 475C.
Solution
If we use the melting point of the steel as a reference temperature, the parameters n2 and 800 and 500C, respectively) become:
n2 =
e (at
3.40
soo
500
= 0),
(_1_-_1_J
9500 9S00
= 3.40(_1 0.32
and
..1tS/5 Cooling rate at 475C
= ..1't2 X
th
= 4.09
0.3s
1.23s
The cooling rate at a specific temperature is obtained by differentiation of equation (1-18) with respect to time. When 0"2 = 0 the cooling rate at e = 0.3 (475C) becomes:
-de d't2
=~
('t2)
= ~ = (0.30)2 = 0.026
n2 3.4
and C.R.=
(Tc - To)
th
1 ~
130Cs-1
Since the cooling curves in Fig. I.S are virtually parallel at temperatures below SOOC(i.e. for 8/n2 < 0.15), the computed values of L\tS/5 and C.R. are also valid for positions outside the weld centre-line. In the present example the centre-line solutions can be applied down to (0"2m)2 ~ 2. According to equation (1-19), this corresponds to a lower peak temperature of:
9p = n2
e(0'2m)
2
= 3.40 =0.63
2e
14
It should be emphasised that the present heat flow model represents a crude oversimplification of the spot welding process. In a real welding situation, most of the heat is generated at the interface between the two plates because of the large contact resistance. This gives rise to the development of an elliptical weld nugget inside the joint as shown in Fig. 1.9. Moreover, since the model neglects transfer of heat into the electrodes, the mode of heat flow will be mixed and not truly two-dimensional as assumed above. Consequently, equation (1-18) cannot be applied for reliable predictions of isothermal contours and zone widths. Nevertheless, the model may provide useful information about the cooling conditions during spot welding if the efficiency factor 11 and the voltage drop between the electrodes can be estimated with a reasonable degree of accuracy. A more refined heat flow model for spot welding is presented in Appendix 1.2.
In thermit welding the time interval between the ignition of the powder mixture and the completion of the reduction process will be short because of the high reaction rates involved. Assume that a groove of width 2Ll is filled instantaneously at time t = a by liquid metal of an initial temperature T, (see Fig. 1.10). The metal temperature outside the fusion zone is To. If heat losses to the surroundings are neglected, the problem can be treated as uniaxial conduction where the heat source (extending from -Ll to +L1) is represented by a series of elementary sources, each with a heat content of: (1-20) At time tthis source produces a small rise of temperature at position x, given by equation (1-5):
dT = ( dQ / All 2 JexP[ -(x - x,)2 /4at] pe( 4nat)
(1-21)
=
exp[ -(x
- x')
/4at]
The final temperature distribution is obtained by substituting u = (x-x)/( 4at) 112 (i.e. dx' = - due 4at) 112) into equation (1-21) and integrating between the limits x' = -Ll and x' = +L1 This gives (after some manipulation): (1-22)
15
TTT1~~~~
0)
+ .
~
3
0')
3 3
~~.........:=iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii~J-.J.....J...
Fig. 1.9. Calculated peak temperature contours in spot welding of steel plates (numerical solution). Operational conditions: 1= 23kA, 64 cycles. Data from Bently et al.7
t= 0
.. x
Fig. 1.10. Idealised heat flow model for thermit welding of rails.
where erf(u) is the Gaussian error function. The error function is defined in Appendix 1.3*. Because of the complex nature of equation (1-22), it is convenient to present the different solutions in a dimensionless form by introducing the following groups of parameters:
*The error function is available in tables. However, in numerical calculations it is more convenient to use the Fortran subroutine given in Appendix 1.3.
16 -
Dimensionless temperature:
e'=
Dimensionless time:
cr, -To)
(T-To)
(1-23)
(1-24)
Substituting these parameters into equation (1-22) gives: (1-26) Equation (1-26) has been solved numerically for different values of 0 and 't3. The results are presented graphically in Fig. 1.11. As would be expected, the fusion zone itself (0 :::; 1) cools in a monotonic manner, while the temperature in positions outside the fusion boundary (0) 1) will pass through a maximum before cooling. The locus of the HAZ peak temperatures in Fig. 1.11 is defined by ae'la't3 = O. Referring to Appendix 1.3, we may write:
The peak temperature distribution is obtained by solving equation (1-27) for different combinations of Om and 't3m and inserting the roots into equation (1-26).
Example (1.3)
Consider thermit welding of steel rails (i.e. reduction of Fe203 with Al powder) under the following operational conditions:
Calculate the cooling time from 800 to 500C in the centre of the weld, and the total width of the fully transformed region adjacent to the fusion boundary. The AC3-temperature of the steel is taken equal to 890C.
17
Definition of parameters:
e=
l
0.8
t =
V4at
L1
L1
.Q=~
0.4
0.2
o ~~~~ o
__ ~
2
-L
__ ~
= 0) equation
Cooling time ~t8/5 From the above relation it is possible to calculate the cooling time from T; = 2200C to 800 and 500C, respectively:
, = elf ( - 1 9S00
't3
= (800 - 20)
(2200 - 20)
and
, = elf ( 1 8500
"""""*
't3
J=
~tS/5:
~t8/5
* =-[('t3) 4a
LI
2 -('t3)2] =-(6)2
. 4x5
[(5.05)
2 -(3.04)
2] s=29s
18
The computed value for fltS/5 is also valid for positions outside the weld centre-line, since the cooling curves at such low temperatures are reasonably parallel within the fusion zone. Total width of fully transformed region The fusion boundary is defined by:
The locus of the 890C isotherm in temperature-time Taking the ordinate equal to 0.40, we get:
't3m ~
1.50
By inserting this value into equation (1-27), we obtain the corresponding coordinate of the isotherm:
= LI
(Q~-Qm)
= 6 (1.24 -
l)mm
= 1.44mm
Unfortunately, measurements are not available to check the accuracy of these predictions. Systematic errors would be expected, however, because of the assumption of instantaneous release of heat immediately after powder ignition and the neglect of heat losses to the surroundings. Nevertheless, the present example is a good illustration of the versatility of the concept of instantaneous heat sources, since these solutions can easily be added in space as shown here or in time for continuous heat sources (to be discussed below).
The model considers a continuous (plane) heat source in a long rod as shown in Fig. 1.12(a). The heat is liberated at a constant rate q'o in the plane x = 0 starting at time t = O. If we subdivide the time t during which the source operates into a series of infinitesimal elements dt' (Fig. 1.12b), each element will have a heat content of:
dQ=q~dt'
(1-28)
19
(a)
Continuous heat source
-x
+x
(b)
q I~ ql a
I dQ=q~dtl~
I I I I I
I I
t'
dtl
til
=t -
tl
.1 I
I
Fig. 1.12. Idealised heat flow model for friction welding of rods; (a) Sketch of model, (b) Subdivision of time into a series of infinitesimal elements dt~
At time t this heat will cause a small rise of temperature in the material, in correspondance with equation (1-5):
dT = q~ dt' / A x ] exp [ pc~ 41ta(t - t') '4a(t - t')
2
(1-29)
If we substitute t" = t - t'into equation (1-29), the total temperature rise at time t is obtained by integrating from t" = t (t' = 0) to t" = 0 (t' = t):
-q~ / A pc,) 41ta
J
o
{?exp
dt"
[x2]
- 4a;"
(1-30)
In order to evaluate this integral, we will make use of the following mathematical transformation:
II =
wdp=wp-
pdw
20
where
w=2ex _X2 ] dw= [X2] -p[ 4at" , 2aCt,,)2 2 _X ] dt" ex [ -p 4at"
and
p =..Jt",dp
Hence, we may write:
=--
2..Jt"
dt"
I[ = ..Jt" exp[ t
2[
4::" ]]- J
t
2a(;/~)3/2
The latter integral can be expressed in terms of the complementary error function" erfc(u) by substituting:
u=_x_ du= -x dt" ~ 4at" , 4-ra (t,,)3/2
and integrating between the limits u = x I (4at)1/2 and u = 00. This gives (after some manipulation):
T-T
Apc-[iW
q~
{i (exp [x2] --
4at
-J4ii
(1-31)
If the temperature of the contact section at the end of the heating period is taken equal to Th, equation (1-31) can be rewritten as:
T - To =
cr, - To
(1-32)
where tlz denotes the duration of the heating period (t ~ tlz). Measured contact section temperatures for different metal/alloy combinations are given in Table 1.2. Equation (1-32) may be presented in a dimensionless form by the use of the following groups of parameters: Dimensionless temperature:
8"=
r, -r;
t / tlz
T-To
(1-33)
Dimensionless time:
't4 =
(1-34)
*The complementary
21
Table 1.2 Measured contact section temperatures during friction welding of some metals and alloys. Data from Tensi et al.lO
MetaUAlloy Combination Steel Steel-Nickel Steel- Titanium Copper-Al Copper-Nickel AI-Cu-2Mg AI-4.3Cu AI-12Si AI-5Mg
1Based
Measuring Method Thermocouples Direct readings Direct readings Direct readings Direct readings Thermocouples Thermocouples Thermocouples Thermocouples
1 1 1 1
Temperature Level [OC] 1080-1340 1260-1400 1080 548 1083 506 562 575 582
Dimensionless x-coordinate:
Q'=_X_
(1-35)
~4ath
-~(Q')2] - (
[il']J
(1-36)
Equation (1-36) describes the temperature in different positions from the weld contact section during the heating period. However, when the rotation stops, the weld will be subjected to free cooling, since there is no generation of heat at the interface. As shown in Fig. 1.13(a) this can be accounted for by introducing an imaginary heat source of power +q' 0 at time t = liz which acts simultaneously with an imaginary heat sink of negative power -q' o- It follows from the principles of superposition (see Fig. 1.13b) that the temperature during the cooling period is given by:" (1-37) where 8"("(4) and 8"( 't4 - 1) are the temperatures calculated for the heat source and the heat sink, respectively, using equation (1-36). Equations (1-36) and (1-37) have been solved numerically for different values of 0.' and 't4. The results are presented graphically in Fig. 1.14. Considering the contact section (0.'= 0), the temperature increases monotonically with time during the heating period, in correspondance with the relationship: (1-38)
22
(a)
(b)
t:
'-'"
.-... <0
He~ting penod
Cooling period
Fig. 1.13. Method for calculation of transient temperatures during friction welding; (a) Sketch of imaginary heat source/heat sink model, (b) Principles of superposition.
Similarly, for the cooling period we get: (1-39) Outside the contact section (Q"> 0), the temperature rise will be smaller andthe cooling rate lower than that calculated from equations (1-38) and (1-39).
23
1.0
Cooling
0.8
1
-<D
0.6
0.4
0.2
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Example (1.4)
Calculate the peak temperature distribution across the joint. Assume in these calculations that the thermal diffusivity of the AI-Cu-2Mg alloy is 70mm2 s'. Solution Readings from Fig. 1.14 give:
x=O x=6mm x= 12mm x=24mm x=36mm
(n~
(Q~
= 0)
=0.1)
(n~
e;: = 1, i.e. Tp = 506C e;: = 0.83, i.e. Tp = 423C e;: = 0.69, i.e. Tp = 355C e;: = 0.48, i.e. Tp = 253C e;: = 0.35, i.e. Tp = 190C
24
In this particular case, it is possible to check the accuracy of the calculations against in situ thermocouple measurements carried out on friction welded components made under similar conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling cycles cannot be reproduced with the same degree of precision. This has to do with the fact that the present analytical solution omits a consideration of the plastic straining occurring during friction welding, which displaces the coordinates and alters the heat balance for the system.
25
- Flux cored arc (FCA) welding. - Submerged arc (SA) welding. The main purpose of this section is to review the classical models for the pseudo-steady state temperature distribution around moving heat sources. The analytical solutions to the differential heat flow equations under conditions applicable to arc welding were first presented (a)
800
C,)
0
I
E
a.
1..I
PI :::J ~ CD ~
400
200
8
Time,s ~
10
12
14
16
18
(b)
600
o
0
1
Observed relationship
Q)
~ ~
~- 500 ::J
a.
E $
Q)
a..
<0
400
Predicted relationship /
..0
.
....
300~------~--------L-------~--------~------~ 10 246 8 o
Distance from contact section, mm Fig. I.l S, Comparison between measured and predicted temperatures in friction welding of AI-Cu-2Mg
alloys; (a) Temperature-time pattern, (b) Peak temperature distribution. Data from Tensi et al.lO
26
by Rosenthal.P'!" but the theory has later been extended and refined by a number of other investigators.9,11,15-20 1.10.1 Arc efficiency factors In arc welding heat losses by convection and radiation are taken into account by the efficiency factor 'Yl, defined as: (1-40) where qo is the net power received by the weldment (e.g. measured by calorimetry), I is the welding current (amperage), and U is the arc voltage. For submerged arc (SA) welding the efficiency factor ('Yl) has been reported in the range from 90 to 98%, for SMA and GMA welding from 65 to 85%, and for GTA welding from 22 to 75%, depending on polarity and materials. 11 A summary of ranges is given in Table 1.3. 1.10.2 Thick plate solutions
Model (after Rykalin9)
According to Fig. 1.17, the general thick plate model consists of an isotropic semi-infinite body at initial temperature To limited in one direction by a plane that is impermeable to heat. At time t = 0 a point source of constant power q 0 starts on the surface at position 0 moving in the positive x-direction at a constant speed v. The rise of temperature T - To in point P at time t is sought. During a very short time interval from z'to t' + dt'the amount of heat released at the surface is dQ = q.dt'. According to equation (1-7) this will produce an infinitesimal rise of temperature in P at time t:
dT= 2q 0 dt'
exp [(R )2
4a(t - t')
(1-41)
-2qo dt"
= pc( 41tat,,)3/2
exp -
[(R )2
1
4at"
where
for conduction
topointP. For a convenient presentation of the pseudo-steady state solution, the position P should be referred to that of the moving heat source. This is achieved by changing the coordinate system from 0 to 0' (see Fig. 1.17):
R'=~(XO-1.)t')2+y;+Z;
and
y=Yo'
Z=Zo'
x=xo-vt
27
x=vt
Fig. 1.16. Schematic diagram showing weld centre-line temperature in different positions from the heat source during steel welding at pseudo-steady state.
arc efficiency factors for different welding processes. Data from Refs 11, 21.
Arc efficiency factor 11
Range 0.91-0.99
Mean 0.95
0.66-0.85
0.80
0.75-0.93
0.85
0.66-0.70
0.70
0.25-0.75
0.40
0.55-0.80
0.60
0.22-0.46
0.40
28
METALLURGICAL
MODELLING
OF WELDING
=
where R=~x2+y2+z2.
(1-42)
into equation (1-42), and integrating between the limits u (after some manipulation): T-To
(R2/4at")112 and u
00.
This gives
2!M(
2
In)exP(-uX,2a)f
u
exp(-u2-m2
/u2)du
(1-43)
It is well-known that:
OOJ
exp(-u
-m
lu )du=-" -exp(-2m)=-exp(-uRI2a)
j1t
2
j1t
2
Hence, the general thick plate solution can be written as: T-T
o
=.!lsL.(~)eXp(-UXI2a) 21tA R
x[
exp( _u
m2 / u2 )du
(1-44)
If u is sufficiently small (i.e. when welding has been performed over a sufficient period), we obtain the pseudo-steady state temperature distribution:
(1-45)
This equation is often referred to as the Rosenthal thick plate solution, 13, 14 in honour of D. Rosenthal who first derived the relation by solving the differential heat flow equation directly for the appropriate boundary conditions. 1.10.2.1 Transient heating period It follows from the above analysis that the pseudo-steady
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
29
tz
Fig. 1.17. Moving point source on a semi-infinite slab.
attained after a transient heating period. The duration of this heating period is determined by the integral in equation (1-44). Taking the ratio between the real and the pseudo-steady state temperature equal to K1, we have:
K}
2 cexp
'V1t
( uR )JU exp(-u 2a
o
-m
/u )du
(1-46)
Equation (1-46) can be expressed in terms of the following parameters: Dimensionless radius vector:
0'3=-
uR
2a
(1-47)
K 1 = 1-
J1n
exp( (53)
f
't
(1-49)
30
Equation (1-49) has been solved numerically for a limited range of 0"3 and 'to The results are presented graphically in Fig. 1.18. A closer inspection of Fig. 1.18 reveals that the duration of the transient heating period depends on the dimensionless radius vector 0"3' In practice this means that the Rosenthal equation is not valid during the initial period of welding unless the distance from the heat source to the observation point is very small. It should be noted, however, that a dimensionless distance 0"3 may be 'short' for one combination of welding speed and thermal diffusivity, while the same position may represent a 'long' distance for another combination of v and a. Similarly, the dimensionless time 't may be 'short' or 'long' at a chosen number of seconds, depending on the ratio vl2a.
Example (1.5)
Consider stringer bead deposition on a thick plate of aluminium at a constant welding speed of 5 mm S-I. Calculate the duration of the heating period when the distance from the heat source to the point of observation is 17 mm.
Solution
Taking a = 85 mm?
S-I,
2x85
0.8
t
~
u)
0.6
.........
;:5'
~
II
0.4
0.2
Fig. 1.lS. Ratio between real and pseudo-steady different combinations of 3 and T.
(J'
state temperature
31
It is seen from Fig. 1.18 that the pseudo-steady state temperature distribution is approached when 't = 3, which gives:
2a 2x85 t = 2't = --2-3s ~20s U (5)
L2 =
in = 5
X 20 = 100mm
The above calculations show that the Rosenthal equation is not valid if the ratio between R and L2 exceeds a certain critical value (typically 0.15 to 0.30 for aluminium welds and 0.4 to 0.6 for steel welds). This important point is often overlooked when discussing the relevance of the thick plate solution in arc welding. 1.10.2.2 Pseudo-steady state temperature distribution The Rosenthal equation gives, with the limitations inherent in the assumptions, full information on the thermal conditions for point sources on heavy slabs. Accordingly, in order to obtain a general survey of the pseudo-steady state temperature distribution, it is convenient to present the different solutions in a dimensionless form. The following parameters are defined for this purpose: 11 - Dimensionless operating parameter:
(1-50)
Dimensionless x-coordinate:
~=
Dimensionless y-coordinate:
\If
Ux
2a
(1-51)
= uy 2a
(1-52)
Dimensionless z-coordinate:
~=
Uz
2a
(1-53)
(j3
)exp(
-(j3 -~)
(1-54)
where e and (J 3 are the dimensionless temperature and radius vector, respectively (previously defined in equations (1-9) and (1-47. Equation (1-54) has been solved numerically for chosen values of (J3 and . A graphical presentation of the different solutions is shown in Fig. 1.19. These maps provide a good
32 (a)
1.4 1.2
1.0 0.8
as
0.6
0.4
(t)
Sp/n3
0.2
o~==~====~~==~~~----~~~~..10 -8 6 ~4 -2
Fig. 1.19. Dimensionless lel to the -axis.
temperature maps for point sources on heavy slabs; (a) Vertical sections paral-
overall indication of the thermal conditions during thick plate welding, but are not suitable for precise readings. Consequently, for quantitative analyses, the following set of equations can be used;'!
Isothermal zone widths
= 0:
a~
.~=o
d(J3
fI3
we have:
33
(b)
..10
*=0.01
3
Fig. 1.19. Dimensionless temperature maps for point sources on heavy slabs (continued): (b) Isothermal contours in the -'V-plane for different ranges of eln3.
= __
1_ -1-1.
(J3m ~m
= O,i.e.~m
= -( (J3m)
(j3m
(j3m
+1
34 and
8p = --exp 1 n3
0"3m
[ -0"3m / (0"3m
+ 1)]
(1-55)
Equation (1-55) can be used for calculations of isothermal zone widths 'Vm and cross sectional areas At. From Fig. 1.20 we have:
A graphical presentation of equations (1-55), (1-56), and (1-57) is shown in Fig. 1.21. Length of isothermal enclosures Referring to Fig. 1.20, the total length of an isothermal enclosure
~t
where ~' and ~"are the distance from the heat source to the front and the rear of the enclosure, respectively.
graphical representation
35
A =1L'I'
12m
=A~
4a2
(j3m
v = Rm2a
1<
10
0.1
0.01 0.01
Fig. 1.21. Dimensionless distance
0.1
n3/8p
U3m'
10
n3
100
lep-
half width
'I'm
<J"3
~ in equation
~'=
and
tIn(
n3~~
e)
(1-60) Volume of isothermal enclosures Since the assumption of a point heat source involves semi-circular isotherms in the plane, the volume of an isothermal enclosure is obtained by integration over the total length from ~" to ~':
't'-~
(1-61)
d~=-d0'3
36
r = -~
Noting that
0"3
J
s'
~"
J
s'
~"
~2d~
(1-62)
= - ~"= n3/8
r is related
(1-64) Cooling conditions close to weld centre-line For points located on the weld centre-line behind the heat source ur = Hence, equation (1-54) reduces to": 8
n3
~ = 0, and
0"3
= -~
= 'to
(1-65)
Equation (1-65) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from 81 to 82): (1-66) For welding of low alloy steel the cooling time from 800 to 500 e is widely accepted as an adequate index for the thermal conditions under which the austenite to ferrite transformation takes place. From equation (1-66), we have:
0
~'t8/5
=n3(_I 8
500
8800
I_)
(1-67)
By substituting reasonable average values for a and pc into equation (1-67), we obtain the following expression for I1t8/5 when To = 20oe:
~t8/5 (s) = 5.111E
(1-68)
Here n is the arc efficiency factor, and E (kJ mrrr") is the gross heat input per unit length of the weld, defined as:
time
length coordinate E.
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
37 (1-69)
E=
UxI 1000'\)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation (1-65) with respect to time:
-d(81 n3) _ ~ _ (81 )2 2 n3 d1; 1;
(1-70)
(1-71)
Example (1.6) Consider stringer bead deposition (GMAW) on a thick plate of low alloy steel under the following conditions: 1= 300A, U
= 28V,
Sketch the contours of the fusion boundary and the AC3-isotherm (910C) in the plane at pseudo-steady state.
Solution
~-'l' (x-y)
As shown in Fig. 1.22(a) it is sufficient to calculate the coordinates in four different (characteristic) positions to draw a sketch of the isothermal contours. If we neglect the latent heat of melting, the SIn3 ratio at the melting point becomes, according to equations (1-9) and (1-50):
4nx(5) x7.5 300x28xO.8x4
2
~=
n3
=0.088
End-points The end-points are readily obtained from equations (1-59) and (1-60):
~'=
tln(--l-~-J'i.e.~'
0.088'-,,'
-n3
= 1.15(x = 2.87mm)
and
~"= -
-1 = -= -11.36(x 0.088
= -28.4Imm)
Maximum width The maximum width of an isothermal enclosure can generally be calculated from equations (1-55) and (1-56), or read from Fig. 1.21. Taking Sp In3 = SIn3 = 0.088, we obtain:
38
(a)
Pas. (3)
POSt
(4)
Fig. 1.22. Calculation of isothermal contours from Rosenthal thick plate solution; (a) Calculation procedure, (b) Sketch of fusion boundary and Acrisotherm in position ~ = 0 (Example 1.6).
'I'm and
~m
= 2.75
(ym
(xm
= 6.88mm)
= 4.13
= 10.33mm)
39
Intersection point with lfI (y)-axis In this case = ~ = 0, and (f3 = Hence, equation (1-54) reduces to:
'1'.
1 = -exp(-\IJ)
n3 which gives
'I'
'I' = -In(
83(y
= 4.58mm)
Similarly, the contour of the AC3-isotherm can be determined by inserting e/n3= 0.052 into the same set of equations. Figure 1.22(b) shows a graphical representation of the computed isothermal contours.
Example (1.7)
Consider GTA welding on a thick plate of low alloy steel under the following conditions: 1= 150A, U = 23V,
Calculate the weld pool volume, the weld bead cross section, the width of the fully transformed HAZ, the cooling time from 800 to 500C, and the cooling rate at the onset of the austenite to ferrite transformation (e.g. at 650C). The arc efficiency factor is taken equal to 0.5.
Solution
If we neglect the latent heat of melting, the e/n3 ratio at the melting point becomes:
e
n3
150x23xO.5x3
0.455
Weld pool volume The coordinate "may be calculated from equation (1-59):
~'= tln(-l-~-J'i.e.~'
0.455~'
= 0.627
We can now use equations (1-63) and (1-64) to calculate the weld pool volume. This gives:
r = l~ [3(1/
and
3
= 3.22
v=~r=~x3.22mm3
\)3
5 (3)3
::::::120mm3
40
Weld bead cross section The weld bead cross section can be read from Fig. 1.21. Taking obtain: A I = - 'II m
2
1t 2 =
n3
lap
= 1/0.445 = 2.196,
we
1.72
and
4a2 4x(5)2 2 2 A=-2 Al =--2-x1.72mm :::;20mm U (3)
Width offully transformed HAZ When Tp = 910C, the n3 lap ratio becomes:
8p
= 2.196( 1520-20) =
910-20
3.70
'I'm = 1.45
which gives
f2
= DAD
2x5 =--0.40mm:::;1.3mm 3
Cooling time from 800 to 500C The cooling time, ~tS/5' can be obtained from equation (1-68): 150x 23 8=2.98 1000x3
IltS/5
= 5.1llE=5.1 xO.5
This value is also valid for positions outside the weld centre-line, since the cooling curves at such low temperatures are reasonably parallel down to Tp ~ 910C (see Fig. l.l9(a)). Cooling rate at 650C The cooling rate at a specific temperature is given by equation (1-71). Taking the transformation start temperature equal to 650C, we get:
C.R.(OCs -1)
(2'7TA)
q/v
41
It follows from Fig. 1.19(b) that the isotherms behind the heat source become increasingly elongated as the arc power q() and the welding speed '\J increase. In the limiting case, when q 0 ~ 00, v ~ 00 and q 0 Iv remains finite, the isotherms will degenerate into surfaces which are parallel to the welding x direction as shown in Fig. 1.23(a). Conduction of heat will then occur exclusively in directions normal to the x-axis. In a short time interval dt, the amount of heat released per unit length of the weld is equal to:
dQ dx
= qo dt = qo
vdt v
(1-72)
According to the assumptions this amount of heat will remain in a slice of thickness dx due to the lack of a temperature gradient in the welding direction. Since symmetry demands that the isotherms in the y-z plane are semi-circles", the situation becomes identical to the temperature field around a linear instantaneous heat source in a thin plate, provided that the space above the slab is replaced by solid material and the strength of the source is doubled (see Fig. 1.23(b)). The solution is then given by equation (1-6) if we replace Qld by Zq; Iv:
/ 4at]
= qo / u (!)exp[-(r*)2
21tA.
/4at]
(1-73)
where r* is the two-dimensional radius vector in the y-z plane. Equation (1-73) represents the simplified solution for a fast moving high power source on a semi-infinite slab, and is valid within a limited range of the more general Rosenthal equation for three-dimensional heat flow (equation (1-45)). By introducing the dimensionless radius vector <T4 = vr*/2a, equation (1-73) transforms to:
~ =
n3
(.!.) ex
r
p[- (0'4 )2 ]
2't
(1-74)
A graphical presentation of equation (1-74) gives a family of curves which resembles the thermal cycles shown in Fig. 1.19(a). Although the cooling conditions close to the weld centre-line are similar to those calculated from the Rosenthal equation, the predicted width/depth of the isotherms will always be greater than that inferred from the general thick plate solution as illustrated in Fig. 1.24 due to the assumption of 2-D heat flow. The parameter <T 4m in Fig. 1.24 is obtained by differentiating equation (1-74) with respect to time:
*The isotherms must meet the plate surface at right angles in the absence of a temperature gradient across the boundary.
42 (a)
z
(b)
J...--l...--
Symmetry prane
-..,---------,
I
,,
Fig. 1.23. Fast moving high power source on a semi-infinite slab; (a) Sketch of model, (b) Analogy between a fast moving high power source and an instantaneous line source.
43
6
0'
4m
/'V m = rm'Y.m
0.1
10
100
In
Fig. 1.24. Theoretical width of isotherms under 2-D and 3-D heat flow conditions, respectively at pseudosteady state (thick plate welding).
which gives
(1-75) It is interesting to note that the dimensionless width IT 4m will approach 'I'm when the 8p In3 ratio becomes sufficiently small (i.e. less than about 0.1). Under such conditions the isotherms will be strongly elongated in the welding direction (see Fig. 1.19(b)), which forces the heat to flow primarily in directions normal to the x-axis. A general graphical representation of the weld thermal programme can be obtained by combining equations (1-74) and (1-75):
(1-76)
Equation (1-76) has been plotted in Fig. 1.25. This graph provides a basis for calculating the retention time within specific temperature intervals under various welding conditions.
44
1.0
0.8 0.6
0.4
0.2
4
2t/(0'4m)
2
pattern in thick plate welding at high arc power and high welding speed.
Example (1.8)
Consider SA welding on a thick plate of low alloy steel under the following conditions: I
= 400A,
= 25V, v = 5mm
s-1, To = 20C
Calculate the retention time within the austenite regime (T ~ 910C) for points located 1mm outside the fusion boundary.
Solution
If we use the melting point of the steel as a reference temperature, the parameter = 400x25xO.95x5 41t x (5)2 x 7.5 =20.16
n3
becomes:
n3
8p
A comparison with Fig. 1.24 shows that the assumption of 2-D heat flow is justified when 1. Hence, the width of the fusion zone (Sp = 1) can be calculated from equation (1-75):
which gives
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
45
rm =-(j4m
2a
2xS =--3.8Smm=7.70mm S
8 = 2n3 (
P
:a
u(rm
+1)
J2 = 2X20 16(
0
2xS Sx(8.7)
J2 =0.78
The total time spent in the thermal cycle from 6 = 0.59 (T= 910Ce) to 6p = 0.78 and down again to 6 = 0.59 can now be read from Fig. 1.25. Taking the ordinate 6/6p equal to 0.76, we obtain:
2~'tr
j4m)
2 =
which gives
~'tr = 1.8 j4m) 2
2
and
2a 2xS ~tr=2~'tr=--217.1s~7s U (S)
As shown in Fig. 1.26, the general thin plate model considers a line source in a wide sheet of thickness d and initial temperature To. At time t = a the source starts to move at a constant speed u in the positive x-direction. The rise of temperature T =T; in point P at time t is sought. According to equation (1-6) the elementary source dQ = q 0 dt' released at position t' will cause a small rise of temperature dT in point P at time t:
dT
[(r')2]
4a(t - t')
exp
[(r,)2
_ 4at"
(1-77)
tlf
for conduction
y;
to point P.
If we refer the position P to that of the heat source at time t, we shall expect a solution independent of time at pseudo-steady state. This is achieved by changing the coordinate system from 0 to 0' (see Fig. 1.26):
y
= Yo,
= Xo -
ut,
Xo - ur'
=x
1)t -
ur'
=x
+ uz"
46
2.5
2:
1.5
\...
\ \
\
K1 (u)
=so
....
~ 1.0
0.5
0.5
Fig. 1.27. Graphical representation
1.0
1.5
2.0 .
2.5
3.0
3.5
47
Hence, we may write: dT= -qo dt" [ -[(x exp dpc( 41Tat")
+ v(')2 + y2] ]
4a('
(1-78) where
r=~x2+y2
For integration of all contributions from t " = t (t ' = 0) to t " = 0 (t '= t), we introduce:
(a5)=-z-, 4a
uZr2
W=--
u2t" 4a
d t "-- 4a z d w,
U
.i;.:
4at"
(0"5)2
4w
q0 I d =--exp
4nA
[UX --
WI
2a w=o
4w
Jdw
w
(1-79)
OOf
JdW -=2K
woo
(0"5)-2K
..L
(ur)
-
2a
where Ko( <T5) is the modified Bessel function of the second kind and zero order. A graphical representation of Ko(u) is shown in Fig. 1.27. Hence, the general thin plate solution can be written as: T - T = qo I d exp(o
2nA
Ux)
2a
(1-80)
When w is sufficiently large (i.e. when welding has been performed over a sufficient period), we obtain the pseudo-steady state temperature distribution:
48
T-T
o
q0 I d =--exp 21tA
('OX --
2a
J ( 2a J
K
'Or 0
(1-81)
Equation (1-81) is often referred to as the Rosenthal thin plate solution, in memory of its originator D. Rosenthal.l+!" It follows that this model is applicable to all types of welding processes (including electron beam, plasma arc and laser welding), provided that a full throughthickness penetration is achieved in one pass. 1.10.3.1 Transient heating period In thin plate welding the duration of the transient heating period is determined by the integral in equation (1-80). Taking the ratio between the real and pseudo-steady state temperature equal to K2, we obtain:
(1-82)
where w = 't/2, and dw = dt/2. Equation (1-82) has been solved numerically for a limited range of (J"sand presented graphically in Fig. 1.28.
Example (1.9)
t,
Consider butt welding of a thin aluminium plate at a constant travel speed of 5mm S-I. Calculate the duration of the transient heating period when the distance from the heat source to the point of observation is 17mm.
Solution
Taking a = 85 mm-
S-I,
It follows from Fig. 1.28 that the pseudo-steady state temperature distribution is approached when 't 5, which gives:
Z
t= and
-t
2a '02
2x 85 = --5s (5)2
= 34s
~ = ut = 5 X
34mm
= 170mm
This minimum bead length is nearly twice as large as that calculated for 3-D heat flow for the same combination of welding speed, thermal diffusivity and radius vector (see Example
49
t
c6
0.8 0.6
t:,
l-
,{
"-~
0.4
O's=vr/2a
II
0.2
O~~~~--~~-----L----~------L-----~----~----~
234
't=v2tJ2a
567
Fig. 1.28. Ratio between real and pseudo-steady combinations of (J'sand 'T.
state temperature
(1.5. Consequently, the duration of the transient heating period is significantly longer in thin plate welding than in thick plate welding due the pertinent differences in the heat flow conditions. 1.10.3.2 Pseudo-steady state temperature distribution A graphical presentation of the pseudo-steady state temperature distribution similar to that shown in Fig. 1.19 for three-dimensional heat flow may be obtained by introducing the dimensionless plate thickness 8 = vdl2a, which is a measure of the relative speeds of the arc and the heat flow in the material. By rearranging equation (1-81), we get:
e~ = exp(-~)K
n3
0
(as)
(1-83)
Plots of this equation are shown in Fig. 1.29. It follows that the pseudo-steady state temperature distribution in thin plate welding depends on the parameter 8'&/n3. In practice, this means that the shape of the isotherms is not influenced by the welding speed, since both '& and n3 are proportional to u, Isothermal zone widths The maximum width of an isothermal enclosure is obtained by setting a(88 In3) la~ = O. Noting that (a/au)Ko(u) = -K, (u), where K, (u) is the modified Bessel function of the second kind and first order (shown in Fig. 1.27), we get: a(eo/n3) a~ a(eo/n3) a<J5 a<J5 a~
~m K1J5m)]=O
---=--=
.--
(1-84)
=-exp(-~m)[KoJ5m)+
<Jsm
\
\ \
\
\
\
\
C (t5o.
cP
I
('t)
cf)
GO cD 4-4 0
00
~ .s
0,)
en
t::
c t.oo.
CD
C\I
~ ~ :.a ~ c.8
00
d 0,)
~ ~
u/ge
0
-a
~
t::
Cd
0,)
.S
o
0
00
0,)
00
~ ::l
0,)
~ c.8
00
0..
8 B ~
2
00 00
a
0,)
~ ~
LO
T-
~,
~ co
c
(f)1
0 C\J
,
'(I)
.9
00
"2
t::
0,)
c;oo..
CD
ee 0. c:t>
=-
Q\
0,)
.~
~
~ ~
51
Equation (1-86) can be used for calculations of isothermal zone widths ('I'm) and cross sectional areas (A2) in thin plate welding. Referring to Fig. 1.30, we have:
(1-87) and (1-88) Figure 1.31 shows a graphical presentation of equations (1-86), (1-87), and (1-88). Length of isothermal enclosures The distance from the heat source to the front (g') and the rear (g') of an isothermal enclosure is obtained by substituting (J".5 = g into equation (1-83). This gives:
~' = In(
and
n3K;i~')}
~'> 0
(1-89)
~"=
In( n3K~~~")}
~"< 0
(1-90)
52
A2
= 2B'I'm = A v2/4a2
O'Sm
= rm v/2a
= Ym v/2a
'I'm 10
t
b
E io
0.1
~~~~~~~--~~~~~ 100
1000
distance
0"5m'
half width
"'m
100
uP
10
~ ~
0.1
;t=~I_~1I
0.0 1 ~"""""-.a...&"""""'Lo....--~....a.-""'&'~""""""""l""'O
---~...---1"---'.
&....&..1.0
.... 1 00
---1
n3/90
Fig. 1.32. Dimensionless distance from heat source to front (~') and rear (~") of isothermal enclosure vs ni9S (thin plate welding).
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
53
A graphical presentation of equations (1-89) and (1-90) is shown in Fig. 1.32. Included is also a plot of the total length of the enclosure ~t vs the parameter n3 18B. Cooling conditions close to weld centre-line For points located on the weld centre-line behind the heat source (J5 = -~ = 'to When r is larger than about 1, (i.e. t> 2alrr), it is a fair approximation to set Ko ('t) exp( -'t)~1t / 2't (see Fig. 1.27). Hence, equation (1-83) reduces to:
Z
-~-y1t/2't
n3
8B
(1-91)
Equation (1-91) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from 81 to 92):
(1-92)
~'t
8/5
= -1t (n 2 3
I u)
~ 2[
1
(8500)
(1-93)
(500~ To)2
(1-94)
Taking as average values A = 0.025 W mm' e-1, pc for welding of low alloy steels, we have:
= 0.005
J mm='
e-1,
and To
= 20 e
0
(1-95) Similarly, the cooling rate at a specific temperature is obtained by differentiating equation (1-91) with respect to time: (1-96)
54
(1-97)
(T _ T )3
0
(1-98)
Example (1.10)
Consider GTA butt welding of a 2mm thick sheet of cold-rolled aluminium (AI-Mg alloy) under the following conditions: 1= 110A, U=15V; v=4mms-l, 'Yl=0.6, To=20C
Sketch the contours of the fusion boundary and the Ar-isotherm in the ~-'" (x-y) plane at pseudo-steady state. The recrystallisation temperature Ar of the base material is taken equal to 275C. Calculate also the cross sectional area of the fully recrystallised HAZ and the cooling rate at 275C for points located within this region. Solution Referring to Fig. 1.22(a) (Example (1.6 it is sufficient to calculate the coordinates in four different (characteristic) positions to sketch the contour of the fusion boundary. If we neglect the latent heat of melting, the n3/8B ratio at the melting point becomes:
!!leo - 2n'
qo
a d (H m
-
= H 0)
End-points The end-points can be read from Fig. 1.32: ~' = 0.25 (x = 6.88mm) and ~" = - 0.90 (x = -24.75mm) Maximum widths The maximum width of an isothermal enclosure can generally be calculated from equations (1-86) and (1-87) or read from Fig. 1.31. When n3/8pB = 0.84, we obtain: "'m = 0.41 (ym = 11.41mm) and
gm
-0.29 (xm
-7.87mm)
55
Intersection
= '1'.
which gives
'I' = 0.37 (y
IO.I8mm)
Similarly, the contour of the Ar-isotherm can be determined by inserting n3/08 = 2.08 into the same set of equations. Figure 1.33 shows a graphical representation of the calculated isothermal contours.
Fig. 1.33. Calculated contours of fusion boundary and Ar-isotherm in GTA butt welding of a 2mm thick aluminium plate (Example 1.10).
56
Cross sectional area of fully recrystallised HAZ In general, cross sectional areas can be read from Fig. 1.31. Taking the n3/8p8 ratio equal to 0.84 (8p = 1) and 2.08 (ap = 0.48), respectively, we have: 4x2 M.2 =8(3-0.8)=--2.2=0.16 2x55 which gives 4(55)2 M=0.16--mm2 (4)2 =121mm2
Cooling rate at 275C The cooling rate at a specific temperature can be calculated from equation (1-97). In the present case, we obtain: c.R.rCs-I)= 21txO.149xO.0027
(110x15xO.6/4x2)2
(275-20)3
Cs-I ~3Cs-I
It follows from Fig. 1.29 that the isotherms behind the heat source become increasingly elongated as the 881n3 ratio decreases. In the limiting case the isotherms will degenerate into surfaces which are parallel to the welding x direction, as shown in Fig. 1.34. In a short time interval dt the amount of heat released per unit length of the weld is equal to: dQ = qo dt dA ddx
=.!!.SL
(1-99)
According to the assumptions this amount of heat will remain in a rod of constant cross sectional area due to the lack of a temperature gradient in the welding direction. Under such conditions the mode of heat flow becomes essentially one-dimensional, and the temperature distribution is given by equation (1-5):
T - To
112
(1-100)
Equation (1-100) represents the simplified solution for a fast moving high power source * in a thin sheet, and is valid within a limited range of the more general Rosenthal equation for twodimensional heat flow (equation (1-81. By substituting the appropriate dimensionless parameters into equation (1-100), we obtain:
*Since the shape of a given isotherm in the x-y plane is determined by the q.Id ratio, the minimum welding speed which is required to maintain I-D heat flow increases with decreasing qJvd ratios. Hence, the term 'fast moving high power source' is also appropriate in the case of the thin plate welding.
57
1,, ,,
I, t I I I I
I,
_
_
90 n3
The locus of the peak temperatures is readily evaluated from equation (1-101) by setting dln(8B1n3)ld't = 0:
*'
(1-101)
which gives
and
9po
n3
_ ~_
~2et;;; ~2;
f1C_1
'II m
(1-102)
It is evident from the plot in Fig. 1.35 that the predicted width of the isotherms is always greater than that inferred from the general thin plate solution (equation (1-83)) due to the assumption of one-dimensional heat flow. However, such deviations become negligible at very small 8pB1n3 ratios because of a small temperature gradient in the welding x direction compared to the transverse y direction of the plate. A general graphical representation of the weld thermal programme (similar to that shown in Fig. 1.25 for a fast moving high power source on a heavy slab) can be obtained by combining equations (1-101) and (1-102):
58 6
5'
~
I
5 4 3
~E ---
;e
9' ,...
L----------=--=-:-::-=-=-:-:-=-=--
=_-=_-=_:-:-
- - -
~~Y.!Y''p~o~e -- - -
o~~~--~~~~wu~--~--~~~~--~--~~~~~
0.01 0.1
6po/n3
10
Fig. 1.35. Theoretical width of isotherms under 1-D and 2-D heat flow conditions, respectively at pseudosteady state (thin plate welding).
1.0 . (T T ) (Tp-TO)
0
0.8
0.6
<J:)
'""
ale
ct>
0-
0.4 0.2
10
15 2t/('I1 )2 m
20
25
30
pattern in thin plate welding at high arc power and high welding speed.
59 (1-103)
~ = ~
8p
~~
\If
m
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes (SMAW) under the following conditions: 1= 80A, U = 25V, v = 5mm s', To = 20C Calculate the retention time within the critical temperature range for chromium carbide precipitation (i.e. from 650 to 850C) for points located at the 850C isotherm.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3/8 becomes:
n3 =
80x25xO.8 21tx4x7.4x2
=4.30
A comparison with Fig. 1.35 shows that the assumption of 1-D heat flow is justified when 8p ~ 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 650C) to 8p = 0.56 (Tp = 850C) and down again to 6 = 0.43 can be read from Fig. 1.36. Taking the ordinate 6/6p equal to 0.76, we obtain:
which gives
~'t
r
~ 112
and ~t
r r
= --112s (5)2
~ 36s
1.10.4 Medium thick plate solution In a real welding situation the assumption of three-dimensional or two-dimensional heat flow inherent in the Rosenthal equations is not always met because of variable temperature gradients in the through thickness z direction of the plate.
Model (after Rosenthal!")
The general medium thick plate model considers a point heat source moving at constant speed across a wide plate of finite thickness d. With the exception of certain special cases (e.g. watercooling of the back side of the plate), it is a reasonable approximation to assume that the
60
plate surfaces are impermeable to heat. Thus, in order to maintain the net flux of heat through both boundaries equal to zero, it is necessary to account for mirror reflections of the source with respect to the planes of z = 0 and z = d. This can be done on the basis of the 'method of images' as illustrated in Fig. 1.37. By including all contributions from the imaginary sources ... 2q-2 , 2q-l , Zq, , 2q2 ,.. .located symmetrically at distances 2id below and above the upper surface of the plate, the pseudo-steady state temperature distribution is obtained in the form of a convergent series *: T-T
o
=~
2nA.
exp(-
ux) 2a
(1-104)
where R, =~x2 + y2 +(z-2id)2 . Note that equation (1-104) is simply the general Rosenthal thick plate solution (equation (1-45)) summed for each source.
+
2d 2d
T ~~----~~--~~--~
Fig. 1.37. Real and imaginary point sources on a medium thick plate.
s, and
*The number of imaginary heat sources necessary to achieve the required accuracy depends on the chosen values of vd/2a.
61
By substituting the dimensionless parameters defined above into equation (1-104), we obtain: (1-105) where
It follows from equation (1-104) that the thermal conditions will be similar to those in a thick plate close to the centre of the weld. Moreover, Rosenthal 13, 14 has shown on the basis of a Fourier series expansion that equation (1-104) converges to the general thin plate solution (equation (1-81) for points located sufficiently far away from the source. However, at intermediate distances from the heat source, the pseudo-steady state temperature distribution will deviate significantly from that observed in thick plate or thin plate welding because of variable temperature gradients in the through-thickness direction of the plate. Within this 'transition region', the thermal programme is only defined by the medium thick plate solution (equation (1-104)). 1.10.4.1 Dimensionless mapsfor heatflow analyses Based on the models described in the previous sections, it is possible to construct a series of dimensionless maps which provide a general outline of the pseudo-steady state temperature distribution during arc welding. 20 Construction of the maps The construction of the maps is done on the basis of the medium thick plate solution (equation (1-105)). This model is generally applicable and allows for the plate thickness effect in a quantitative manner. Since the other solutions are only valid within specific ranges of this equation, they will have their own characteristic fields in the temperature-distance or the temperature-time space. The extension of the different fields can be determined from numerical calculations of the temperature distribution by comparing each of these models with the medium thick plate solution, using a conformity of 95% as a criterion. Similarly, when the 95% conformity is' not met between the respective solutions, the fields are marked 'transition region'. Since any combination of dimensionless temperature, operating parameter, and plate thickness locates a point in a field, it means that the dominating heat flow mechanism can readily be read off from the maps. Peak temperature distribution The variation of peak temperature with distance in the 'V(y )-direction has been numerically evaluated from equation (1-105) for different values of the dimensionless plate thickness (8 = vd/2a). The results are shown graphically in Fig. 1.38(a) and (b) for the two extreme cases of ~ = 0 (z = 0) and, = 8 (z = d), respectively. An inspection of the maps reveals that the temperature-time pattern in stringer bead weldments can be classified into three main categories:
62
(a)
0.1
.... ",,..,.,
0.01 0.01
0.1 'I'm
.,
:
I I I
10
100
(b)
100
1-D heat flow
c. <l>
'c:
('t)
0.1
0.01 0.01
0.1
10
100
Fig. 1.38. Peak temperature distribution in transverse direction ('V = 'tim) of plate; (a) Upper plate surface (~ = 0), (b) Lower plate surface (~ = 8).
63
1. Close to the heat source, the thermal programme will be similar to that in a thick plate (Fig. 1.38(a)), which means that the temperature distribution is determined by equation (154). For large values of the dimensionless plate thickness, the mode of heat flow may become essentially two-dimensional. This corresponds to the limiting case of a fast moving high power source in a thick plate (equation (1-74)). Under such conditions the slope of the 8p /n3-'Vm curves in Fig. 1.38(a) attains a constant value of -2. 2. With increasing distance from the heat source, a transition from three-dimensional to two-dimensional heat flow may occur, depending on the dimensionless plate thickness and the operational conditions applied. Considering the upper surface of the plate (Fig. 1.38(a)), the extension of the transition region is seen to decrease with increasing values of 8 as the conditions for thick plate welding are approached. The opposite trend is observed for the bottom plate surface (Fig. 1.38(b )), since a small dimensionless plate thickness generally results in a more rapid equalisation of the temperature gradients in the t(z) direction. When the curves in Fig. 1.38(b) become parallel with the x-axis, the temperature at the bottom of the plate reaches its maximum value. Note that within the transition region, reliable predictions of the pseudo-steady temperature distribution can only be made from the medium thick plate solution (equation (1-105)). 3. For points located sufficiently far away from the heat source, the temperature gradients in the through-thickness direction of the plate become negligible. This implies that the temperature distribution at the upper and lower surface of the plate is similar, and can be computed from the thin plate solution (equation (1-83)). When the conditions for onedimensional heat flow are approached (equation (1-101)), the slope of the 8p/n3-'Vm curves in Fig. 1.38(a) and (b) attains a constant value of -1. Cooling conditions close to weld centre-line Figure 1.39 contains a plot of the cooling programme for points located on the weld centre-line ('V = t = 0), as calculated from equation (1-105). A closer inspection of Fig. 1.39 reveals that the slope of the cooling curves increases gradually from -1 to -0.5 with increasing distance from the heat source. This corresponds to a change from three-dimensional to one-dimensional heat flow. From Fig. 1.39 it is possible to read-off the cooling time within specific temperature intervals for a wide range of operational conditions. These results are also valid for positions outside the weld centre-line, since the cooling curves are virtually parallel in the transverse 'V direction of the plate. A requirement is, however, that the peak temperature of the thermal cycle is significantly higher than the actual temperature interval under consideration. Retention times at elevated temperatures The retention time, Ll'tr' is defined as the total time spent in a thermal cycle from a chosen reference temperature 8 to the peak temperature 8p and down again to 8. This parameter can readily be computed from equation (1-105) by means of numerical methods. The results of such calculations (carried out in position t = 0) are shown graphically in Fig. 1.40 for 8 = 0.5 Sp. An inspection of Fig. 1.40 reveals a complex temperature-time pattern. In this case it is not possible to determine the exact field boundaries between the respective solutions, since the
64 1000
METALLURGICAL
MODELLING
OF WELDING
100
..
i
a,
c:
C')
1.0
0.1
't~
Fig. 1.39. Cooling programme for points located on the weld centre-line ('" = , =0).
0.1
1.0
A'tr
10
65
mode of heat flow may vary within a single thermal cycle. Hence, the extension of the different fields is not indicated in the graph. The results in Fig. 1.40 provide a systematic basis for calculating the retention time within specific temperature intervals under various welding conditions. Isothermal contours Because of the number of variables involved, it is not possible to present a two-dimensional plot of the isotherms without first specifying the dimensionless plate thickness. Examples of calculated isotherms in different planes are shown in Figs. 1.41 and 1.42 for 8 equal to 0.5 and 5, respectively. It is evident that an increase in the dimensionless plate thickness from 0.5 to 5 has a dramatic effect on the shape and position of the isothermal contours. However, in order to explain these observations in an adequate manner, it is necessary to condense the results into a two-dimensional diagram. As shown in Fig. 1.43, this can be done by plotting the calculated field boundaries in Fig. 1.38(a) at maximum width of the isotherms vs the parameters Op In3 and vdl2a. It is seen from Fig. 1.43 that a large plate thickness generally will favour three-dimensional heat flow. With decreasing values of Op In3' the conditions for a fast moving high power source are approached before the transition from the thick plate to the thin plate solution occurs. In such cases the isotherms at the bottom of the plate will be strongly elongated in the welding ~ direction and shifted to positions far behind the heat source. The opposite trend is observed at small values of vdl2a, since a rapid equalisation of the temperature gradients in the throughthickness direction of the plate will result in elliptical isotherms at both plate surfaces, located in an approximately equal distance from the heat source. In either case the temperature at which the cross-sectional isotherms approach a semi-circle or become parallel with the ~(z)axis can be obtained from Fig. 1.43 by reading-off the intercept between the line for the dimensionless plate thickness and the respective field boundaries. Limitations of the maps Since the maps have been constructed on the basis of the analytical heat flow equations, it is obvious that they will apply only under conditions for which these equations are valid. The simplifying assumptions inherent in the models can be summarised as follows: (a) The parent material is isotropic and homogeneous at all temperatures, and no phase changes occur on heating. The thermal conductivity, density, and specific heat are constant and independent of temperature. The plate is completely insulated from its surroundings, i.e. there are no heat losses by convection or radiation from the boundaries. The plate is infinite except in the directions specifically noted. The electrode is not consumed during welding, and all heat is concentrated in a zero-volume point or line.
(b)
(c)
(d) (e)
66 (a)
-1
-0.5
'I' o
0.5
(b)
-6
I
-5
-4
-3
I
-2
-1 0
6/n3=
1.51
21
~.75~3~1
0 0.5
(c)
-6
-5
-4
-3
-2
..
-1
Fig. 1.41. Computed isothermal contours in different sections for 8 = 0.5; (a) Front view ('JI 'JIm), (b) Side view ('JI = 0), (c) Top view (~ = 0) and bottom view (~ = 8).
67
(a)
\(I
-10
-5
10
5 (b)
-20 -16 -12 -8 -4
(c) ~=O
9/n3 =0.05
-20
Fig. 1.42. Computed isothermal contours in different sections for B = 5; (a) Front view ('V = 'Vm); (b) Side view ('V 0); (c) Top view (~ = 0) and bottom view (~ = B).
68
1000
100
10
~ a.
c:D
('t)
1.0
0.1
0.01
0.1
1.0
10
100
1000
8=vd/2a~
= tVm)
in transverse
direction (tV
(f)
Pseudo-steady state, i.e. the temperature does not vary with time when observed from a point located in the heat source ..
In general, the justification of these assumptions relies on a good correlation between theory and experiments. However, since the analytical solutions ignore the important role of arc energy distribution and directed metal currents in the weld pool, predictions of the weld thermal programme should be restricted to positions well outside the fusion zone where such effects are of less importance (to be discussed below).
Example (1.12)
Consider stringer bead welding (GMAW) on a 12mm thick plate of aluminium (> 99% AI) under the following conditions:
I = 260A, U = 25V,
1)
3mm s",
To = 20C
Based on Fig. 1.43, sketch the peak temperature contours in the transverse section of the weld at pseudo-steady state.
69
Solution If we neglect the latent heat of melting, the parameter (T; = Tm) becomes: n3 =
n3
= 0.1, I.e.
= 10
S-1
and a
85mm2
S-l
=0.21
Tp (OC)
340 --. 660
Model System Medium thick plate solution Thin plate solution (2- D heat flow) Thin plate solution (l-D heat flow)
Comments Heat flow in x and y directions, partial heat flow in z direction Heat flow in x and y directions, negligible heat flow in z direction Heat flow in y direction, negligible heat flow in x and z directions
<0.17
< 130
A sketch of the HAZ peak temperature contours in the transverse section of the weld is shown in Fig. 1.44.
Case Study (1.1)
Consider stringer bead GMA welding on 12.5mm thick plates of low alloy steel and aluminium (i.e. a AI-Mg-Si alloy), respectively under the conditions E = 1.5 kJ mm ' and 11 = 0.8. Details of welding parameters for the four series involved are given in Table 1.4. Computed peak temperature contours in the transverse section of the welds are given in Figs. 1.45 and 1.46.
-y
o
('I)
Fig. 1.44. Schematic diagram showing specific peak temperature contours within the HAZ of an aluminium weld at pseudo-steady state (Example 1.12).
70 (a) .. 30
WELD A1
y(mm)
.. 20 ..10 0
10
20
30 0 .-..
T(
""-"
E
N
12.5 (b)
WELD A2
-30 -20 -10
y(mm) 0
10
20
30
0 .-..
"'-"
E E
12.5
Fig. 1.45. Computed peak temperature contours in aluminium welding at pseudo-steady state (Case study 1.1); (a) Weld AI, (b) WeldA2. Black regions indicate fusion zone.
Aluminium welding In general, the maximum width of the isotherms at the upper and lower surface of the plate can be obtained from Fig. 1.38(a) and (b), although these maps are not suitable for precise readings. A comparison with the computed peak temperature contours in Fig. 1.45(a) and (b) reveals a strong influence of the welding speed on the shape and position of the cross-sectional isotherms at a constant gross heat input of 1.5 kJ mrrr '. It is evident that the extension of the fusion zone and the neighbouring isotherms becomes considerably larger when the welding speed is increased from 2.5 to 5 mm s'. This effect can be attributed to an associated shift from elliptical to more elongated isotherms at the plate surfaces (e.g. see Fig. 1.43), which reduces heat conduction in the welding direction. It follows from Fig. 1.43 that the upper left comer of the map represents the typical operating region for aluminium welding. Because of the pertinent differences in the heat flow conditions, the temperature-time pattern will also vary significantly between the respective series as indicated by the maps in Figs. 1.39 and 1.40. Hence, in the case of aluminium welding the usual procedure of reporting arc power and travel speed in terms of an equivalent heat input per unit length of the bead is highly questionable, since this parameter does not define the weld thermal programme. In general, the correct course would be to specify both q 0' v and d, and compare the weld thermal history on the basis of the dimensionless parameters n3 and 8. Steel welding If welding is performed on a steel plate of similar thickness, the operating region will be shifted to the lower right comer of Fig. 1.43. Under such conditions, the isotherms adjacent to the fusion boundary will be strongly elongated in the x-direction even at very low welding speeds (see Fig. 1.42). This implies that the thermal programme approaches a state where the temperature distribution is uniquely defined by the net heat input l1E, corresponding to the
71
(a)
WELD 81
y(mm)
.E E .......
N
(b)
WELD 82
y(mm)
-20 -10
...-..r....,...-,........................ ~
10
20
........
E ......
N
Fig. 1.46. Computed peak temperature contours in steel welding at pseudo-steady (a) Weld S I, (b) Weld S2. Black regions indicate fusion zone.
v
(mm s")
d
(mm)
n3
s
0.50 0.25
5 2.5 8
]0
5
4800
72
limiting case of a fast moving high power source. As a result, the calculated shape and width of the fusion boundary and neighbouring isotherms are seen to be virtually independent of choice of qo and vas illustrated in Fig. 1.46(a) and (b). 1.10.4.2 Experimental verification of the medium thick plate solution It is clear from the above discussion that the medium thick plate solution provides a systematic basis for calculating the temperature distribution within the HAZ of stringer bead weldments under various welding conditions. In the following, the accuracy of the model will be checked against extensive experimental data, as obtained from in situ thermocouple measurements and numerical analyses of a large number of bead-on-plate welds. Weld thermal cycles Examples of measured and predicted weld thermal cycles in aluminium welding are presented in Fig. 1.47. It is evident that the medium thick plate solution predicts adequately the HAZ temperature-time pattern under different heat flow conditions for fixed values of the peak temperature. This, in tum, implies that the model is also capable of predicting the total time spent in a thermal cycle within a specific temperature interval as shown in Fig. 1.48. Weld cooling programme At temperatures representative of the austenite to ferrite transformation in mild and low alloy steel weldments, the conditions for a fast moving high power source are normally approached before the transition from thick plate to thin plate welding occurs (see Fig. 1.39). In such cases, it is possible to present the different solutions for L\'t8/5 (at = ~= 0) in a single graph by introducing the following groups of variables ":
'I'
Ordinate:
L\'t8/5
Abscissa: Relevant literature data for the cooling time between 800 and 500C are given in Fig. 1.49. A closer inspection of the figure reveals a reasonable agreement between observed and predicted values in all cases. For welding of thick plates, the ordinate attains a constant value of 1. Similarly, in thin sheet welding, the slope of the curve becomes equal to 1. In aluminium welding, the thermal conditions will be much more complex because of the resulting higher base plate thermal diffusivity (see Fig. 1.39). Hence, it is not possible to describe the weld cooling programme in terms of equations (1-74) and (1-101), which apply to fast moving high power sources. The plot in Fig. 1.50 confirms, however, that the medium thick plate solution is also capable of predicting the cooling time within specific temperature intervals (e.g. from 300 to 100C) in aluminium weldments.
*These groups of variables can be obtained from equations (1-74) and (1-101).
73
600
6 400 e.... ~
1
e
(I)
GMAW: qo
d =8mm
:::J
Co
{E.
200
10 Time (sec)
20 ~
30
Fig. 1.47. Comparison between measured and predicted weld thermal cycles in aluminium welding for fixed values of Tp- Data from Myhr and Grong.20
60
"C
t
~
Q) Q)
40
-~
0
+'"
CIl ..c
!!1 20
<1
T-~
p t I
,- I
0.5 T p
t
.6.
t,
20
8. t r (s), predicted
40
60
Fig. 1.48. Comparison between measured and predicted retention times in aluminium welding for fixed values of Tp- Data from Myhr and Grong.l"
74 10
t
p
6
4 8 2
~lcO
1--1
<1
~1~Vl
c::
I 8
~
1 0.8
0.6
0.4 0.01
0.1
~[1
10 9800
100
02
+_1
~OO
Fig. 1.49. Comparison between observed and predicted cooling times from 800 to 500C in steel welding (solid lines represent theoretical calculations). Data from Myhr and Grong.P'
t
"0
200
160
(1)
o C
..0
en
Q)
120
80
40
t(sec)
t3/1
o~~-----~-----~--~--~~~~-----~--~--~-----~~-----~ o 40
80 .6.t
3/1
200
240
Fig. 1.50. Comparison between observed and predicted cooling times from 300 to lOOC in aluminium welding. Data from Myhr and Grong.20
75
Peak temperatures and isothermal contours Figure 1.51 shows a comparison between measured and predicted HAZ peak temperatures in aluminium welding. It is evident that the relative positions of the HAZ isotherms can be calculated with a reasonable degree of accuracy from the medium thick plate solution, provided that the equation is precalibrated against a known isotherm (i.e. the fusion boundary). Additional information on the HAZ peak temperature distribution in aluminium welding can be obtained from the data of Koe and Lee21 reproduced in Fig. 1.52. These numerical calculations * showed a good correlation with experimental measurements. A comparison with the medium thick plate solution in Fig. 1.52 reveals a fair agreement between numerically and analytically computed isothermal contours. It is interesting to note that even though the plate thickness is small (i.e. 3.2mm), the mode of heat flow becomes essentially three-dimensional close to the fusion boundary. This important point is often overlooked when discussing the relevance of the analytical heat flow models in thin sheet welding. 1.10.4.3 Practical implications The following important conclusions can be drawn from the results presented in Figs. 1.381.52:
500
t
"'C Q)
400
.0
e m
o
300
9
1-0.
200
100
100
500
600
Fig. 1.51. Comparison between observed and predicted HAZ peak temperatures in aluminium welding. Data from Myhr and Grong.i''
*Based on the finite difference method (FDM).
76
y(mm)
E
N
3 Fig. 1.52. Comparison between numerically and analytically computed peak temperature contours in GTA welding of a 3.2mm thin aluminium sheet. (Broken lines: numerical model; solid lines: analytical model.) Data for welding parameters and material properties are given in Ref. 21.
1. Considering heat flow and temperature distribution in fusion welding, there exists no defined plate thickness which can be regarded as 'thick' or 'thin'. Accordingly, in a real welding situation, the mode of heat flow will vary continuously with increasing distance from the heat source. 2. Close to the centre of the weld, the thermal conditions will be similar to those in a heavy slab. This means that the temperature distribution is approximately described by the Rosenthal thick plate solution. 3. At intermediate distances from the heat source, the temperature distribution will deviate significantly from that observed in thick plate welding because of variable temperature gradients in the through-thickness direction of the plate. Within this transition region, reliable predictions of the pseudo-steady state temperature distribution can only be made from the medium thick plate solution. 4. For points located sufficiently far away from the heat source, the temperature gradients in the through-thickness direction of the plate become negligible. Under such conditions, the weld thermal programme is approximately defined by the Rosenthal thin plate solution. 5. In general, a full description of the weld thermal history requires that both the arc power q 0' the travel speed v, and the plate thickness d are explicitly specified. Hence, the usual procedure of reporting welding variables in terms of an equivalent heat input per unit length of the bead, E(kJ mrrr '), is highly questionable, since this parameter does not define the weld thermal programme. An exception is welding of thick steel plates, where the temperature distribution approaches that of a fast moving high power source because of a low thermal diffusivity of the base metal. 6. A comparison between theory and experiments shows that the medium thick plate solution predicts adequately both the peak temperature distribution and the temperaturetime pattern within the HAZ of stringer bead weldments for a wide range of operational conditions (including aluminium and steel welding). A requirement is, however, that the equation is calibrated against a known isotherm (e.g. the fusion boundary) due to the simplifying assumptions inherent in the model.
77
1.10.5 Distributed heat sources In some cases it is also necessary to consider the important influence of filler metal additions, arc energy distribution, and convectional heat flow in the weld pool on the resulting bead morphology to obtain a good agreement between theory and experiments. Of particular interest in this respect is the formation of the so-called weld crater/weld finger, frequently observed in GMA and SA stringer bead weldments (see Fig. 1.53). Although the nature of these phenomena is very complex, they can readily be accounted for by applying an empirical correction for the effective heat distribution in the weld pool. 1.10.5.1 General solution
Model (after Myhr and Grong22)
The heat distribution used to simulate the weld crater/weld finger formation is shown schematically in Fig. 1.S4(a). Here we consider two discreate distributions of elementary point sources, which extend in the y- and z-direction of the plate, respectively. The contribution from each source to the temperature rise in an arbitrary point P located within the plate is calculated on the basis of the "method of images", as shown in Fig. 1.S4(b) and (c). For a heat source displaced in the y-direction (Fig. 1.S4(b, the temperature field is given by equation (1104): T(q ) = ~exp(a 21tA
ux) 2a
(1-106)
where
z
Fig. 1.53. Schematic diagram showing the weld crater/weld finger formation during stringer bead welding.
78
(a)
2d
2d
(b)
2d
Fig. 1.54. General heat flow model for welding on medium thick plates; (a) Physical representation of the heat distribution by elementary point sources, (b) Method for calculating the temperature field around an elementary point source displaced along the y-axis.
79
(c)
2d
I
Fig. 1.54. General heat flow model for welding on medium thick plates( continued): (c) Method for calculating the temperature field around an elementary (submerged) point source displaced along the z-axis,
Similarly, for a submerged point source located along the z-axis (Fig. I.S4(c)), we obtain:
T(Qb)=3..L
41tA
exp (-
'\)X)[j~=(1/ 2a .
...J
R.) exp
J
(-~R.)
2a
J=-OO
(1-107)
80 where
and
Note that equation (1-107) correctly reduces to equation (1-106) when ~ approaches zero. The total temperature rise in point P is then obtained by superposition of the temperature fields from the different elementary heat sources, i.e.: T(t)=To+ l:rT(q~)+T(qt)l
i
(1-108)
where qo = L(q~
i
+qt)=TJUI
In practice, we can subdivide the heat distributions into a relatively small number of elementary point sources, and usually a total number of 8 to 10 sources is sufficient to obtain good results (i.e. smooth curves). However, the relative strength of each heat source and their distribution along the y- and z-axes must be determined individually by trial and error by comparing the calculated shape of the fusion boundary with the real (measured) one. Figure 1.55 shows the results from such calculations, carried out for a single pass (bead-ingroove) GMA steel weld. It is evident that the important effect of the weld crater/weld finger formation on the HAZ peak temperature distribution is adequately accounted for by the present model. A weakness of the model is, of course, that the shape and location of the fusion boundary must be determined experimentally before a prediction can be made. 1.10.5.2 Simplified solution Similar to the situation described above, the point heat source will clearly not be a good model when the heat is supplied over a large area. Welding with a weaving technique and surfacing with strip electrodes are prime examples of this kind.
Model (after Grong and Christensen'")
As a first simplification, the Rosenthal thick plate solution is considered for the limiting case of a high arc power q0 and a high welding speed v, maintaining the ratio q0/ v within a range applicable to arc welding. Consider next a distributed heat source of net power density q 0 /2L extending from -L to +L on either side of the weld centre-line in the y-direction", as shown schematically in Fig. 1.56. It follows from equation (1-73) that an infinitesimal source dq located between y' and y' + dy' will cause a small rise of temperature in point P at time t, as: dT , y where (r')
= (dqyl
/ UJ(!)exp[21tA. t
(r')2] 4at
(1-109)
= ~(y
- y')2
* Alternatively,
81
0
2
4 6
8
g
N
10 12
14 16 18
20
12
14
16
18
20
Fig. 1.55. Calculated peak temperature contours in the transverse section of a GMA steel weldment (Operational conditions: 1= 450A, U = 30V, v = 2.6mm s',d = 50mm).
q dy' =_0__ -y
y' 2L
z
2-D heat flow
z
Fig. 1.56. Distributed heat source of net power density qj2L on a semi-infinite body (2-D heat flow).
82 Substituting u=(y-
y')/ ~4at
,dy'=-du~
= -L
to y'
= +L
gives:
T _ To = (qo / 21.)L)[
pc.fJi
(1-110)
Peak temperature distribution Because of the complex nature of equation (1-110), the variation of peak temperature with distance in the y-z plane can only be obtained by numerical or graphical methods. Accordingly, it is convenient to present the different solutions in a dimensionless form. The following parameters are defined for this purpose: Dimensionless operating parameter:
(1-111)
Dimensionless time:
't
=--
-JLkii
L
(1-112)
Dimensionless y-coordinate:
B=
Dimensionless z-coordinate:
.l
L
(1-113)
(1-114)
(1-115)
where
e is the dimensionless
83
The variation of peak temperature 8p with distance in the Y-Z plane has been numerically evaluated from equation (1-115) for chosen values of J3 and v (i.e. J3 = 0, J3 = 3/4, J3 = 1, and 'Y= 0). The results are presented graphically in Figs. 1.57 and 1.58 for the through thickness (z = zm) and the transverse (y = Ym) directions, respectively. These figures provide a systematic basis for calculating the shape of the weld pool and neighbouring isotherms under various welding conditions. In Fig. 1.59 the weld width to depth ratio has been computed and plotted for different combinations of nw and 8p It is evident that the predicted width of the isotherms generally is much greater, and the depth correspondingly smaller than that inferred from the point source model. Such deviations tend to become less pronounced with decreasing peak temperatures (i.e. increasing distance from the heat source). At very large value of nw, the theoretical shape of the isotherms approaches that of a semi-circle, which is characteristic of a point heat source.
Example (1.13)
Consider GMA welding with a weaving technique on a thick plate of low alloy steel under the following conditions: 1= 315A, U = 25V, v
= 5mm
s-l, 2L
= 20mm,
To = 20C
Sketch the contours of the fusion boundary and the ACrisotherm (710C) in the Y-Z plane. Compare the shape of these isotherms with that obtained from the point heat source model. Solution If we neglect the latent heat of melting, the operating parameter at the chosen reference temperature (T; = Tm) becomes: n
w
0.47
e = -- 1 = 2.11
0.47
Readings from Figs. 1.57 and 1.58 give: Zm(y zm(y zm(y Ym(z
= 0) = 7.5mm) = 10mm) = 0)
= = = =
Ac j-temperature In this case the peak temperature should be referred to 720C, i.e.:
84
o ~------~------~~------~------~--------~------~--~ 0.4 o
0.8
1.2
Fig. 1.57. Calculated peak temperature distribution in the through-thickness ferent positions along the weld surface.
o1 ------------------------------------------------------~-1.08
1.24
Fig. 1.58. Calculated peak temperature distribution in the transverse direction of the plate at position 'Y (z) = o.
85
t:
4 2
e
p
from which
Readings from Figs. 1.57 and 1.58 give: Zm(y zm(y zm(y Ym(z
= 0) = 7.5mm) = 10mm) = 0)
Similarly, equation (1-75) provides a basis for calculating the width of the isotherms in the limiting case where all heat is concentrated in a zero-volume point. By rearranging this equation, we obtain: r"
m
= 2a
U
~ 2n3
e8p
= 2x 5
5
.J9;
6.27
which gives
86 and
Figure 1.60 shows a graphical presentation of the calculated peak temperature contours.
Implications of model
It is evident from Fig. 1.60 that the predicted shape of the isotherms, as evaluated from equation (1-110), departs quite strongly from the semi-circular contours required by a point heat source. Moreover, a closer inspection of the figure shows that inclusion of the heat distribution also gives rise to systematic variations in the weld thermal programme along a specific isotherm, as evidenced by the steeper temperature gradient in the y-direction compared with the z-direction of the plate. This point is more clearly illustrated in Fig. 1.61, which compares the HAZ temperature-time programme for the two extreme cases of z = 0 and y = 0, respectively. It is obvious from Fig. 1.61 that the retention time within the austenite regime is considerably longer in the latter case, although the cooling time from 800 to 500C, J1.tS/5' is reasonably similar. These results clearly underline the important difference between a point heat source and a distributed heat source as far as the weld thermal programme is concerned.
M odellimitations
In the present model, we have used the simplified solution for a fast moving high power source (equation 1-73 as a starting point for predicting the temperature-time pattern. Since the equations derived later are obtained by integrating equation (1-73), they will, of course, apply only under conditions for which this solution is valid. Moreover, a salient assumption in the model is that the heat distribution during weaving can be represented by a linear heat source orientated perpendicular to the welding direction. Although this is a rather crude approximation, experience shows that the assumed heat dis-
2L=20 mm
-y, mm
-10
-5
10
+y,mm
z,mm Fig. 1.60.Predicted shape of fusion boundary and ACt-isotherm during GMA welding of steel with an oscillating electrode (Example 1.13). Solid lines: Distributed heat source; Broken lines: Point heat source.
87
1500 ~--------~----------~----------~---------.
o o
Q)
- - - - - - 1350
e
E ~
e :::J
1000
0-
500
I ___)______r
~1l.t8/5
.-I
b. t
S/5
o ~----------~--------~--~--------~---------------o
5 10
15 .
20
Time,s
Fig. 1.61. Calculated HAZ thermal cycles in positions y = 0 and
= 0 (Example 1.13).
tribution is not critical unless the rate of weaving is kept close to the travel speed. However, for most practical applications weaving at such low rates would be undesirable owing to an unfavourable bead morphology.
Case Study (1.2)
Surfacing with strip electrodes makes a good case for application of equation (1-110). Specifically, we shall consider SA welding of low alloy steel with 60mm X 0.5mm stainless steel electrodes. The operational conditions employed are listed in Table 1.5. It is evident from the metallographic data presented in Fig. 1.62 that neither the bead penetration nor the HAZ depth (referred to the plate surface) can be predicted readily on the basis of the present heat flow model when welding is carried out with a consumable electrode, owing to the formation of a reinforcement. This situation arises from the simplifications made in deriving equation (1-110). The problem, however, may be eliminated by calculating the depth of the Ac, and Ac, regions relative to the fusion boundary, i.e. flzm = zm(8p) - Zm (8p = 1), or L\Ym = Ym (8p) - Ym (8p = 1), for specific positions along the weld fusion line, as shown by the solid curves in Fig. 1.62 for 8p = 0.54 and 0.45, respectively. An inspection of the graphs reveals satisfactory agreement between theory and experiments in all three cases, which implies that the model is quite adequate for predicting the HAZ thermal programme as far as strip electrode welding is concerned. This result is to be expected, since the assumption of twodimensional heat flow is a realistic one under the prevailing circumstances.
Case Study (1.3)
As a second example we shall consider GTA welding (without filler wire additions) at various heat inputs and amplitudes of weaving within the range from 1 to 2.5 kJ mrrr ' and 0 to 15mm, respectively. Data for welding parameters are given in 'fable 1.6.
88
Weld S3
.. Fusion line
ACa
o AC1
Weld 54
Weld 85
,,
...
....... - --
ep=o.45~
z,mm Fig. 1.62. Comparison between observed and predicted Ac, and Ac, contours during strip electrode welding (Case study 1.2). Data from Grong and Christensen.l?
Table 1.5 Operational conditions used in strip electrode welding experiments (Case study 1.2).
Base metall filler metal combination Weld No. I (A) U (V)
u
(mm
S-I)
2L (mm)
(11 = 0.7)
nw
S3
Low alloy steell stainless steel
27 27 27
60 60 60
S4 S5
89
Table 1.6 Operational conditions used in GTA welding experiments (Case study 1.3).
I U (V) 13.5 14.0 13.5 12.5
v
(mm
2L
nw
Weld No.
B1 B2 B3 B4
s')
11 = 0.23
0040 0040
! I
Calibration procedure In general, a comparison between theory and experiments requires that the arc efficiency factor can be established with a reasonable degree of accuracy. Unfortunately, the arc efficiency factor for GTA welding has not yet been firmly settled, where values from 0.25 up to 0.75 have been reported in the literature (see Table 1.3). Additional problems result from the fact that only a certain fraction of the total amount of heat transferred from the arc to the base plate is sufficiently intense to cause melting. This has led to the introduction of the melting efficiency factor 11m' which normally is found to be 30-70% lower than the total arc efficiency of the process, depending on the latent heat of melting, the applied amperage, voltage, shielding gas composition, or electrode vertex angle.P Consequently, since these parameters cannot readily be obtained from the literature, the following reasonable values for 11m and 11 have been assumed to calculate nw in Table 1.6, based on a pre-evaluation of the experimental data: 11m = 0.12 (fusion zone), 11 = 0.23 (HAZ). It should be noted that the above values also include a correction for three-dimensional heat flow, since the assumption of a fast moving high power source during low heat input GTA welding is not valid. Hence, both the arc efficiency factor and the melting efficiency factor used in the present case study are seen to be lower than those commonly employed in the literature. Full weaving (welds Bl and B2) The results from the metallographic examination of the two GTA welds deposited under full weaving conditions are presented graphically in Fig. 1.63. Note that the shape of the fusion boundary as well as theAc3 and the Ac, isotherms can be predicted adequately from the present model for both combinations of E and L (an exception is the HAZ end points in position z = 0), provided that proper adjustments of 11m and 11 are made. The good correlation obtained in Fig. 1.63 between the observed and the calculated peak temperature contours justifies the adaptation of the model to low heat input processes such as GTA welding, despite the fact that the assumption of two-dimensional heat flow is not valid under the prevailing circumstances. No weaving (welds B3 and B4) For the limiting case of no weaving (Fig. 1.64), the concept of an equivalent amplitude of weaving has been used in order to calculate the peak temperature contours from the model. This parameter (designated Leq ) takes into account the effects of convectional heat flow in the weld pool on the resulting bead geometry, and is evaluated empirically from measurements of the actual weld samples. At low heat inputs (Fig. 1.64(a)), the agreement between theory and
90
o AC1
2L= 15
mm
Weld 82
Fig. 1.63. Comparison between observed and predicted fusion line, AC3 and ACI contours during GTA welding under full weaving conditions (Case study 1.3). Data from Grong and Christensen.l?
experiments is largely improved by inserting 2Leq = 7.5mm into equation (1-110), when comparison is made on the basis of the point source model. In contrast, at a heat input of 2.5 kJ mrrr ' (Fig. 1.64(b)), the measured shape of the HAZ isotherms is seen to approach that of a semi-circle, and hence the deviation between the present model and the simplified solution for a fast moving high power point source is less apparent. Intermediate weaving At intermediate amplitudes of weaving (2L = 5 and 7.5mm, respectively), convectional heat flow in the weld pool will also tend to increase the bead width to depth ratio beyond the theoretical value predicted from the present model, as shown in Fig. 1.65. The plot in Fig. 1.65
91
.. Fusion line
e AC3
o
AC1
Fig. 1.64.Comparison between observed and predicted fusion line, AC3 and ACI contours during GTA welding with a stationary arc (Case study 1.3). Solid lines: Distributed heat source, Broken lines: Point heat source. Data from Grong and Christensen.l?
includes all data obtained in the GTA welding experiments with an oscillating arc, as reported by Grong and Christensen. 19 These results suggest that the applied amplitude of weaving must be quite large before such effects become negligible. Consequently, adaptation of the model to the weld series considered above would require an empirical calibration of the weaving amplitude similar to that performed in Fig. 1.64 for stringer bead weldments to ensure satisfactory agreement between theory and experiments. 1.10.6 Thermal conditions during interrupted welding Rapid variations of temperatures as a result of interruption of the welding operation can have an adversely effect on the microstructure and consequently the mechanical properties of the weldment.
92
METALLURGICAL
MODELLING
OF WELDING
e
o
.c
Q)
6
4 A 2L=5.5 mm h. 2L=7.5mm CD 2 L=9.5 mm 0 2 L = 15.0 mm 0.2 0.4 0.6 2
"'0
-0
~
.c
2 0.1
nw
.,
Fig. 1.65. Comparison between observed and predicted weld width to depth ratios during GTA welding with an oscillating arc (Case study 1.3). Data from Grong and Christensen.l? L...-
qo
+q-'
r,o
I
of-I- _ ~ _--: ...:;r-. -. -
I,.-t I
t**, 't**
I I
t.r
I I
Fig. 1.66. Idealised heat flow model for prediction of transient temperatures during interrupted welding.
HEAT
FLOW
AND TEMPERATURE
DISTRIBUTION
IN WELDING
93
The situation existing after arc extinction may be described as shown in Fig. 1.66. From time ( = (* there is no net heat supply to the weldment. This condition is satisfied if the real source q 0 is considered maintained by adding an imaginary source +q0 and sink -q 0 of the same strength at t". The temperature at some later time t" in a given position R; (measured from the origin a") is then equal to the difference of temperatures due to the positive heat sources qo and +qo and the negative heat sink -qo. Each of these temperature contributions will be a product of a pseudo-steady state temperature Tp.s, and a correction factor 1(1 or 1(2 (given by equations (1-49) and (1-82), respectively). Hence, for 3-D heat flow, we have: B( 0"3,'T**) where
'OR '0 (* a = __ r" = __ 3 u:' 2a
O 0
=Q 0 up.s. ( 0"3,'T
** -
Kl (0 0"3,T ** -
T *)]
(1-116)
and
r'" =--
'02(**
2a
Similarly, for 2-D heat flow, we get: Q( 0"5,T**) - up.s. Q ( 0"5,T 0 ** _ ,. *)[ K2 ( (J" 5'" 0 **)
o uro 2a
K2 (0 (J" 5'"
** -,. *)]
(1-117)
where
0"5=-
(r 0 is the position of the weld with respect to the imaginary heat source at time t" in the x-y plane).
Example (1.14)
Consider repair welding of a heavy steel casting with covered electrodes under the following conditions: 1= 120A, U = 23V, v = 2mm s', To = 20C Suppose that a 50mm long bead is deposited on the top of the casting. Calculate the temperature in the centre of the weld 5 s after arc extinction.
Solution
The pseudo-steady state temperature for points located on the weld centre-line ('I' be obtained from equation (1-65). When t" - (* = 5 s, we get:
= ~= 0) can
e
p.s.
r'" _ r"
n3
= 120x23xO.8x2x2x5
=0.94
X
41t X (5)2
7.5
(2)2
Referring to Fig. 1.67, the position of the weld with respect to the imaginary heat source at time t" is 10mm, which gives:
94
I-
50 mm
I..
10 mm
--1
(a)
f = 180/(360-2$) ~-------------------~
(b) 3-D heat flow
= 180/270
(c)
95
<J3 ----
2 X 10 - 2 2x5
r'" = (2)2
and
30 = 12
2x5
X 5 = 2 2x5 At these coordinates, the correction factor Kl is seen to be 1 and 0.62, respectively (Fig. 1.18). The temperature in the centre of the weld 5 s after arc extinction is thus:
e = 0.94(1-0.62)
which is equivalent to
= 0.36
T= [20 + 0.36 (1520-20)]OC::::: 560C 1.10.7 Thermal conditions during root pass welding During conventional bead-on-plate welding the angle of heat conduction is equal to 180 due to symmetry effects (e.g. see Fig. 1.23). In order to apply the same heat flow equations during root pass welding, it is necessary to introduce a correction factor, f, which takes into account variations in the effective heat diffusion area due to differences in the joint geometry. Taking f equal to 1 for ordinary bead-on-plate welding (b.o.p.), we can define the net heat input of a groove weld as:"
[~ ]=f[ ~]
Recommended values of the correction factor Fig. 1.68.
(1-118)
b.o.p.
Example (1.15)
Consider deposition of a root pass steel weld in a double- V-groove with covered electrodes (SMAW) under the following conditions: 1= 120A, U
= 22V,
v = 2.6mm
S-I,
To = 20C
Calculate the cooling time from 800 to 500C (11t8/5), and the cooling rate (C.R.) at 650C in the centre of the weld when the groove angle is 60.
Solution
The cooling time, 11t8/5 , and the cooling rate, C.R., can be obtained from equation (1-68) and (1-71), respectively: Cooling time, ..1t 8/5
t1t8/5
=5.1xll
(180) 240
qo / u
= 102 Cs0
The above calculations show that the thermal conditions existing in root pass welding may deviate significantly from those prevailing during ordinary stringer bead deposition due to differences in the effective heat diffusion area. These results are in agreement with general experience (see Fig. 1.69). 1.10.8 Semi-empirical methods for assessment of bead morphology In fusion welding fluid flow phenomena will have a strong effect on the shape of the weld pool. Since flow in the weld pool is generally driven by a combination of buoyancy, electromagnetic, and surface tension forces (e.g. see Fig. 3.10 in Chapter 3), prediction of bead morphology from first principles would require detailed consideration of the current and heat flux distribution in the arc, the interaction of the arc with the weld pool free surface, convective heat transfer due to fluid flow in the liquid pool, heat of fusion, convective and radiative losses from the surface, as well as heat and mass loss due to evaporation. Over the years, a number of successful studies have been directed towards numerical weld pool modelling, based on the finite difference, the finite element, or the control volume approach.24-31 Although these studies provide valuable insight into the mechanisms of weld pool development, the solutions are far too complex to give a good overall indication of the heatand fluid-flow pattern. The present treatment is therefore confined to a discussion of factors affecting the nominal composition of single-bead fusion welds. This composition can be obtained from an analysis of the amount of deposit D and the fused part of the base material B, from which we can calculate the mixing ratio BI(B + D) or DI(B + D). Methods have been outlined in the preceding sections for handling such problems by means of point or line source models. The following section gives a brief description of procedures which can be used for predictions of the desired quantities in cases where the classic models break down, or where the calculation will be too tedious. 1.10.B.1 Amounts of deposit andfused parent metal The heat conduction theory does not allow for the presence of deposited metal. The rate of deposition, dMwldt, is roughly proportional to the welding current I, and is often reported as a coefficient of deposition, defined as:
k'(g A-I s-I) = dMw / dt I (1-119)
1 'Up
__ w =-1
dM
k'
dt
'Up
(1-120)
97
o d =60
lld=50mm
mm
<i
~
10
:
C')
cD
.5 (5 o 5
Fig. 1.69. Comparison between observed and predicted cooling times from 800 to 500C in root pass welding of steel plates (groove preparation as in Fig. 1.68(b). Data from Akselsen and Sagmo.l"
Recommended values of k'/p for some arc welding processes are given in Table 1.7. In practice, the deposition coefficient k' / p will also vary with current density and electrode stickout due to resistance heating of the electrode. Consequently, the numbers contained in Table 1.7 are estimated averages, and should therefore be used with care.
Example (1.16)
Consider stringer bead deposition (SAW) on a thick plate of low alloy steel under the following conditions:
1= 600A, U = 30V,
v = 5mm
S-1, To =
20C
Table 1.7 Average rates of volume deposition in arc welding. Data from Christensen.F
Welding Process SMAW GMAW, steel GMAW, aluminium SAW, steel k'/p (mm ' A-I s+) 0.3-0.5 0.6-0.7 -0.9 -0.7
98
+ D) at pseudo-steady state.
The amount of fused parent material can be obtained from equation (1-75). If we include an empirical correction for the latent heat of melting, the dimensionless radius vector (J" 4m becomes:
o
4m This gives:
=(3.n
e
)112
3
=4.59
B=!!..(r*)2
=~(2a(j4m)2 2 U
=!!..(2X5X4.59)2 2 5
mm? = 132mm2
Similarly, the amount of deposited metal can be calculated from equation (1-120). Taking k'/p equal to 0.7mm3 A-I s-1 for SAW (Table 1.7), we get: D= 0.7 600mm2 = 84mm2 5 The mixing ratio is thus:
B
(B+D)
Consider stringer bead deposition with covered electrodes (SMA W) on a thick plate of low alloy steel under the following conditions: 1= 180A, U=22V, v=2mms-1, To='20C
In this particular case the conditions for a fast moving high power source are not met. Thus, in order to eliminate the risk of systematic errors, the amount of fused parent metal should be calculated from the general Rosenthal thick plate solution (equation (1-45)) or read from Fig. 1.21. When T; = Tm (i.e. Sp = 1), we obtain:
n
3
=2.12
99
and
4a A - 4 X (5)2 1 7 2 - 42 2 B- \)2 1 - (2)2 . mm mm
2
Moreover, the amount of deposited metal can be calculated from equation (1-120). Taking k'/p equal to 0.4mm3 A-I s' for SMAW (Table 1.7), we get:
D and
=
B
0.4180mm2 2
42 (42+36)
= 36mm2
0.54
(B+D)
The above calculations indicate a small difference in the mixing ratio between SA and SMA welding, but the data are not conclusive. In practice, a value of BI(B + D) between 1/3 and 1/2 is frequently observed for SMAW, while the mixing ratio for SAW is typically 2/3 or higher. The observed discrepancy between theory and experiments arises probably from difficulties in estimating the amount of fused parent metal from the point heat source model. 1.10.8.2 Bead penetration It is a general experience in arc welding that the shape of the fusion boundary will depart quite strongly from that of a semi-circle due to the existence of high-velocity fluid flow fields in the weld pool. 24-31 For combinations of operational parameters within the normal range of arc welding, a fair prediction of bead penetration h can be made from the empirical equation derived by Jackson et al.:33 (1-121) A summary of Jackson's data is shown in Table 1.8. It is seen that the constant C in equation (1-121) has a value close to 0.024 for SAW and SMAW with E6015 type electrodes, and about 0.050 for GMA W with CO2 -shielding gas. Penetration measurements of GMAI Ar + O2, GMAI Ar, and GMAlHe welds, on the other hand, show a strong dependence of polarity, and shielding gas composition, to an extent which makes the equation useless for a general prediction. Such data have therefore not been included in Table 1.8.
Example (1.18)
Based on the Jackson equation (equation (1-121)), calculate the bead penetration for the two specific welds considered in Examples (1.16) and (1.17). Use these results to evaluate the applicability of the point heat source model under the prevailing circumstances.
Solution
= 0.024
[(600)4/5(30)2]113
mm
= 7.4mm
100
Table 1.8 Recommended bead penetration coefficients for some arc welding processes. Data from Jackson.P
Welding Process SAW, steel C -0.024 Comments Various types of fluxes (h from 3 to 15mm) Wide range of I, U, and v (h from 0.7 to 5mm) Electrode positive (h from 6.5 to Smm)
-0.024
-0.050
and h (SMAW)
= 0.024
[(180)4/2(22)2]113mm
= 2.5mm
The corresponding values predicted from the point heat source model are: r:(SAW) and r:(SMAW)
= 9.2mm = 5.2mm
Provided that the Jackson equation gives the correct numbers, it is obvious from the above calculations that the point heat source model is not suitable for reliable predictions of the bead penetration during arc welding. This observation is not surprising. 1.10.9 Local preheating So far, we have assumed that the ambient temperature To remains constant during the welding operation (i.e. is independent of time). The use of a constant value of To is a reasonable approximation if the work-piece as a whole is subjected to preheating. In many cases, however, the dimensions of the weldment allow only preheating of a narrow zone close to the weld. This, in turn, will have a significant influence on the predicted weld cooling programme, particularly in the low temperature regime where the classic models eventually break down when T approaches To.
Model (after Christensen 32)
The idealised preheating model is shown in Fig. 1.70. Here it is assumed that the weld centreline temperature is equal to the sum of the contributions from the arc and from the field of preheating. The former contribution is given by equation (1-45) for R = -x = in, provided that the plate thickness is sufficiently large to maintain 3-D heat flow. Similarly, the temperature field due to preheating can be calculated as shown in Section 1.7 for uniaxial heat conduction from extended sources (thermit welding). By combining equations (1-45) and (1-22), we obtain the following relation for the weld centre-line:
T - T
o
= qo / '\)
21CAt
+ (T*
0
- T )erf[L*
0
/ ~]
(1-122)
101
Temperature at t
=0
profile
__
-+--..... To
-y
....... --To
, +y Weld
where To* is the local preheating temperature, and L * is the half width of the preheated zone. Equation (1-122) can be written in a general form by introducing the following groups of parameters: Dimensionless temperature: 8* = T-T T* -T o
0 0
(1-123)
~
By inserting these parameters into equation (1-122), we obtain:
(1-126)
e* =...l + eif[
't6
QN /
('"6 )112]
(1-127)
102
It is evident from the graphical representation of equation (1-127) in Fig. 1.71 that the predicted weld cooling programme falls within the limits calculated for Q"~ 0 (no preheating) and Q" ~ (global preheating). The controlling parameter is seen to be the dimensionless half width of the preheated zone Q'~ which depends both on the actual width L *, the base plate thermal properties a, A, and the net heat input qo Iv.
00
Example (1.19)
Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low alloy steel under the following conditions: 1= 155A, U=22V, v=2mms-1, To*= 100C, To=20C, L*=200mm
Calculate the cooling time from 800 to 500C (L\tS/5), and the cooling time t100 measured from the moment of arc passage to the temperature in the centre of the weld reaches 100C. Solution First we calculate the time constant to from equation (1-124): t =
o
12
10 -
*CD6
0.1
1.0
't6 ----
10
~
100
of equation (1-127).
103
Cooling time, Lit 8/5 The dimensionless temperatures conforming to 800 and 5000e are: 9* =800-20=9.75 soo 100 - 20
9*
500
= 500 - 20 = 6 0 100 - 20 .
from which
L\tS/5
L\'t6 X to
This cooling time is only slightly longer than that calculated from equation (1-68) for To = 200e (6.9s), showing that moderate preheating up to 1000e is not an effective method of controlling ~tS/5. Cooling time, t100 When T = To * = 100oe, the dimensionless temperature e* = 1. Reading from Fig. 1.71 gives 't6 ~ 10, from which: t100 = 't6 X to = 10 x 108.5 s = 1085 s The above value should be compared with that evaluated from the numerical data ofYurioka al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling programme in practice is also a function of the plate thickness d, an effect which cannot readily be accounted for in a simple analytical treatment of the heat diffusion process. For the specific case considered above the parameter t100 varies typically from 500 to 900s, depending on the chosen value of d. This cooling time is significantly shorter than that calculated from equation (1-127), indicating that the analytical model is only suitable for qualitative predictions.
et
References
1. 2. 3. 4. 5. 6. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University Press. British Iron and Steels Research Association: Physical Constants of some Commercial Steels at Selected Temperatures; 1953, London, Butterworths. R. Hultgren, R.L. Orr, P.D. Anderson and K.K. Kelly: Selected Values of Thermodynamic Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons. E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83-121. J.E. Hatch (ed.): Aluminium - Properties and Physical Metallurgy; 1984, Metals Park (Ohio), American Society for Metals. Metals Handbook, 9th edn., Vol. 2,1979, Metals Park (Ohio), American Society for Metals.
104
Heat input:
10000 5000
E=1.7 kJ/mm
t
~o
..J
cD
1000
(5
.5 o
C)
500
2L*=400 mm
I--
100
50
50
100
150
200
T~ (OC)
Preheating temperature,
Fig. 1. 72. Cooling time to 100cC, t 100, in steel welding for different combinations Data from Yurioka et al. 35
7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.
K.P. Bentley, I.A. Greenwood, P. McKnowlson and R.G. Bakes: Brit. Weld. 1963,10,613619. N.N. Rykalin, A.I. Pugin and V.A. Vasil'eva: Weld. Prod., 1959,6,42-52. N.N. Rykalin: Berechnung der Wiirmevorgiinge beim Schweissen; 1953, Berlin, VEB Verlag Technik. H.M. Tensi, W. Welz and M. Schwalm: Aluminium, 1981,58,515-518. N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. t., 1965,12,54-75. Welding Handbook, 8th edn., Vol. 2, 1991, Miami (Florida), American Welding Society. D. Rosenthal: Weld. i., 1941,20, 220s-234s. D. Rosenthal: Trans. ASME, 1946,68,849-866. C.M. Adams: Weld. i, 1958, 37, 210s-215s. P.S. Myers, O.A. Uyehara and G.L. Borman: Weld. Res. Bull., 1967, 123, 1-46. T.W. Eagar and N.S. Tsai: Weld. t., 1983,62, 346s-355s. M.F. Ashby and K.E. Easterling: Acta Me ta II., 1984,32, 1935-1948. 0. Grong and N. Christensen: Mater. Sci. Tech., 1986,2, 967-973. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1990,38,449-460. S. Kou and Y. Le: Meta II. Trans., 1983, 14A, 2245-2253.
s..
105
22. 23. 24. 25. 26. 27. 28. 29. 30. 32. 33. 34. 35.
O.R. Myhr and 0. Grong: Unpublished work, 1990, University ofTrondheim, The Norwegian Institute of Technology. R.W. Niles and C.E. Jackson: Weld. J., 1975, 54, 25s-32s. G.M. Oreper, T.W. Eagar and J. Szekely: Weld J., 1983,62, 307s-312s. Y.H. Wang and S. Kou: Proc. Int. Conf. on Trends in Welding Research, Gatlinburg, TN, May, 1986, pp. 65-69, Publ. ASM International. S.A. David and J.M Vitek: Int. Mater. Rev., 1989,34,213-245. K.C. Mills and B.J. Keene: Int. Mater. Rev., 1990,35,185-216. R.L. Ule, Y. Joshi and E.B. Sedy: Metall. Trans., 1990, 21B, 1033-1047. T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243-257. A. Matsunawa: Proc. 3rd Int. Con! on Trends in Welding Research, Gatlinburg, TN, 1992, pp.3-16, Publ. ASM International. N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Norwegian Institute of Technology. C.E. Jackson: Weld. J., 1960,39, 226s-230s. O.M. Akselsen and G. Sagmo: Technical Report STF34A89147, 1989, Trondheim (Norway), Sintef-Division of Metallurgy. N. Yurioka, M. Okumura, S. Ohshita and S. Saito: IIW Doc. XII-E-I0-81, 1981.
s')
n.:,
cooling rate (OC s') plate thickness (mm) natural logarithm base number Gaussian error function complementary function Gaussian error
K}
m n
Ms
106
p
y-axis/transverse
direction (mm)
qo Q r
'Yl
two-dimensional (mm)
radius vector
a
8800
rm R
locus of peak temperature in T-r space (mm) three-dimensional (mm) radius vector
dimensionless temperature conforming to 800C dimensionless temperature conforming to SOOC dimensionless temperature conforming to the melting point dimensionless peak temperature volume heat capacity (J mrrr' C-l) thermal conductivity (W mrn ' C-I)
8500
Rm M T
locus of peak temperature in T-R space (mm) isothermal zone width (mm) temperature (OC) reference temperature (OC) ambient temperature (OC) melting point (OC) peak temperature (OC) time (s)
8m 6p
pc
t;
To Tm Tp
't
~'t
~'t8/5
r
t"
~t
dimensionless cooling time from 800 to SOOC dimensionless 300 to 100C cooling time from
time variable, (s) cooling time (s) cooling time from 800 to
Specific symbols
~t8/5
soooe
(s)
~t3/1
cooling time from 300 to 100C (s) integration parameter voltage (V) integration parameter x-axis/welding direction (mm)
dimensionless
operating parameter
u U
w
arc ignition time (s) isothermal zone width (mm) dimensionless R-vector locus of peak temperature in 8-(J" I space
Mlm
al
aim
107
8<:TIm
dimensionless width
isothermal zone
locus of peak temperature in O'-Q space (star denotes a specific peak temperature) 't3 dimensionless time dimensionless T; to 500aC 't3m cooling time from
'ti 'tIm
dimensionless time locus of peak temperature in O-'tl space dimensionless cooling time locus of peak temperature in 0 '-'t3 space
8'ti
Spot Welding
dt thickness of overlapping plates (mm) dimensionless operating parameter
Friction Welding
II integral in equation (1-30) net power generation at weld interface (W) duration of heating period (s) contact section temperature at the end of heating period (ac) dimensionless temperature dimensionless peak temperature dimensionless x-coordinate locus of peak temperature in Q'space dimensionless time
n2 th
:
th
<:T2
<:T2m
r-vector
locus of peak temperature in O-<:T2 space dimensionless time locus of peak temperature in O-'t2 space dimensionless cooling time
r,
e" e" p
Q' Q'm
't2 't2m
Ll't2
e"-
Thermit Welding
LI half width of groove (mm) initial temperature of liquid metal (ac) distance from reference point to infinitesimal source (mm)
't4
Arc Welding
B amount of fused parent metal
(mm-)
C D E
constant in Jackson equation amount of deposited metal (mm-) gross heat input per unit length of weld (kJ mnr ') correction factor for the net heat input during root pass welding
0'
108
FCAW GMAW GTAW h i,j, k k*
qi
imaginary heat source of net arc power : qm or qb (W) maximum intensity of distributed (Gaussian) heat source (W mrrr ') power density of distributed (Gaussian) heat source (W mrrr") infinitesimal heat source (W) two-dimensional y-z plane (mm) radius vector in
qm gas tungsten arc welding bead penetration (mm) integer variables .... -1, 0, 1.... dqyl constant in heat distribution function (mrrr=) coefficient of weld metal deposition (g A-I s') amplitude of weaving or half width of strip electrode (mm) half width of preheated zone (mm) equivalent amplitude of weaving (mm) half width of linear source in Gaussian heat distribution model (mm) length of weld bead (mm) R;, Rj, mass of weld metal (g) dimensionless operating parameter Ro dimensionless operating parameter in weaving model reference point in stationary coordinate system reference point in moving coordinate system arbitrary reference point to arbitrary point of observation strength of elementary heat sources (W) tlOO SAW SMAW 0' r* q(y)
k'
L*
distance from infinitesimal heat source to point P in x-y or y-z plane (mm) position of weld end-crater with respect to imaginary heat source at time t" in x-y plane (mm) distance from infinitesimal heat source to point Pin x-y-z space (mm)
i;
Lo
t;
R'
t.,
Mw n3 nw
0
s,
distances from real and imaginary heat sources to point P in x-y-z space (mm) position of weld with respect to imaginary heat source at time t** in x-y-z space (mm) submerged arc welding shielded metal arc welding time at moment of arc extinction (s) time constant in preheating model (s) cooling time to 100C (s)
0" P qa' qb
109
t0*
time constant in heat distribution function (s) time referred to moment of arc ignition (s) retention time (s) preheating temperature (OC)
~
<p
Kl
displacement of elementary heat source in z-direction (mm) groove angle ratio between real and pseudo-steady state temperature (thick plate welding) ratio between real and pseudosteady state temperature (thin plate welding) dimensionless temperature in preheating model dimensionless cross sectional area of isothermal enclosure (thick plate welding) dimensionless cross sectional area of isothermal enclosure (thin plate welding) density of weld metal (g mm=') dimensionless distance from real and imaginary heat sources to point
P
t**
~tr
T* 0
K2
v
V
8*
Xo
welding direction in stationary coordinate system (mm) x-coordinate at maximum width of isotherm (mm)
Al
Xm
A2
Yo
transverse direction in stationary coordinate system (mm) distance from infinitesimal heat source to point P in y-direction (mm) y-coordinate at maximum width of isotherm (mm)
0"3
y'
p
O"i
Ym
dimensionless R-vector dimensionless Ro-vector locus of peak temperature in 8-0"3 space dimensionless r*-vector
~ym Zo
isothermal zone width (mm) 0"3 through-thickness direction in stationary coordinate system (mm) z-coordinate at maximum width of isotherms (mm) isothermal zone width (mm) dimensionless y-coordinate weaving model in
0"5
0"3m
zm
0"4
&m
J3
0"4m
'Y
0"5 0"5m
110
~
~o
~a
dimensionless
Za
-axis
-axis
~m
dimensionless z-coordinate at maximum depth of isotherm melting efficiency factor dimensionless volume of is othermal enclosure dimensionless half width of preheated zone dimensionless time in weaving model dimensionless time in preheating model locus of peak temperature in space dimensionless retention time dimensionless time at moment of arc extinction dimensionless time referred to moment of arc ignition
8-'t
~m
'TIm
~t
r
n"
't5
~'
dimensionless distance from heat source to front of isothermal enclosure dimensionless distance from heat source to rear of isothermal enclosure dimensionless y-axis
~"
't6
'tm
'Va
'I'm
dimensionless
Ya
-axis
~'tr 't*
T - To where
L exp[ _(R.)2
I
I 4at]
(AI-I)
1=-00
111
and i is an integer variable (...-1, 0, 1...). A numerical solution of equation (ALl) gives a peak temperature distribution similar to that shown in Fig. 1.9.
in J
u o
exp(-u2)du
= 1- erf(u)
00
112
The functions have the following limiting values and symmetries: erf(O) = 0, erf(OO) = 1, erf(-u) = -erf(u) and eifc(O) = 1, eifc(oo) = 0, eifc(-u) = 2-erfc(u)
The following Fortran subroutine can be used for calculations of the error functions with a fractional error less than 1.2 X 10-7: FUNCTION ERFC(U) Z=ABS(U) T=l'/(1.+0.S*Z) ERFC=T*EXP(-Z*Z-1.26551223+ * * T*( 1.00002368+ T*( .37409196+ T*(.09678418+ T*( -.18628806+ T*(.27886807+ T*( -1.13520398+ T*(1.48851587+ T*(-.82215223+ T*.17087277)
(Al-2)
The total power of the source qo is obtained by integration of equation (Al-2). Substituting p={k* y,dp=-{k* dy and integrating from u = -00 to u = +00, gives:
from which
2]
(Al-3)
113
v' + dy/
It follows from equation (1-73) that an infinitesimal source located between will cause a small rise of temperature dTy' in point P at time t, as:
dqy'
v' and
dT ,y
(AI-4)
where
(r')=~(y_y')2
+z2, A=apc
and
dqy'
= q0,fii
{i1
exp] -k
* (y')2
]dy'
Integration of equation (AI-4) between the limits y' = -00 and v' = +00 gives:
(AI-5)
-y
+y
p
2-D heat flow
z
Fig. A1.2.Distributed heat source of net power density q(y) on a semi-infinite body.
114 where m
constant)and n
= 4at.
y'-
~~
2:J n
dy'
from which
2 y2 _ -u ------(y,)2 (y _ y,)2
m+n
and integrating between the limits u = -00 and u = +00. This gives (after some manipulation):
T-T
o
q0 /
'U
exp [
21tA~(t+t;)t
(A1-6)
If we replace the Gaussian heat distribution by a linear source of the same strength, which extends from -Lo to +L; on either side of the weld centre-line in the y-direction (see Fig. A1.3), we may write:
t"
o
(Lo)2
'Ita
(A1-7)
In practice, the parameter Lo has the same physical significance as the weaving amplitude L in equation (1-110). Consequently, these solutions are equivalent in the sense that they predict a similar temperature-time pattern.
115
qo ={i'q m IIk* V
f\
-y
-L
+y
Fig. AI.3.Physical
representation
2.1 Introduction
The weld metal composition is controlled by chemical reactions occurring in the weld pool at elevated temperatures, and is therefore influenced by the choice of welding consumables (i.e. combination of filler metal, flux, and/or shielding gas), the base metal chemistry, as well as the operational conditions applied. In contrast to ladle refining of metals and alloys where the reactions occur under approximately isothermal conditions, a characteristic feature of the arc welding process is that the chemical interactions between the liquid metal and its surroundings (arc atmosphere, slag) take place within seconds in a small volume where the metal temperature gradients are of the order of 1000 e mm ' with corresponding cooling rates up to 1000 e s'. The complex thermal cycle experienced by the liquid metal during transfer from the electrode tip to the weld pool in GMA welding of steel is shown schematically in Fig. 2.1. As a result of this strong non-isothermal behaviour, it is very difficult to elucidate the reaction sequences during all stages of the process. Consequently, a complete understanding of the major controlling factors is still missing, which implies that fundamentally based predictions of the final weld metal chemical composition are limited. Additional problems result from the lack of adequate thermodynamic data for the complex slag-metal reaction systems involved. However, within these restrictions, the development of weld metal compositions can be treated with the basic principles of thermodynamics and kinetic theory considered in the following sections.
0 0
(ii)
CHEMICAL REACTIONS IN ARC WELDING Contact tube Gas nozzle Shielding gas Filler wire Arc plasma temperature-1 aaooc Cold part of weld pool 1 900C) Electrode tip droplet (1600-2000C) Falling droplet (2400C) Hot part of weld pool (1900-2200C)
117
: ~
Base plate
Fig. 2.1. Schematic diagram showing the main process stages in GMA welding. Characteristic temperature ranges at each stage are indicated by values in parenthesis.
average
As indicated in Fig. 2.2 the high temperature stage comprises both gas/metal and slag/metal interactions occurring at the electrode tip, in the arc plasma, or in the hot part of the weld pool, and is characterised by extensive absorption of elements into the liquid metal. During the subsequent stage of cooling following the passage of the arc, a supersaturation rapidly increases because of the decrease in the element solubility with decreasing temperatures. The system will respond to this supersaturation by rejection of dissolved elements from the liquid, either through a gas/metal reaction (desorption) or by precipitation of new phases. In the latter case the extent of mass transfer is determined by the separation rate of the reaction products in the weld pool. It should be noted that the boundary between the two stages is not sharp, which means that phase separation may proceed simultaneously with absorption in the hot part of the weldpool. In the following sections, the chemistry of arc welding will be discussed in the light of this two-stage reaction model.
118
weldpool
t
Q)
cti CD
"0
~ ~
::;
Q)
CO 3:
0-
en
(5
Absorption of elements
t
0
e
'E
Q)
c:
]i
Q)
I
Peak concentration
g
Q)
E
"0
co
I
I
o c:
3:
en
(5
I I
Equilibrium concentration at melting point
-------------r---------------------------------I----
i
Time ---.
Fig. 2.2. Idealised two-stage reaction model for arc welding (schematic).
Table 2.1 Temperature for 90% dissociation of some gases in the arc column. Data from Lancaster.'
Gas CO2 H2 2 N2 Dissociation Temperature (K)
Next, consider a shielding gas which consists of two components, i.e. one inert component (argon or helium) and one active component X2- When the fraction dissociated is close to unity, the partial pressure of species X in the gas phase Px is equal to:
119
Px
=n
n +Xn PlOt.
(2-2)
where n] and nx are the total number of moles of components I (inert gas) and X, respectively in the shielding gas, and Ptot. is the total pressure (in atm). It follows from equation (2-1) that two moles of X form from each mole of X2 that dissociates. Hence, equation (2-2) can be rewritten as:
(2-3)
where n X2 is the total number of moles of component X2 which originally was present in the shielding gas. If nX2 and n] are proportional to the volume concentrations of the respective gas components in the shielding gas, equation (2-3) becomes: P x
X2)ptot.
(2-4)
= 1 atm, we obtain
)
=----~-2
100+vol%
2(vol%X
X2
(2-5)
Similarly, if X2 is replaced by another gas component of the type YX2, we get: (2-6) and Px= vol%YX2 100 + vol% YX2 (2-7)
It is evident from the graphical representations of equations (2-5) and (2-7) in Fig. 2.3 that the partial pressure of the dissociated component X increases monotonically with increasing concentrations of X2 and YX2 in the shielding gas. The observed non-linear variation of Px arises from the associated change in the total number of moles of constituent species in the gas phase due to the dissociation reaction. Moreover, it is interesting to note that the partial pressure Px is also dependent on the nature of the active gas component in the arc column (i.e. the stoichiometry of the reaction). This means that the oxidation capacity of for instance CO2 is only half that of O2 when comparison is made on the basis of equal concentrations in the shielding gas (to be discussed later).
120
>< a.
Vol% ~ , Vol% y~
Fig. 2.3. Graphical representation of equations (2-5) and (2-7).
2.4.1 Thinfilm model In cases where the rate of element absorption is controlled by a transport mechanism in the gas phase (step one), it is a reasonable approximation to assume that all resistance to mass transfer is confined to a stagnant layer of thickness 8 (in mm) adjacent to the metal surface, as shown in Fig. 2.4. Under such conditions, the overall mass transfer coefficient is given by:2 D k =----K.
a
(2-8)
where Dx is the diffusion coefficient of the transferring species X (in mm? S-I). Although the validity of equation (2-8) may be questioned, the thin film model provides a simple physical picture of the resistance to mass transfer during gas absorption.
121
Px I '-'-'-'-'-'-'r'-'-'-'-'-'-'-'-'-'-'-'I
I I I I I I I I I
I
I I I. I
Distance
Fig. 2.4. Film model for mass transfer (schematic).
2.4.2 Rate of element absorption Referring to Fig. 2.5, the rate of mass transfer between the two phases (in mol written as:
S-I)
can be
(2-9)
where A is the contact area (in mm- ), R is the universal gas constant (in mm ' atm K-l mot"), T is the absolute temperature (in K), Px is the partial pressure of the dissociated species X in the bulk phase (in atm), and is the equilibrium partial pressure of the same species at the gas/ metal interface (in atm). Based on equation (2-9) it is possible to calculate the transient concentration of element X in the hot part of the weld pool. Let m denote the total mass of liquid weld metal entering/ leaving the reaction zone per unit time (in g s').If Mx represents the atomic weight of the element (in g mol:"), we obtain the following relation when Px>
Px
Px :
[wt%X.]
I
(2-10)
It follows from equation (2-10) that the transient concentration of element X in the hot part of the weld pool is proportional to the partial pressure of the dissociated component X in the plasma gas. Since this partial pressure is related to the initial content of the molecular species X2 or YX2 in the shielding gas through equations (2-5) and (2-7), we may write:
122
Arc column
BulkQas phase
Metal phase
[wt%X. ]
I
=c
1(
(2-11)
and [wt%X]i
= C2 (
(2-12)
where C 1 and C2 are kinetic constants which are characteristic of the reaction systems under consideration.
(2-13)
By introducing the equilibrium constant K, for the reaction and setting the activity coefficient to unity, we get:
(2-14)
This equation should be compared with equation (2-10) which predicts a linear relationship
123
between wt% X and Px- If the above analysis is correct, one would expect that the partial pressure pCX at the gas/metal interface is directly proportional to the partial pressure of the dissociated component in the bulk phase. Unfortunately, the proportionality constant is difficult to establish in practice.
(2-15)
where
kd
pCX2
Metal phase Cold part of weld pool Fig. 2.6. Idealised kinetic model for gas desorption in arc welding (schematic).
124
The partial pressure pJc2 can be calculated from chemical thermodynamics the following reaction: 2K(dissolved) from which
by considering (2-16)
= X2 (gas)
(2-17) where K2 is the equilibrium constant, and [wt% X] is the concentration of element X in the liquid metal (in weight percent). Note that the activity coefficient has been set to unity in the derivation of equation (2-17). The equilibrium constant K2 may be expressed in terms of the solubility of element X in the liquid metal at 1 atm total pressure Sx. Hence, equation (2-17) transforms to:
(2-18)
(2-19)
Data for the solubility of hydrogen and nitrogen in some metals up to about 2200C are given in Figs. 2.7 and 2.8, respectively. It is evident that the element solubility decreases steadily with decreasing metal temperatures down to the melting point. This implies that the desorption reaction is thermodynamically favoured by the thermal conditions existing in the cold part of the weld pool. 2.6.2. Sievert's law
PX2
becomes kinetically
unfeasible
when
X] = SX~PX2
(2-20)
Equation (2-20) is known as the Sievert's law. This relation provides a basis for calculating the final weld metal composition in cases where the resistance to mass transfer is sufficiently small to maintain full chemical equilibrium between the liquid metal and the ambient (bulk) gas phase.
2.7 Overall Kinetic Model for Mass Transfer during Cooling in the Weld Pool
Because of the complexity of the rate phenomena involved, it would be a formidable task to derive a complete kinetic model for mass transfer in arc welding from first principles. How-
(ij
r------r---r--,---..--.--...,---,--,
Q) E
"C
CD E
(J)
"C
CD
"'N
.2
...... r.C\J
E 0.05
0.02
O.O!OL-O-""--- 6-..10-0-'----7-'-0-0-'----8...L0-0--"-9--100
:c
o ,...
0>
E
1500 2000
Temperature, C
Temperature, C
(c)
(d)
~.
CD
"C
s
CD
.2
o
o
C)
CD
(J)
40
E
"C
(J)
CD
0)
40
.2
o
o
~ r.C\J 20
~ 20 :cC\J
E
1500
1000
1000
1500
2000
Temperature, C
Temperature, C -----
Fig. 2.7. Solubility of hydrogen in some metals; (a) Aluminium, (b) Copper, (c) Iron, (d) Nickel. Data compiled by Christensen."
ever, for the idealised system considered in Fig. 2.9, it is possible to develop a simple mathematical relation which provides quantitative information about the extent of element transfer occurring during cooling in the weld pool. Let [%X]eq. denote the equilibrium concentration of element X in the melt. If we assume that the net flux of element X passing through the phase boundary A per unit time is proportional to the difference ([%X] - [%X]eq), the following balance is obtained.'
- V d[%X]
dt
= k d A([%X])
- [%X]
eq.
(2-21)
where V is the volume of the melt (in mm'' ), kd is the overall mass transfer coefficient (in mm s'), and A is the contact area between the two phases (in mm? ).
126
Temperature.vc
~~~~
t
z cfl. !
.Q
0)
-1.6
-2.0
-2.4
-2.8 ~ 4
a....- __
~,--
8 104rr,"K
12 .
16
)'
'0
Phase I >< :::s
><
Volume (V)
(5
ctS (j)
o c:
Q)
;:
Ci5
Concentration
Fig. 2.9. Idealised kinetic model for mass transfer in arc welding (schematic).
By rearranging equation (2-21) and integrating between the limits [%Xl; (at t = 0) and [%X] (at an arbitrary time t), we get: [%X] - [%X]eq. [%X]. I - [%X] eq.
] = exp [t-- ]
t
0
(2-22)
127
It is evident from the graphical representation of equation (2-22) in Fig. 2.10 that the rate of mass transfer depends on the ratio VI kfi, i.e. the time required to reduce the concentration of element X to a certain level is inversely proportional to the mass transfer coefficient kd This type of response is typical of a first order kinetic reaction. Although the above model refers to mass transfer under isothermal conditions, it is also applicable to welding if we assume that the weld cooling cycle can be replaced by an equivalent isothermal hold-up at a chosen reference temperature. Thus, by rearranging equation (222), we get:
[%x]
-[%X]
eq.
)exp(-~J to
(2-23)
It follows that the final concentration of element X in the weld metal depends both on the cooling conditions and on the intrinsic resistance to mass transfer, combined in the ratio tlt.; When [%X]eq. is sufficiently small, equation (2-23) predicts a direct proportionality between [%X] and [%X]i (i.e. the initial concentration of element X in the weld pool). This will be the case during deoxidation of steel weld metals where separation of oxide inclusions from the weld pool is the rate controlling step. Moreover, when ttt; 1 (small resistance to mass transfer), equation (2-23) reduces to:
[%X] = [%X]eq.
(2-24)
Under such conditions the final weld metal composition can be calculated from simple chemical thermodynamics. Because of this flexibility, equation (2-23) is applicable to a wide range of metallurgical problems at the same time as it provides a simple physical picture of the resistance to mass transfer during cooling in the weld pool.
. 1.0
..-... 0xQ)
t
I
0.8
:::;::
. xC
0.6
0XQ}
0.4
0.2
12
16
t ;s
Fig. 2.10. Graphical representation of equation (2-22).
128
129
I y~:yI Low I
Medium
High
Fig. 2.11. Ranking of different welding processes in terms of hydrogen level (schematic). The diagram is based on the ideas of Coe." (i) The Japanese method (JIS Z 313-1975), which has been adopted with important adjustments from the former ASTM designation A316-48T. This method involves collection of released hydrogen from a single pass weld above glycerine for 48h at 45C. The total volume of hydrogen is reported in ml per 100g deposit. Only 5 s of delay are allowed from extinction of the arc to quenching.
(ii) The French method (N.F.A. 81-305-1975) where two beads are deposited onto core wires placed in a copper mould. Hydrogen released from this bead is collected above mercury, and the volume is reported in ml per 100g fused metal (including the fused core wire metal). (iii) The International Institute o/Welding (IIW) method (ISO 3690-1977), where a single bead is deposited on previously degassed and weighed mild steel blocks clamped in a quickrelease copper fixture. The weldment is quenched and refrigerated according to a rigorously specified time schedule. Hydrogen released from the specimens is collected above mercury for 72 h at 25C, and the results are reported in ml per 100g deposit, or in g per ton fused metal. To avoid confusion, it is recommended to use the symbol H DM for the content reported in terms of deposited metal (ml per 100g deposit), and HFM for the content referred to fused metal (ml per 100 g or g per ton fused metal). The relationship between HDM and HFM is shown in Fig. 2.12. As would be expected, these three methods do not give identical results when applied to a given electrode. Approximate correlations have been established between the IIW criteria HDM and HFM and the numbers obtained by the Japanese and the French methods (designated HJIS and HFR, respectively). For covered electrodes tested at various hydrogen levels, we have:"
130
HDM
and
HFM
(2-25)
[HDM] :::: 1.27[HJ1s]
+ 2.19
(2-26)
The conversion factor from H FR to H FM applies to a ratio of deposited to fused metal, D/(B + D), equal to 0.6, which is a reasonable average for basic electrodes. The use of HFM in preference of HDM is normally recommended, because it is a more rational criterion of concentration. Moreover, H DM values would be grossly unfair, if applied to high penetration processes like submerged arc welding. In GTA welds made without filler wire HDM cannot be used at all, since there is no deposit. It should be noted that the present IIW procedure gives the amount of 'diffusible hydrogen'. For certain purposes the total hydrogen content may be wanted. It is obtained by adding the content of 'residual hydrogen' determined on the same samples by vacuum or carrier gas extraction at 650C. A very small additional amount may be observed on vacuum fusion of the sample, tentatively labelled 'fixed hydrogen'. There is no clear line of demarcation between these categories of hydrogen. As will be discussed later, the extent of hydrogen trapping depends both on the weld metal constitution and the thermal history of the metal. In singlebead basic electrode deposits the diffusible fraction is usually well above 90%. 2.8.3 Reaction model Normally, measurements of hydrogen in weld metals are carried out on samples from solidified beads. Due to the rapid migration of hydrogen at elevated temperatures, such data do not represent the conditions in the hot part of the weld pool. Quenched end crater samples would be better in this respect, but they are not representative of normal welding. Further complications arise from the presence of hydrogen in different states (e.g. diffusible or residual hydrogen) and the lack of consistent sampling methods. Nevertheless, experience has shown that pick-up of hydrogen in arc welding can be interpreted on the basis of the simple model outlined in Fig. 2.13. According to this model, two zones are considered: (i) An inner zone of very high temperatures which is characterised by absorption of atomic hydrogen from the surrounding arc atmosphere.
CHEMICAL REACTIONS IN ARC WELDING Electrode Cold part of weld pool Desorption of hydrogen (controlled by PH2 in ambient gas phase)
131
Weld pool
Fig. 2.13. Idealised reaction model for hydrogen pick-up in arc welding.
(ii) An outer zone of lower temperatures where the resistance to hydrogen desorption is sufficiently small to maintain full chemical equilibrium between the liquid weld metal and the ambient (bulk) gas phase. Under such conditions, the final weld metal hydrogen content should be proportional to the square root of the initial partial pressure of diatomic hydrogen in the shielding gas, in agreement with Sievert's law (equation (2-20). 2.8.4 Comparison between measured and predicted hydrogen contents It is evident from the data in Table 2.2 that the reported ranges for hydrogen contents in steel weld metals are quite wide, and therefore not suitable for a direct comparison of prediction with measurement. For such purposes, the welding conditions and consumables must be more precisely defined. 2.8.4.1 Gas-shielded welding In GTA and GMA welds the hydrogen content is usually too low to make a direct comparison between theory and experiments. An exception is welding under controlled laboratory conditions where the hydrogen content in the shielding gas can be varied within relatively wide limits. The results from such experiments are summarised in Fig. 2.14, from which it is seen that Sievert's law indeed is valid. A closer inspection of the data reveals that the weld metal hydrogen content falls within the range calculated for chemical equilibrium at 1550 and 2000C, depending on the applied welding current. This shows that the effective reaction temperature is sensitive to variations in the operational conditions. An interesting effect of oxygen on the weld metal hydrogen content has been reported by Matsuda et al. 9 Their data are reproduced in Fig. 2.15. It is evident that the hydrogen level is significantly higher in the presence of oxygen. This is probably due to the formation of a thin (protective) layer of slag on the top of the bead, which kinetically suppresses the desorption of hydrogen during cooling.
132
Table 2.2 Measured arc atmosphere compositions in steel welding. Also included are typical ranges for the weld metal hydrogen content. Data compiled by Christensen."
Arc Atmosphere Composition (vol%) Method Primary Source of Hydrogen Moisture introduced through the shielding gas Firmly bound water in the electrode coating Firmly bound water in the electrode coating Loosely bound water in the electrode coating Firmly bound water in flux Loosely bound water in flux Loosely bound water in flux Weld Metal Hydrogen Content (ppm) Range Average
CO2
CO
H2+ H2O
GMAW* (CO2) SMAW (acid) SMAW (rutile) SMAW (basic) FCAW (rutile) FCAW (basic) SAW (basic)
3 25 25 3-5
-
can vary within wide limits, depending on the operational conditions applied.
t
E
"C
Q)
14
GTAW (low-alloy steel)
J:
~ u,
10 8 6
o !=100
01=200 A
S
~
Q) U)
a a ,.......
C)
r.N 4
E
2
0
6
...
Fig. 2.14. Hydrogen pick-up in GTA welding at different concentrations of H, in the shielding gas. Data from Chew and Willgoss.8
133
t
r.='
~ E
"0
8.0
(/)
6.0
Ar+2.5 vol% H2
(l)
Q)
.2 4.0
C)
"""-
en
o ,...
r.
C\I
2.0
___
.-
- Ar+0.3 vol% H2
0.01
0.02
0.03
0.04
0.05
0.06
Weld metal oxygen content, wt% ---Fig. 2.15. Hydrogen pick-up in GTA welding at different levels of oxygen in the weld metal. Data from Matsuda et al.9 . Example (2.1)
Consider GTA welding (Ar-shielding) on a thick plate of low-alloy steel under the following conditions: 1= 200A, U
= 15V, v = 3 mm
The shielding gas contains 0.1 vol% moisture (H20) and is supplied at a rate of 15NI min-I. Calculate the 'potential' hydrogen level, assuming that all hydrogen introduced through the shielding gas is absorbed in the weld metal.
Solution
10-3 x l1(NI 60
H20)_1_(mol s 22.4
H20J!(-S-)2.02( NI H20 3 mm
gH J mol H20
= 7.5.10-6
gHmm-I
The resulting bead cross section and total mass of weld metal per mm can be estimated from the Rosenthal equation by considering the dimensionless operating parameter at the melting point (equation (1-50)):
n
3
= 200x15xO.5x3
41tx(5)2x7.5
=1.91
134
mm
~14mm
Taking the density of the steel equal to 7.85 X 10-3 g mrrr ', we obtain: pB = 7.85 X 10-3 X 14 g mm"! = 0.11 g mm ' The 'potential' hydrogen level is thus:
7.5 5 ppm --ppm:::::7 0.11
It is evident from the above calculations that the 'potential' hydrogen level is at least one order of magnitude higher than the expected weld metal hydrogen content (1 to 3 ppm). This shows that the hydrogen pick-up in GTA welding is not determined by the total amount of hydrogen which is introduced through the shielding gas, but is mainly controlled by the resulting partial pressure of hydrogen in the ambient (bulk) gas phase. 2.8.4.2 Covered electrodes In SMA welding the partial pressure of hydrogen is more difficult to assess due to the presence of trapped moisture and hydrogen-containing compounds in the electrode coating. Such compounds will loose their identity at the stage of introduction into the arc atmosphere. Since very little information is available on the species present in the arc column, we shall base our estimate on a simple thermodynamic approach, including only the molecular species H2 and H20 which can be determined by analysis (see data in Table 2.2). It follows that the combined partial pressure of H2 and H20 in the gas phase is given by: (2-27) The parameter Pw can be estimated on the basis of combustion measurements of the electrode coating, assuming that no carbon is picked up or lost from the system in excess of the amount calculated from an analysis of the base plate and the electrode wire. For a recorded content of mw g H20 and me g CO2 per 100 g of electrode coating, we obtain:
(2-28) From a thermodynamic standpoint, replacement of P H2 by Pw in the expression for Sievert's law requires the use of a modified solubility of hydrogen, defined as: K
H
S =S
w
K3 +[%0]
(2-29)
where K3 is the equilibrium constant for the H20-H reaction, and [%0] is the weld metal oxygen content. In practice, the correction term ~ K3 / (K3 + [%0]) does not depart significantly from unity, which means that S; ~ SHe
135
During welding with basic covered electrodes considerable amounts of CO2 may form as a result of decomposition of calcium carbonate, according to the reaction: CaC03
= CaO
+ CO2
(2-30)
Modern basic electrodes contain between 20 to 40 weight percent CaC03, which is equivalent with a CO2 content of 9 to 18 percent. Taking as an average me equal to 15 g CO2 per 100 g electrode coating, we obtain:
[%H]=Sw~W
mw
+ 6.2
(2-31)
In Fig. 2.16 the validity of equation (2-31) has been checked against relevant literature data (compiled by Chew 10). A closer inspection of the data reveals that the weld metal hydrogen content falls within the range calculated for chemical equilibrium at 1520 to 2000C, taking Sw equal to the solubility of hydrogen in pure iron at the indicated temperatures (i.e. 27 and 40 ml H2 per 100 g fused metal, respectively). Although the observed scatter in the effective reaction temperature is admittedly large, equation (2-31) points out a very interesting effect, namely that the hydrogen content of SMA steel weld metals is controlled by the combined partial pressure of H2 and H20 in the ambient gas phase. For this reason it is frequently recommended that calcium carbonate is added to the electrode coating, which on decomposition produces considerable amounts of shielding gas in the form of CO2, Hydrogen shielding can also be achieved by additions of volatile alkali-fluorides, which on heating will evaporate and dilute the atmosphere with respect to hydrogen.
t
~
16
J: 12
LL
S
Q)
E
Q)
en .a
C)
0 0
,-
:c
C\I
Water content in electrode coating, wt% ---Fig. 2.16. Hydrogen pick-up in SMA welding at different water contents in the electrode coating. Data compiled by Chew. 10
136
METALLURGICAL
MODELLING
OF WELDING
Example (2.2)
Consider SMA welding on mild steel with basic covered electrodes. The electrode coating contains 35 wt% CaC03 and 0.5 wt% H20 in the as-received condition. After drying at 350C for 1 h the water content is reduced to 0.2 wt% H20. Estimate the weld metal hydrogen content (in ppm) both before and after drying of the electrode. Assume in these calculations an effective reaction temperature of 1800C.
Solution
First we calculate the CO2 content per 100 g of electrode coating. Taking the atomic weight of CaC03 and CO2 equal to 100.1 and 40.0, respectively, we obtain:
me =--35g=14g
The combined partial pressure Pw can now be estimated from equation (2-28). drying we have: ___ 0_.5 atm = 0.080 atm 0.5+(0.41 x14) After drying of the electrode, the partial pressure Pw becomes: 0.2 0.2 + (0.41 Before
40 100.1
Pw
x 14) atm
0.034 atm
From Fig. 2.7 (c) it is evident that the solubility of hydrogen in liquid iron at 1800C is about 37 ml H2 per 100 g fused metal. This corresponds to a modified solubility Sw (in ppm) of:
s, ~SH
= O.37(mi H2
g Fe
ton
Fe) ~ 33ppm
Substituting this value into the expression for Sievert's law gives: [H]
ppm
= 33
V;;
F;
={
33 33
\ID.034 ~ 6 ppm
\IQ.OsO ~ 9 ppm
It follows from the above calculations that a low weld metal hydrogen level requires the use of 'dry' basic electrodes. In practice, this can be achieved by protecting the electrodes against moisture pick-up during storage (see Fig. 2.17). However, in certain cases it is necessary to differentiate between strongly bound and loosely adsorbed moisture in the coating of basic electrodes. This point is more clearly illustrated in Fig. 2.18, which shows the H DM content of hydrogen in basic electrode deposits at various levels of coating moisture. It is seen that water remaining from an insufficient baking treatment is more dangerous than moisture picked up by exposure of a properly dried coating. This has to do with the fact that loosely adsorbed mois-
137
i
Q)
~8 0.8 ~
~ 0.6
c:
c;
CD .5
C S c: o
o
0.2
Exposure time, days -----Fig. 2.17. Moisture content in basic electrode coating as a function of exposure time and relative humidity (R.H.) in ambient gas phase. Data from Evans. 11
SMAW(low-alloy steel)
E ::J :0 E
Q)
Fig. 2.18. Hydrogen pick-up in SMA welding at different levels (states) of adsorbed water in the electrode coating. Data from Evans and Bach.12
138
ture will tend to evaporate during the welding operation (before it enters the arc column) because of resistance heating of the electrode, a process which is not feasible when the water is bound in rust on the surface of the electrode wire or the iron powder. 2.8.4.3 Submerged arc welding This method is usually classified as a pure slag-shielded process, because carbonates or other gas-producing compounds are not present in large quantities. A closed arc cavity does exist, however, as indicated by the falling volt-ampere curve characteristic of open arcs, and by observations made by probes inserted through the flux cover. It is reasonable to assume that the gas contained within this enclosure consists of metal vapour, volatile constituents originating from the flux, and relatively small fractions of carbon monoxide and water vapour. Acid fluxes of the calcium silicate type will probably generate silicon monoxide, while agglomerated fluxes bonded with alkali silicate will produce volatile alkali fluorides. In addition, carbon monoxide may be present as a result of oxidation of carbon, or decomposition of carbonates. A small but important contribution to the cavity atmosphere is the trace of moisture remaining in the flux even after careful drying. No direct measurements of partial pressures are available, and the gas composition must therefore be inferred from observations of hydrogen absorption in the weld metal. Hydrogen pick-up during SA welding has been examined by Evans and Bach.I? Their data are replotted in Fig. 2.19. The shape of the observed curve of hydrogen vs residual water content would seem to indicate a relationship similar to that predicted by Sievert's law. In fact, a very close fit can be obtained through empirical calibration of the dilution term in equation (2-28). This, however, implies unreasonable amounts of CaC03. Carbon monoxide in addition to that delivered by carbonates could be formed by oxidation of carbon. Again, an unreasonable amount of carbon loss would be required. Therefore, it must be concluded that further research is needed for a proper interpretation of the factors controlling hydrogen pick-up in SA welding.
30 25
10
::E
Q
:c
S .B
I
E a.
.~
....: 20
~ L1.
c. ~ 15 o
6 "E .!
c
0 0
.....
o .,-
:cC\J 10
4
"0
c Q)
0)
E
5 2
0
J:
>,
0.1
0.5
Fig. 2.19. Hydrogen pick-up in SA welding at different water contents in the flux. Data from Evans and Bach.12
CHEMICAL
REACTIONS
IN ARC WELDING
139
Example (2.3)
Consider SA welding on a thick plate of low-alloy steel under the following conditions: 1= SOOA, U = 30V, v = 8mm
S-I, 'T) =
0.95, To = 20C
The flux contains 0.04 wt% H20 and is consumed at a rate of 0.6 g per g weld deposit. Estimate both 'potential' and 'equilibrium' hydrogen levels when the total oxidation loss of carbon in the weld pool is 0.03 wt%.
Solution
First we calculate the total amount of fused parent metal and weld deposit formed on welding. From equations (1-75) and (1-120), we have: B = !!:.(r* )2 2 m and D=
mm? = 108mm2
= 44mm2
When the dilution ratio D/(B + D) is known, it is possible to calculate the total flux consumption per gram fused weld metal: 0.6 D (B+D)
= 0.6
44 (108+44)g
= 0.17
pp
pp
If we assume that all CO produced by reactions between dissolved carbon and oxygen is infiltrated in the arc column, the following balance is obtained: Total number of moles of CO per g fused weld metal: 0.03 I -1 25 0 10-6 1-1 -10-0-x-I-2 mo g = . x mo g Total number of moles of H20 per g fused weld metal: 0.17 x O.04 mo Ig -1 = 3 . 8 X 10-6 mo 1-1 ---g 100 x 18
p
w
(25
= 0.132atm
Since the effective reaction temperature of hydrogen absorption in SA welding is not known, the maximum solubility of hydrogen at 1 atm total pressure is taken equal to 33 ppm, similar to that in the previous example. By inserting this value in the expression for Sievert's law, we obtain: [H]ppm = 33-V0.132 ~ 12 ppm In practice, the 'potential' hydrogen level represents an upper limit for the hydrogen concentration which cannot be exceeded. Thus, the contradictory results obtained in the present example clearly illustrate the difficulties involved in estimating the effective partial pressure of hydrogen in SA welding. 2.8.4.4 Implications of Sievert's law An important implication of Sievert's law is that the fraction of hydrogen picked up from the arc atmosphere is very high at low hydrogen pressures: (2-32) As seen from equation (2-32), the first traces of hydrogen added to the atmosphere are completely absorbed in the metal. At increasing partialpressures the fraction of hydrogen picked up in the metal will gradually decrease, finally attaining a threshold of (SB12) in the case of pure H2. This shows that the concept of 'potential' hydrogen content frequently used to characterise filler materials (see Fig. 2.11) is a dangerous one, since the rates of absorption are so different in the high and low ranges of the hydrogen potential. 2.8.4.5 Hydrogen in multi-run weldments So far, no standardised method is available for the determination of hydrogen in multi-layer welds. Early measurements by Roux.P using an arrangement similar to that subsequently adopted in French standards, indicate a constant ratio of extracted hydrogen to the mass of fused metal, regardless of the number of passes. If hydrogen is reported on the basis of deposited metal, this ratio may vary by a factor of 2.5 when comparing a deposit made in five passes to a single bead. Exploratory measurements of local hydrogen contents in large-size joints have been made by Skjolberg.!" who butt welded a 40 mm plate with a self-shielding flux cored wire at an interpass temperature of 200C. Samples were cut from a refrigerated part of the weldment at mid-thickness, including positions in the weld metal close to the fusion line and samples in the HAZ. His results are summarised in Table 2.3. Normal testing of the filler wire according to ISO 3690 gave fused metal hydrogen contents of 3.3 ppm (diffusible) and 1.7 ppm (residual). A comparison with Table 2.3 shows that the multi-run content of diffusible hydrogen is much lower than the corresponding ISO value, probably as a result of a high interpass temperature which facilitates loss of hydrogen to the surroundings through diffusion.
141
Table 2.3 Measured hydrogen contents in multi-run FeA steel weldment. Data from Skjolberg.!"
HAZ Distance from fusion line (mm) Condition As-welded Weld Metal 0.6 ppm diffusible 0.9 ppm residual 0.35 ppm diffusible 2.25 ppm residual
o to 5
0.25 ppm diffusible 0.15 ppm diffusible
10 to 15
-
2.8.4.6 Hydrogen in non-ferrous weldments The solubility of hydrogen in metals and alloys of industrial importance increases with temperature, and passes through a maximum in the vicinity of the boiling point, where the opposing trends of increasing solubility and increasing dilution by metal vapour balance. Solubility curves for hydrogen in aluminium, copper, and nickel up to about 2200C have previously been presented in Fig. 2.7. Since all these metals can dissolve considerable amounts of hydrogen, the risk of hydrogen absorption during welding is imminent if moisture is present in the shielding gas. Results obtained from arc melting experiments with Cu, AI, Ni in Ar-H2 gas atmospheres indicate that hydrogen is absorbed at a high temperature zone under the arc and is transported by fluid flow to the outer, cooler regions of the pool.l" Rejection of the gas in the supersaturated outer regions is slower than the absorption in the hot zone, so the gas content throughout the pool approximates to that in the absorption zone. Typical estimates of the effective reaction temperature of hydrogen desorption (based on the Sievert's law) gave the following result: 15 Copper: Aluminium: Nickel: At present, it is not known whether these reaction temperatures also apply to conventional GTA or GMA welding of the same materials or are mainly restricted to the operational conditions employed in the arc melting experiments. 1900C
142
2.9.1 Sources of nitrogen Since the total nitrogen level in most welding consumables and shielding gases is quite low, the main source of nitrogen contamination is air infiltrated in the arc column. For this reason, the weld metal nitrogen content is very sensitive to variations in the operational conditions (e.g. arc length, electrode stick-out, shielding gas flow rate etc.). The overall reaction of nitrogen absorption is similar to that of hydrogen: N2 gas = 2~(dissolved) By introducing the equilibrium constant K4 for the reaction, we get: (2-33)
(2-34)
where SN is the maximum solubility of nitrogen at 1 atm total pressure.jj, is the activity coefficient, and PN2 is the resulting partial pressure of diatomic nitrogen in the gas phase. The solubility of nitrogen in liquid iron is approximately given by: log SN(ppm) = 2.75 - 188/T (2-35)
where T is the temperature in K. At 1600 and 2000C, this equation gives equilibrium concentrations of 446 and 465 ppm, respectively. In alloyed steel containing large amounts of nitride-forming elements (e.g. austenitic stainless steel), the activity coefficient of nitrogen fN is about 1/4 and hence, the solubility will be about 4 times higher than that calculated from equation (2-35). From a primitive model of pseudo-equilibrium between gaseous N2 and dissolved N a maximum solubility of about 465 ppm would be expected in welding under 1 atm total pressure. Thus, the maximum pick -up of nitrogen in deposition of bare wire in air would be of the order of 465-i<l8 ppm or 416 ppm. If a tentative estimate of air infiltration in the arc column is made at 1 vol% N2, the expected pick-up of nitrogen would be 465..)0.01 or about 47 ppm. A comparison with the data in Table 2.4 shows that the measured weld metal nitrogen contents are much higher than predicted from Sievert's law. This implies that the mechanism of nitrogen desorption is different from that of hydrogen. 2.9.2 Gas-shielded welding Information on the factors controlling nitrogen pick-up may be obtained from the work of Kobayashi et al., 16 who examined the GMA welding process in a systematic manner. Some of their results are shown in Fig. 2.20. Figure 2.20(a), for low-alloy steel, reveals that the square root relationship is a fair approximation only for welding in mixtures of N2 and H2 (curve No.5). Mixtures of N2 + Ar (curve No.3), N2 + CO2 (curve No.2) and N2 + O2 (curve No.1) show increasing deviation from the predicted behaviour. Pure N2 under reduced pressure gives a curve (No.4) of an entirely different shape including a maximum at p N ~ 0.05.
2
143
Table 2.4 Summary of measured weld metal nitrogen contents. Data compiled by Christensen."
Welding Method SMA W (basic electrodes) Material Low-alloy steel Stainless steel SMAW (rutile electrodes) Low-alloy steel Stainless steel SAW FCAW GMAW Low-alloy steel Low-alloy steel Low-alloy steel Nitrogen Content (ppm)
Similar features are seen from Fig. 2.20(b) for welding of stainless steel. Again, the deviation becomes more pronounced as the oxidation potential of the gas mixture is increased in the sequence H2-Ar-C02-02. Moreover, a comparison with Fig. 2.20(a) reveals that the displacement of the nitrogen concentrations in the presence of chromium is larger than expected from the calculated reduction of the nitrogen activity coefficient. The trends observed in Fig. 2.20 have been confirmed by O'Brien and Jordan'? who studied nitrogen pick-up during CO2-shielded welding of low-alloy steel. As can be seen from Fig. 2.21(a) their curves are similar to those of Kobayashi et al.16 for short circuiting metal transfer, while a mixed spray/globular transfer gives a sharp rise of nitrogen absorption up to PN2 = 0.3 followed by a constant or slightly decreasing concentration (Fig. 2.21(b. Both patterns are clearly not in accordance with predictions based on Sievert's law (equation (2-34)). An interpretation of the observed trends should be made with a view to absorption of hydrogen, where the concept of pseudo-equilibrium has proved useful for a semiquantitative prediction. In both cases the molecular species H2 and N2 are known to dissociate in the arc column (see Table 2.1), and would therefore dissolve in the metal to an extent far beyond the solubility controlled by PH or P N The excess of dissolved hydrogen is probably released as 2 2 gas at weld pool temperatures. This will also be the case with nitrogen in the absence of oxygen, as shown previously in Fig. 2.20(a) and (b). However, under oxidising conditions the desorption of gaseous nitrogen becomes suppressed by the presence of oxygen at the gas/metal interface and hence, nitrogen is retained at a level which by far exceeds the solubility limit at 1 atm total pressure of Nj. This has been confirmed experimentally by Uda and Ohnol'' in their classic work on surface active elements (Le. oxygen, sulphur and selenium) in liquid steel. A similar phenomenon was quoted in Section 2.8.4.1 from the work of Matsuda et al. 9 even in the case of hydrogen, where increased entrapment of hydrogen was observed in the presence of oxygen (see Fig. 2.15). It appears thus that excessive absorption of nitrogen (and in some cases also hydrogen) should be interpreted as a state of incomplete release of solute, as described previously in Sections 2.6 and 2.7. As a consequence, Sievert's law cannot be used for an estimate of nitrogen pick-up in steel welding, unless the weld metal oxygen content is extremely low. 2.9.3 Covered electrodes The nitrogen content of SMA weld deposits is known to be sensitive to variations in the arc
144 (a)
,
~ 0 ~
c: .m c: 0 c c
Q)
0.14 0.12
...;- 0.10
0.08
C) :t:
..
0
0.04 0.02 0 0 20 40
60
80
100
(b)
0.4
t
~ ~
0
0.3
+oS Q) ....
c:
c:
0
Q)
o c:
0.2
z
+J
0)
0.1
1.
p~ +P02 = 1 atm
PAr
= = 1 atm
5. PN2 +P~
= 1 atm
100
20
40
60
80
(a)
0.08
145
0.07
t
...;'
Q) ....
0.06
~ 0.05
3:
c::
c:
<D
8
c:
0.04
~ 0.03
C)
0.02
Experiment 1 A 0.40 kJ/mm 2 0.49 kJ/mm
3 0.52 kJ/mm
0.01
20 (b)
100
t
tfl.
0.05
~ 0.04
c
$
c:
c
Q)
0.03
0')
0.02
Experim~nt 5 0.8 kJ/mm 6 .. 1.1 kJ/mm 7 c 1.6 kJ/mm
0.01
Fig. 2.21. Nitrogen pick-up in GMA welding at different concentrations of N2 in the shielding gas; (a) Short circuting metal transfer, (b) Mixed and free flight metal transfer. Data from O'Brien and Jordan. 17
20
100
146
length (voltage) because of the risk of air infiltration in the arc column. This point is more clearly illustrated in Fig. 2.22, which shows that the resulting weld metal nitrogen level may vary significantly from one weld to another, depending on the operational conditions applied. Consequently, the use of long arcs in SMAW should be avoided in order to prevent excessive pick-up of nitrogen from the surrounding atmosphere. 2.9.4 Submerged arc welding In submerged arc welding the risk of air infiltration in the arc column is less imminent, since welding is performed under the shield of a flux. Hence, in multipass welds the filler wire itself will be the main source of nitrogen (see Fig. 2.23), while in single pass weldments the base plate nitrogen content is more important because of the high dilution involved. The latter point is illustrated by the following numerical example.
Example (2.4)
Consider SA (single pass) welding on a thick plate of low-alloy steel under the following conditions: I = 600A, U = 30V, v = 6mm s','Yl= 0.95, To = 20C Based on the 'rule of mixtures' , calculate the weld metal nitrogen content. Assume in these calculations that the nitrogen content of the base plate and the filler wire is 0.005 and 0.012 wt%, respectively.
E Co.
250------------------------------------~
SMAW (low-alloy steel) 4 and 5 mm basic covered electrodes 200
E
+-"' C
Q)
150
.e ...
o o ~ 100
0>
50
Welder No.
Fig. 2.22. Natural fluctuations in nitrogen pick-up during "SMA welding due to variations in the arc length. Data from Morigaki et al.)?
147
120
E Co.
E S c:
0 0
Q)
c:
C)
'2 Q)
g
E
(ij
"C
20
20
40
60
80
100
120
140
160
Fig. 2.23. Nitrogen pick-up in SA welding at different levels of nitrogen in the electrode wire. Data from Bhadeshia et 20
az.
Solution
First we calculate the total amount of fused parent metal and weld deposit formed on welding. From.equations (1-75) and (1-120), we have:
B = ~(r* 2 and
)2 m
The 'rule of mixtures' gives us the nominal weld metal nitrogen content, which is defined as: [wt%N] B. D
nom.
= (B+D)
[wt%N]plate
+ (B+D) [wt%N]wzo"e
I
173 (173+70)
[0.005]+
70 (173+70)
[0.012]=0.007%
The above calculations show that the nitrogen content of single pass SA steel welds is close to that of the base plate because of the high dilution involved. This is in agreement with general experience.
148
2.10 Absorption
of Oxygen
Partial oxidation almost invariably accompanies the welding of steel. It is well established that considerable interaction takes place between the liquid weld metal and its surroundings (arc atmosphere, slag) when welding is performed in the presence of oxygen. For slag-protected processes, the flux is the main source of oxygen because of its content of easily-reduced oxides, such as iron oxide, manganese oxide, silica, or rutile. In gas metal arc (GMA) welding, oxygen is often deliberately introduced through the shielding gas to improve the arc stability and bead morphology, but at the expense of an increased oxygen content in the weld metal and intensified losses of alloying elements. The oxidation reactions proceed very rapidly under the prevailing conditions owing to the high metal temperatures and the large interfacial contact area available for interactions. A general survey of oxygen contents in fusion welds is shown in Table 2.5. 2.10.1 Gas metal arc welding The GMA welding process offers a special advantage in the way that it allows the reactions to be studied through variations of the shielding gas composition without the complicating presence of a flux. Gas-shielded welding is also interesting form a practical point of view, since the process is readily mechanised and can be applied for welding of nearly all types of steel. Moreover, many of the reactions in the weld pool will be similar to those occurring during welding with covered electrodes or submerged arc fluxes. The amount of oxygen transferred to the metal during the high-temperature stage will clearly depend on the atmosphere. In GMA welding with active gases such as CO2, Ar + O2 or Ar + O2 + CO2 the extent of element absorption is controlled by the oxygen potential of the shielding gas. This type of gas/metal interaction will be discussed below. The analytical weld metal oxygen content is also affected by the presence of deoxidisers, because the transition from the high-temperature stage to the cooling stage is not sharp. As was
Table 2.5 Summary of measured weld metal oxygen contents. Data compiled by Christensen."
Welding Method SMA W (basic electrodes) SMAW (rutile electrodes) SMA W (acid electrodes) GTAW (pure Ar or He) GMAW (Ar-02 mixtures) GMAW(C02) FCAW (self-shielded cored wire) Material Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Low-alloy steel Oxygen Content (ppm) 250-440 500-1300 700-1400 50-100 200-1300* 500-800 130-200 450-2600 400-750 180-350
SAW (calcium silicate fluxes) SAW (manganese silicate fluxes) SAW (basic fluxes)
149
discussed briefly in the introduction to Chapter 2, a separation of non-metallic inclusions may occurtowards the end of stage one and possibly at the beginning of stage two. A necessary condition for removal of oxygen introduced into the metal is, of course, the formation of oxygen-rich phases and compounds (e.g. microslag or carbon monoxide). These matters will be considered in a separate section on weld metal deoxidation reactions. 2.10.1.1 Sampling of metal concentrations at elevated temperatures In order to understand the extent and direction of the oxidation reactions, it is not sufficient to characterise the initial and final conditions (i.e. consumables/parent plate and the weld metal). Accordingly, especially designed experiments are required for assessment of the reactions taking place at the four main process stages: ( 1) Electrode tip (2) Arc column (3) Hot part of the weld pool (4) Cold part of the weld pool. Proper sampling techniques are needed for basic studies of this kind, which allow sampling of the falling droplets during their flight through the arc column. An additional requirement is that the speed of quenching is sufficiently high to freeze-in the metal composition established at elevated temperatures. If not, spontaneous reactions and subsequent losses of dissolved elements due to CO gas formation and manganese silicate slag precipitation may take place on cooling. To overcome these problems, a special 'melt spinning' technique has been developed by Grong and Christensen;' utilising the same principles as those employed in production of amorphous alloy ribbons. By using a water-cooled, fast rotating copper wheel (spinner) as a cathode, rapid crystallisation of droplets released from the electrode tip can be obtained in the absence of a weld pool *. The reactions which normally occur in the pool during regular multirun deposition may then be assessed by comparison of analytical data for chilled metal (i.e. falling droplets) and normal multi-layer weldment, respectively. 2.10.1.2 Oxidation of carbon Filler wires intended for GMAW are usually overalloyed with respect to carbon, silicon, and manganese to compensate for heavy oxidation losses of these elements during the welding operation. It is evident from the data presented in Fig. 2.24(a) and (b) that virtually identical carbon contents are obtained for chilled and multilayer weld metal, except when welding is performed in extremely oxidising atmospheres (beyond 20 vol% O2 in Ar). This implies that oxidation of carbon does not normally take place in the weld pool, but is located to one or both of the preceding stages, i.e. electrode tip or arc column. However, since the carbon oxidation at the electrode tip is more than sufficient to account for the observed total loss t, it is likely that little, if any, carbon is lost in the latter stage. The fact that carbon is found to oxidise readily at the electrode tip but not in later stages, indicates that the supply of oxygen and/or the conditions of nucleation are more favourable in
*The estimated cooling rate is of the order of 105oC
S-I
or higher.
tThe measured electrode tip carbon content also includes oxidation after arc extinction. This accounts for the observed loss of carbon beyond that recorded in chilled metal or multi-layer weld deposit.
150
10
15
20
25
~
30
Vol% 02 in Ar
(b)
A Electrode tip Chilled metal (falling droplet) Multi-rayer weld metal Electrode wire
20
40
60
80
100
Vol% CO2 in Ar
Fig. 2.24. Measured carbon contents in electrode tip droplets, chilled metal and multi-layer weld deposit vs the oxygen potential of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-C02 gas mixtures. Data from Grong and Christensen. 1
151
the former case. Normally, homogeneous nucleation of CO gas within the liquid metal is considered impossible, which means that the CO nucleation in practice must take place heterogeneously. However, the most probable site for CO evolution during droplet formation at the electrode tip will be the gas/metal interface itself, which allows carbon to be oxidised simultaneously with silicon and manganese. It is reasonable to assume that most of the observed carbon oxidation is located to the hot layers facing the arc, where the reaction is thermodynamically favoured. At other surface positions Si and Mn are expected to prevent carbon from reacting due to a rather low metal temperature, stated to be only slightly above the melting point at the time of detachment. 21 It is evident from the data in Fig. 2.24(a) and (b) that the carbon losses increase with increasing O2 or CO2 contents in the shielding gas up to a certain critical level. Hence, supply of oxygen to the tip droplet surface is the rate controlling step for oxidation of carbon at low oxygen potentials. This conclusion is also consistent with calculations made by Corderoy et al.,22 who found that transport of atomic oxygen through a stagnant gaseous boundary layer close to the metal surface controls the oxidation rate of alloying elements at this stage of the process. The carbon oxidation will gradually decline with increasing oxygen concentrations in the shielding gas, probably as a result of build-up of carbon monoxide in the surrounding atmosphere. When the critical CO gas pressure is reached, the carbon reaction is blocked, silicon (and manganese) now exercising control of the oxygen level, as indicated in Fig. 2.25. For ArO2 gas mixtures this criticalpco pressure is attained at about 10 to 15 vol% O2 in the shielding gas, corresponding roughly to 0.05 wt% C oxidised in chilled metal. When welding is performed in Ar-C02 mixtures the reaction is blocked at a much earlier stage of carbon oxidation (equal to about 0.02 wt% C lost in chilled metal), since dissociation of CO2 in this case will produce an additional amount of CO to concentrate in the surrounding gas phase.
t
~
1.2
<a 0.8
0
e +-'
C 0 0
0.4
1500
Fig. 2.25. Break even equilibrium partial pressure of CO vs temperature for silicon control of oxygen level at 0.8 wt% Si and silica saturation. Data from Elliott et al. 23
152
The mechanism indicated above is supported by the data presented in Fig. 2.26, which show that a CO2-rich atmosphere even may act carburising if the initial carbon content of the electrode wire is sufficiently low. Moreover, it is interesting to note that the carbon-oxygen reaction is also influenced by the rate of droplet detachment. Since the highest carbon oxidation losses are normally associated with a coarse globular droplet transfer mode, this suggests that the reaction time is more important than the effective contact area available for interaction which depends on the droplet size. 2.10.1.3 Oxidation of silicon It is evident from the data in Fig. 2.27 that loss of silicon mainly take place in the weld pool, as indicated by the difference between the measured silicon content in chilled and multi-layer weld metal. In Ar-02 mixtures the Si loss increases steadily with increasing oxygen potential of the shielding gas. Thus, at 30 vol% O2 in Ar it amounts to 0.59 wt% Si (or 0.67 wt% 0) removed from the weld pool as a result of deoxidation reactions. A similar situation exists in the case of CO2-shielded welding up to about 20 vol % CO2 in Ar. At higher CO2 contents, the Si loss tends to drop off, finally attaining an upper limit of approximately 0.30 wt% Si corresponding to 0.34 wt% 0 removed from the weld pool. In comparison, the amount of Si lost in the two preceding stages (i.e. electrode tip and arc column) is much smaller, as shown by the data for the chilled metal Si content. Since no slag is formed under the conditions of rapid cooling, silicon must escape in the form of a gaseous
Fig. 2.26. Correlation between filler wire and weld metal carbon contents in C02-shielded Data from Ref.24.
welding.
153
t
CiS
o Multi-layer
1.20
- -
weld metal
Electrode wire
~ 0.801~~~e.::=--~~
-=-~~~~~-=-~
20 25
0.40
10
15
30
Vol% 02 in Ar ------
o Multi-layer weld
1.20 - .Electrode wire
metal
~ 0.80 ~
20
40
60
80
100
Fig. 2.27. Measured silicon contents in chilled and multi-layer weld metals vs the oxygen potential of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-C02 gas mixtures. Data from Grong and Christensen.'
154
product. Evaporation losses can in this case be excluded due to a very low vapour pressure of silicon at the prevailing temperatures. It is therefore reasonable to assume that SiO(g) forms as a result of chemical reactions occurring at the electrode tip. The observed decrease in chilled metal Si content with increasing O2 and CO2 contents in the shielding gas is probably caused by the presence of CO at the gas/metal interface, which facilitates SiO formation according to the reaction: (2-36) Si02 (slag) + CO (gas) = SiO (gas) + CO2 (gas) Figure 2.28 shows a plot of the equilibrium partial pressure of SiO vs temperature at silica saturation for three different CO levels. Note that stoichiometric amounts of CO2 have been assumed to form in order to calculate PSiO. It is evident from the thermodynamical data presented in Fig. 2.28 that the formation of SiO is strongly dependent on the CO partial pressure at the slag/metal interface. Thus, from a thermodynamic standpoint the silicon loss at the electrode tip should be most pronounced during CO2-shielded welding due to the resulting higher CO pressure. The data shown in Fig. 2.27(a) and (b) support this assumption. It is seen that the silicon loss is increased by a factor of 3 to 5 in presence of CO2 when comparison is made on the basis of equal oxygen potential of the shielding gas (i.e. equal loss of de oxidants in the weld pool). Similar observations have also been made by Heile and HiIl27 from determination of silicon in collected GMA welding fumes. The recorded chilled metal Si loss in Fig. 2.27(a) and (b) is in good agreement with the reported fume formation rates of silicon. Since all CO consumed in reaction (2.36) is expected to be regenerated immediately by decomposition of CO2 at the metal surface, the high CO partial pressure required for SiO formation at the electrode tip is maintained even in the case of extensive silicon losses. Consequently, the net reaction can be written as:
t
(f')
E ....
ctS
'0
0 i:i3 c..
1500
Fig. 2.28. Equilibrium partial pressure of SiO vs temperature at different CO levels. Data from Refs. 25 and 26.
155 (2-37)
Si02 (slag)
Q (dissolved)
+ SiO (gas)
When welding is performed in Ar-02 gas mixtures, PSiO and Peo may be taken proportional to the recorded loss of silicon and carbon in chilled metal (see data in Fig. 2.29). It is evident from this plot that the silicon loss is directly proportional to the corresponding loss of carbon up to a certain critical level. Thus, during the initial period of carbon oxidation at the electrode tip the SiO formation is probably controlled by the resulting partial pressure of CO at the slag/ metal interface, according to reaction (2-36). When the carbon reaction is blocked, the chilled metal silicon loss becomes independent of the CO partial pressure (see reaction (2-37)), since all CO consumed in the SiO formation will immediately be recirculated within the system. However, at the break even point for silicon control of the oxygen level at the electrode tip, the CO pressure in the surrounding gas phase will be the same for both Ar-02 and Ar-C02 gas mixtures. Hence, the recorded loss of silicon in chilled metal at 20 vol % CO2 in Ar is seen to be similar to that in Ar + 10 vol% O2, as indicated by the heavy broken line in Fig. 2.29. At the temperatures where liquid steel is normally deoxidised, silicon and manganese have a strong affinity to oxygen. Their ability to form stable oxides decreases rapidly with increasing temperature, and above approximately 1800C silicon and manganese do no longer act as efficient deoxidation agents. Precipitation of manganese silicate slags is therefore favoured by the lower metal temperatures prevailing at the electrode tip and in the cold part of the weld pool. At higher temperatures, these oxides become unstable. Consequently, as a result of the metal superheating occurring during droplet transfer through the arc column, the macroscopic slag phase formed earlier at the electrode tip surface (as reported by Corderoy et al.22) will redissolve in the metal. This gives rise to a relatively high chilled metal oxygen content (to be discussed below).
20
t
~
'0
g w
0 -l
C/) CIJ
c:
50
Fig. 2.29. Correlation between loss of carbon and silicon in chilled metal at different O2 levels in the shielding gas. The corresponding loss of silicon at 20 vol % CO2 in Ar is indicated by the heavy broken line in the graph. Data from Grong and Christensen.'
156
Example (2.5)
Consider CO2-shielded welding on a thin sheet of low-alloy steel with a 0.8mm dia. electrode wire under the following conditions: 1= 100A, U = 26V,
V
= 3mm
S-1
Based on the data presented in Fig. 2.27 (b), calculate the fume formation rate (FFR) of silicon (in mg per min) due to SiO formation at the electrode tip. The wire feed rate is 125mm s-l.
Solution
The total loss of silicon due to SiO formation may be taken equal to the observed difference between the filler wire and the chilled metal silicon contents. For welding in pure CO2, we get: il%Si
= 0.83 -
0.66
= 0.17
The corresponding fume formation rate of silicon (in mg min-I) can readily be calculated when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg mm ", we obtain: FFR
= (WFR)~(d)2(p)
4
~%Si 100
= 125 X 60~(0.8)2
4
7.85(0.17)mgmin-1 100
=50 mg mirr ' A comparison with the measured FFR of silicon in Table 2.6 (at 1= 130A) shows that the calculated value is reasonable correct. Moreover, these data support our previous conclusion that the SiO formation is favoured by a high CO2 content in the shielding gas due to the dissociation reaction. In fact, more detailed studies of the reaction kinetics have confirmed that the rate of SiO formation is proportional to the resulting partial pressure of CO at the gas/metal interface.i" in agreement with equation (2-36). 2.10.1.4 Evaporation of manganese It is seen from the data in Fig. 2.30(a) and (b) that the amount of manganese lost in chilled metal is virtually independent of the oxygen potential of the shielding gas, as indicated by the constant difference of about 0.35 wt% between the filler wire and the chilled metal Mn contents in both graphs. This implies that significant amounts of manganese are lost during droplet transfer through the arc column as a result of evaporation. At the prevailing temperatures, the vapour pressure of iron is also high due to the almost unity activity of Fe. If the average arc metal temperature is taken equal to about 2400C,29 the data in Fig. 2.31 indicate that the vapour pressures of iron and manganese (at 1.27 wt% Mn) are nearly identical and close to 0.05 atm. In the hot surface layers of liquid metal facing the arc the temperature will be even higher, which means that iron vapour will dominate. Measurements of collected GMA welding fume reported by Heile 'and Hi1l27(see data in Table 2.6) show a substantial higher loss of iron than that derived from simple thermodynamical calculations taking the rate of element loss proportional to the vapour pressure. From their results a reasonable value of the average mass ratio Fe to Mn in dust is about 5. Consequently, as a preliminary estimate the loss of iron may be taken 5 times the amount of manganese lost
157
mu=
Table 2.6 Measured fume formation rates in GMA welding of ferrous materials. Data from Heile and
ShieldingGas
Voltage (V) 29 31 35 28 28 29 34 28 28 28 23 27 35 27 30 30 30 30 Mn 1 1 5 15 12 5 18 29 16 10 11 13 29 13 14 19 31 36
Fume Formation Rate (mg min-I) Si 1 0 4 8 10 4 16 33 22 23 20 35 62 59 63 73 112 125 Fe 22 12 51 134 75 35 86 273 129 76 75 105 191 86 120 126 216 214
Argon
Ar + 2%
02
150 200 300 400 100 200 300 100 150 300 130 150 200 250 300
Ar + 5%
02
Ar + 250/0 CO2
Pure CO2
in chilled metal, corresponding to a total loss of 1.7 wt % (Fe + Mn) or 500mg (Fe + Mn) per min. The reported fume formation rate of Fe and Mn under nearly similar conditions is only one half of that stated above, which indicates that the calculated flux of metal' vapour may be somewhat overestimated. However, if these two values are considered to represent borderline cases, the volume of metal vapour is in the range from 300 to 500 times that of the droplets and will therefore be more than sufficient to protect the metal from arc atmosphere oxidation at this stage of the process. This conclusion is consistent with statements made by Distin et al.31 who claim that iron vapour acts as an effective oxygen getter already at about 1900oe. Very large amounts of manganese are also lost in the weld pool stage, as shown by the difference between the measured Mn contents in chilled and multi-layer weld metals. This situation appears to be quite similar to that of silicon. For Ar-02 mixtures the manganese loss increases steadily with increasing oxygen contents in the shielding gas, and reaches a value of about 1.08 wt% Mn (or 0.31 wt% 0) removed from the weld pool at 30 vol% O2 in Ar. In the case of Ar -C02 shielded welding the Mn oxidation loss starts to drop off at a CO2 content of about 20 vol%, finally attaining an upper limit of about 0.50 wt% Mn corresponding to 0.15 wt% 0 removed from the weld pool. A more detailed discussion of oxidation reactions in GMA welding is given in Section 2.10.1.5.
Example (2.6)
Consider GMA welding low-alloy of steel under conditions similar to those in Example 2.5.
158 (a)
1.20
Chilled metal (falling droplet)
o Multj ..layer
10
20 15 Volo/o 02 in Ar
25
30
(b)
~ eft ~
1.2
c:
-.
0
Chilled metal (falling droplet) Multi ..layer weld metal - Electrode wire
20
40
60
Vol% .C02 in Ar
Fig. 2.30. Measured manganese contents in chilled and multi-layer weld metals vs the oxygen potential of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-C02 gas mixtures. Data from Grong and Christensen. 1
80
100
Based on the data presented in Fig. 2.30(b), calculate the fume formation rate (FFR) of manganese due to evaporation losses occurring during droplet transfer through the arc column. Estimate also the effective mass transfer coefficient for manganese evaporation under the prevailing circumstances by utilising the vapour pressure data in Fig. 2.31. The surface temperature of the falling droplets is assumed constant and equal to 2600C.
159
t
E ~
-1
CD
U-
.3
CO)
~ -2
8 ~
~ c.
-3
Fig. 2.31. Equilibrium manganese vapour pressure and corresponding vapour pressure ratio temperature at 1.27 wt% Mn in iron. Data from Kubaschewski and Alcock. 30 Solution
to PFe
VS
The total loss of manganese due to evaporation may be taken equal to the observed difference between the filler wire and the chilled metal manganese contents. For welding in pure CO2, we have: ~%Mn = 1.52 - 1.26 = 0.26
The corresponding fume formation rate of manganese (in mg min-I) can readily be calculated when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg mnr-', we obtain: FFR = (WFR)~(d)2(p) 4
z
~%Mn 100
= (125 x 60)~(0.8)2
77 mg min "
A comparison with the data in Table 2.6 reveals that the reported fume formation rate of manganese (at 1= 130A) is much lower than computed in the present example. This discrepancy can probably be attributed to differences in the filler wire manganese content. Assuming that the evaporation loss of manganese is controlled by a transport mechanism in the gas phase, we can estimate the effective mass transfer coefficient from equation (2-15) by inserting a reasonable average value for the manganese vapour pressure at the gas/metal interface (PMn in the bulk phase is taken equal to zero). Reading from Fig. 2.31 (at T = 2600C) gives:
P~n
= 0.1
atm
160
METALLURGICAL
MODELLING
OF WELDING
If the diameter of the falling droplets is taken equal to the diameter of the filler wire, it is possible to calculate the total loss of manganese associated with one droplet (in mol): n
Mn
=~(d)3() 6
100MMn
I1%Mn
=~(0.8)3(7.85X10-3) 6
0.26 100x54.9
mol
= 9.97 X 10-8 mol Since the average flight time of large, globular droplets through the arc column is of the order of one second, the corresponding flux of manganese vapour per unit time is close to 9.97 X 10-8 mol s-l. Thus, by rearranging equation (2-15), we obtain the following value for the effective mass transfer coefficient: k (Mn)=n
d Mn A
RT
P~n S-1
=9.97X10-8(82056.2X2873)mms-1
1t(0.8)2(0.1) .
=117 mm
Although the above value is rather uncertain, the calculated mass transfer coefficient is of the expected order of magnitude. This supports our previous conclusion that the evaporation kinetics are controlled by a transport mechanism in the gas phase. 2.10.1.5 Transient concentrations of oxygen It is evident from the data summarised in Fig. 2.32 that the oxygen content in both chilled and multi -layer weld metal increases with increasing oxygen potential of the shielding gas. The chilled metal analysis is representative of the oxygen absorption occurring at the electrode tip due to the lack of gas/metal interaction in the arc column. Measurements of manganese oxidation losses during droplet formation have been performed by Corderoy et al. 22 From the curve presented for 20ms tip melting cycles, a reasonable estimate for the manganese oxidation loss in Ar-02 shielding gas mixtures may be: [%Mn]ox. ~ (8.4 x 10-3)(vol %02) (2-38)
Taking the Si to Mn mass ratio in precipitated slag equal to 0.66,22the corresponding loss of silicon is equal to: [%Si]ox.
=::
(0.66)(8.4
(2-39)
Based on equations (2-38) and (2-39) it is possible to calculate the oxygen absorption at the electrode tip which is associated with the MnO and the Si02 slag formation. If corrections also are made for the amount of oxygen simultaneously removed as iron oxide*, we obtain: [%O]abs .
=::
(~X 28.1
5.5
(2-40)
2)
*An approximate correction can be made from an analysis of the Fe to Mn mass ratio in precipitated slag, which under the prevailing circumstances is close to 0.16.22
161
(a)
Chilled metal (falling droplet) Multi-layer weld metal - . - Electrode wire
'#. 0.16 ~
10
15
20
25 -----
30
Vol% 02 in Ar (b)
. -
20
40
60
80
100
Fig. 2.32. Measured oxygen contents in chilled and multi-layer weld metals vs the oxygen potential of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-C02 gas mixtures. Data from Grong and Christensen.'
At 30 vol% O2 in Ar the absorption of oxygen is roughly 0.27 wt% 0, which is reasonably close to that recorded in the chilled metal analysis (about 0.23 wt% 0). The rate controlling step for metal oxidation at this stage of the process is believed to be transport of atomic oxygen through a stagnant gaseous boundary layer of Ar and/or CO adjacent to the metal surfacc.F On the other hand, the measured chilled metal oxygen contents are much too low to account for the heavy oxidation losses of deoxidants observed in multi-layer welds. This indicates that
162
considerable amounts of oxygen are introduced in the hot part of the weld pool immediately beneath the root of the arc. Although the gas/metal interfacial contact area available for reaction is much smaller than that of the electrode tip (or falling) droplets, the strong turbulence existing in the hot part of the pool will provide an effective circulation of the liquid metal through the reaction zone.32 Moreover, absorption of oxygen at this stage of the process will be favoured by the increased time available for gas/metal interaction. If the diameter of the reaction zone is taken equal to about 3mm, the time available for oxygen absorption in the weld pool is of the order of one second for a typical welding speed of 3mm S-I. In comparison, the corresponding reaction time during droplet formation at the electrode tip is only 20 to 50 ms. Calculations of the total oxygen absorption can be done on the basis of the measured difference between silicon and manganese contents in chilled and multi-layer weld metals, designated d[%Si] and d[%Mn], respectively *. Moreover, corrections should be made for the amount of oxygen simultaneously removed as iron oxide from the weld pool. For deoxidation with silicon and manganese the mass ratio wt% Fe to wt% Mn in precipitated slag is equal to about 0.1.1 This leads to the following balance:
[%O]tot. ~ [%O]anal.
+ 28.1
32
L\[%SI]
+ (1.1) 54.9
16
L\[%Mn]
(2-41)
+ 1.1
It is seen from the graphical representation of equation (2-41) in Fig. 2.33(a) and (b) that most of the oxygen pick-up takes place in the weld pool. However, at present it is not clear whether the calculated values represent a real transient concentration in the hot part of the weld pool or a number in concentration units representing precipitation of manganese silicate slags. According to Fischer and Schumacher+' the solubility of oxygen in liquid iron is 0.94 wt% at 1900C and 1.48 wt% at 2000C. This temperature range is probably relevant with respect to the hot part of the weld pool, although the surface metal temperature will be even higher.v' However, in steel weld metals the oxygen concentrations should be well below this solubility limit in the presence of silicon and manganese. Taking the average weld pool silicon content equal to 0.50 wt%, the corresponding oxygen content in equilibrium with a silicasaturated slag is roughly 0.10 and 0.20 wt% at 1900 and 2000C, re~pectively.230n the other hand, under the prevailing circumstances there will be no solid/liquid interface available for nucleation of oxide particles, and hence homogeneous nucleation is the only possibility. According to Sigworth and Elliott35 this requires a relatively high degree of supersaturation, which means that the oxygen concentration most likely will exceed the Si-Mn-O equilibrium value before slag precipitation occurs. The data in Fig. 2.33(a) and (b) indicate that the weld pool oxygen absorption is controlled by a complex transport mechanism in the gas phase. However, since mass transfer in gas-jet! liquid systems is not fully understood.' we shall only consider the limiting case where the resistance to mass transfer is confined to a stagnant gaseous boundary layer adjacent to the metal surface. Under such conditions equation (2-10) predicts that the transient oxygen concentration is determined by the partial pressure of atomic oxygen in the plasma gas, as shown
*Oxygen consumed in CO and SiO formation is not included.
163
~ Calculated total oxygen absorption _.. - Analytical weld metal oxygen content
0.5
10
15
20
25
30
(b)
1..5
0
(fl ....,
~ 0.5
Vol% CO2 in Ar
Fig. 2.33. Calculated total oxygen absorption in GMA welding at different oxygen potentials of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-C02 gas mixtures. The analytical weld metal oxygen content is indicated by the broken lines in the graphs. Data from Grong and Christensen. 1
164
by the plots in Fig. 2.34. If we as a borderline case assume that all oxygen present in the plasma gas is immediately absorbed in the liquid metal, the slope of the curve in Fig. 2.34 for Ar-02 gas mixtures (equal to about 2.67 wt% dissolved oxygen at one atmosphere total pressure of atomic oxygen) is representative of the total amount of gas passing through the arc column. Taking the mass of liquid metalleaving/entering the reaction zone equal to 30g mirr ' under the prevailing circumstances,' the following gas flow rate is obtained (in NI min-I). (2.67)(30)(22.4)Nlmin100 16
1 ~
1.1Nlmin-1
This value corresponds to about 6 per cent of the total shielding gas flow rate. In contrast to the situation described in Fig. 2.34 for Ar-02 gas mixtures, a large deviation from the expected relationship is observed for welding in CO2-rich atmospheres. At present, the reason for this shift in the reaction kinetics is not known. However, it is evident from the data presented in Fig. 2.34 that the effective mass transfer coefficient decreases by a factor of about 2.5 when the CO2 content in the shielding gas increases from 10 to 100 vol %. This implies that the oxidation capacity of pure CO2 is comparable with that of Ar + 13 vol % O2, although the partial pressure of atomic oxygen in the plasma gas is equivalent with an oxygen content of about 33 vol%.
Example (2.7)
Consider GMA welding of low-alloy steel in Ar + 20 vol% CO2 and pure CO2, respectively. Based on the results in Figs. 2.27 (b) and 2.30(b), calculate the total weight of top bead slag (in gram per 100 gram weld deposit) which forms as a result of deoxidation reactions. Assume in these calculations that the iron to manganese mass ratio in precipitated slag is equal to 0.1.
~ ~ ~
1.0
0.8
~ 0.6 0.4
0.2
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Po ' atm -----Fig. 2.34. Calculated total oxygen absorption in GMA welding at different partial pressures of atomic oxygen in the plasma gas. Data from Grong and Christensen.'
165
Solution
The amount of silicon and manganese lost as a result of deoxidation reactions is equal to the observed difference in the chilled and multi-layer weld metal Si and Mn contents. Assuming that these elements are removed from the weld pool as Si02 and MnO, respectively, the following balance is obtained: 60.1 A[%Si]+(l.1) 70.9 A[%Mn] 28.1 54.9
mslag
""
Thus, for welding in Ar + 20 vol% CO2, the total weight of slag amounts to:
msZag
= [60.1
28.1
(0.20)
msZag
= [60.1
28.1
(0.26)
A comparison with the experimental data in Fig. 2.35 shows that there is a fair agreement between the amount of slag recorded by weighing and that calculated from a simple mass balance of silicon, manganese and iron.
'(i)
6.0
Calculated weight of slag <> Measured weight of slag 4.0
..0
Q)
0-
0 0
~ as
2.0
C>
Vol% CO2 in Ar
weld-
Fig. 2.35. Comparison between measured and calculated weight of top bead slag in C02-shielded ing. Data from Grong and Christensen.'
166
METALLURGICAL
MODELLING
OF WELDING
2.10.1.6 Classification of shielding gases The data in Fig. 2.34 provide a basis for evaluating the oxidation capacity of various shielding gases. For welding in Ar + O2 and Ar+C02 gas mixtures up to 10 vol % CO2 in argon, the total oxygen absorption is approximately given by the following equation:
[%O]tot
= 2.67 Po = 5.34
-----=--
= 2.67
(2-42)
Similarly, for welding with CO2-rich shielding gases (i.e, between 10 and 100 vol % CO2), we obtain: [%O]tot.
z
~ 0.88" Po = 0.88
(2-43)
Equal oxidation capacity means that the total weld metal oxygen absorption is the same for both shielding gas mixtures. Hence, we may write: 2 ----~-.= 100+vol%C02 vol %C0 vol %02
2-----=-100+vol%02
(2-44)
when the CO2 content in the shielding gas is less than 10 vol%, and __ v_ol_%_o _C_0-=.2_ = 36. 100+vol%C02
s(
vol % 2 100+vol%02
J2
(2-45)
when welding is performed with CO2-rich shielding gases (more than 10 vol% CO2 in Ar). Based on equations (2-44) and (2-45) it is possible to compare the oxidation capacity of various shielding gases (see Table 2.7). Included in Table 2.7 is also a slightly modified version of the International Institute of Welding (IIW) classification system.l" which is based on an evaluation of retained (analytical) oxygen in the weld deposit. It is evident from these data that both systems are applicable and mutual consistent, although the former one utilises a more rational criterion for the shielding gas oxidation capacity. 2.10.1.7 Overall oxygen balance In GMA welding with solid wires, the eo content in the exhaust gas provides a direct measure of the extent of gas/metal interaction. This CO content should be compatible with that calculated from an overall oxygen balance for the reaction system.'?
Example (2.8)
Consider GMA welding of low-alloy steel under the following conditions: 1= 350A, U
= 30V, v = 4mm
S-I, 'll
= 0.8,
To =
zo-e
167
Table 2.7 Proposed shielding gas classification scheme for GMA welding of low-alloy steel according to equations (2-44) and (2-45). Included is also a modified version of the corresponding IIW's classification system.l"
Shielding Gas Composition Vol% CO2
-
[%0] Vol%Ar balance balance balance balance balance balance balance balance balance balance balance
tot.
Vol%02
(wt%)
0-2
-
0-4 0-2
-
<0.10
0.02)
Oxidising
0-1 2-5
-
4-10 2-5
-
0.1-0.25
(0.025-0.035)
Oxidising
1-21'2 5-8
-
10-25
-
(0.035-0.045)
J.
8-13
-
Strongly oxidising
25-100
-
(0.045-0.07)
Extremely oxidising
> 13
(> 0.07)
*The analytical weld metal oxygen content (in wt%) is given by the values in brackets.
The shielding gas is pure CO2 and is supplied at a constant rate of 15NI min-I. Based on the composition data in Table 2.8 calculate the resulting CO content in the welding exhaust gas.
Solution
First we calculate the nominal weld metal chemical composition by neglecting oxidation loss of alloying elements due to chemical reactions: B B+D
[%X]
nom. -
---[%X]
plate
+--[%X]
B+D
wire
The dilution ratios B/(B + D) and D/(B + D) can be estimated from the classic heat flow theory presented in Chapter 1. From equations (1-75) and (1-120), we have:
Table 2.8 Chemical composition of filler wire, base plate and weld metal used in Example 2.8.
C (wt%) 0 (wt%) Si (wt%) Mn (wt%)
168 and
D= This gives:
= 61.3mm2
[%X]nom.
= O.57[%X]plate + 0.43[%X]wire
The extent of gas/metal interaction can then be evaluated from the observed concentration displacements: [Ll%X]
= [%X] - [%X]nom.
Calculated values for the concentration displacements of carbon, oxygen, silicon and manganese utilising the composition data in Table 2.8 are given below.
Element [%X]nom. [~%X] C
0.12 -0.03
0.008 0.057
Si
Mn
0.63 -0.28
1.39 -0.58
The total CO evolution (in mol min-I) can now be computed from an overall oxygen balance for the reaction system. In these calculations we shall assume that Si and Mn lost as SiO(g) and Mn(vap.) immediately react with CO2 to form Si02 and MnO, respectively*. Taking the density of steel equal to 7.85 X 10-3 g mm ', the total mass of weld metal produced per unit time amounts to: 60up(B + D)
= 60
+ 61.3) g min-I
= 268 g min-I
Overall oxygen balance (i.e. consumption o!C02) Oxidation of carbon: 2x 0.03 (268) mol CO min-I = 13.4 x 10-3 mol CO min-I 100 x 12
= 2CO)
mol CO min-I
169
mm'
=28.3x10-3
mol CO min:"
= MnO
+ CO)
= 9.6x10-3
mol CO min-I
Q + CO)
Total CO evolution (sum):(13.4 + 53.5 + 28.3 + 9.6) X 10-3 mol CO min -I CO min-I
= 104.8 X
10-3 mol
Based on this information it is possible to calculate the resulting CO content in the welding exhaust gas: vol % CO = (104.8
X
x 100% ~ 15%
A comparison with the data in Table 2.2 shows that a CO content of about 15 vol % is reasonably close to that determined by analysis. 2.10.1.8 Effects a/welding parameters So far, gas/metal interactions in GMA welding has mainly been discussed in terms of the oxygen potential of the shielding gas. In the following, some consideration will be given to the effects of welding parameters on the weld metal chemistry.
Amperage
When the welding current is raised, the time available for interaction decreases due to the more rapid detachment of the electrode tip droplets. At the same time the interfacial contact area increases as the average droplet size becomes smaller. From measurements of fume formation rates in GMA welding.i? it has been shown that these two counteracting effects will almost cancel, i.e. the total amount of emitted dust (in mg per g deposit) is found to be constant and nearly independent of the applied amperage. On the other hand, the total fume formation rate is probably not a reliable index for the burn-off of Si and Mn, since the evolution of iron vapour during droplet transfer will tend to conceal the corresponding loss of alloying elements. The effect of amperage (or more correct the droplet detachment frequency) on the burn-off of carbon, silicon, and manganese in CO2-shielded welding has been investigated by Smith et al. 38 They found that the recovery of alloying elements in the weld deposit increased with increasing welding current (i.e. droplet detachment frequency). In view of the previous discussion, it is reasonable to assume that the higher weld metal carbon and silicon contents reported by Smith et ai.38 are a result of a reduced CO and SiD gas evolution at the electrode tip due to the shorter time available for chemical interaction. In the case of manganese reduced evaporation losses because of a more rapid transfer of the droplets through the arc column offers a
170
reasonable explanation to the increased element recovery. This shows that the weld metal chemistry is sensitive to variations in the welding current.
Arc voltage
Since the arc voltage neither affects the melting rate nor the droplet size to any great extent, 39 variations in the arc voltage should only have a minor effect on the weld metal chemistry. This conclusion is apparently in conflict with observations made by Lindborg.t" who found that the oxidation reactions in GMA welding were strongly voltage dependent and at the same time independent of the welding current, the droplet detachment frequency, and the mode of metal transfer (spray or short-circuiting). Consequently, further investigations are required to explain these discrepancies.
Welding speed
It can be inferred from the data of Grong and Christensen 1 that the analytical weld metal carbon and oxygen contents are virtually independent of the welding speed v within the normal range ofGMAW (i.e. from 0.4 to 6 mm S-I). However, the intensified losses of silicon and manganese observed at low welding speeds indicate that more oxygen is absorbed in the weld pool under such conditions. This point is more clearly illustrated in Fig. 2.36 which shows a plot of [0/0 0] tot. vs V for a series of multi-pass GMA welds deposited under the shield of Ar + 10 vol % 02. It is evident that the total oxygen absorption increases nearly by a factor of two when the welding speed decreases from 6 to 0.4mm S-I. This shows that the welding speed has a marked effect on the transient oxygen pick-up in the hot part of the weld pool during GMA welding, since it controls the time available for element absorption. 2.10.2 Submerged arc welding In flux -shielded processes the reaction pattern is much more difficult to assess because of the
Fig. 2.36. Calculated total oxygen absorption in GMA welding at different travel speeds. Data from
171
complicating presence of the slag. For this reason most investigators have chosen to analyse empirically slag/metal reactions in SA welding. Nevertheless, some authors have been able to interpret their results on more theoretical grounds in spite of the complex reaction systems involved.41-45 Unfortunately, these thermodynamic approaches give, at best, only a qualitative description of the compositional changes occurring during the welding operation. Recently, a kinetic model has been developed by Mitra and Eagar'? to account for variations in the element recovery in both single-pass and multi-pass SA steel weldments. From their work it is evident that the transfer of alloying elements between the slag phase and the weld metal cannot be adequately described by means of a primitive model of pseudo-equilibrium without including a more detailed analysis of the reaction kinetics. This shows that the conditions existing in SA welding are quite similar to those prevailing during GMA welding, although the experimental and theoretical challenges are much greater in the former case due to the complicating presence of a macroscopic slag phase. 2.10.2.1 Flux basicity index During SA welding of steel, oxygen may be transferred from the slag to the weld metal due to decomposition of easily reduced oxides at elevated temperatures according to the overall reaction: MxOy
=xM
(dissolved) + y
Q (dissolved)
(2-46)
where MxOy denotes any oxide component in the slag phase (e.g. Si02, MnO or FeO). The basicity index (B.I.), originally adopted from steel ladle refining practice, is most frequently employed for assessment of oxygen pick-up in SA welding, since it gives an approximate measure of the flux oxidation capacity. A number of different expressions exists in the literature, but for the purpose of convenience the basicity index defined by Eagar"? has been adopted here:
B.I.=~
L (%
.J (% non-basic
= CaO+
MgO+ BaO+ SrO+ Na20+ K20+ Li20+ 1/ 2(MnO+ FeO) Si02 + 1 / 2(A1203 + Ti02 + Zr02)
(2-47)
where the concentration of each flux component is given in weight percent. It is evident from Fig. 2.37, which shows a typical correlation between the weld metal oxygen content and B .1., that the oxygen level of welds produced under acid fluxes (i.e. low B.I.) is strongly dependent on the basicity index. In contrast, the oxygen concentrations of welds deposited under basic fluxes are seen to be essentially independent of B.I., as indicated by the horizontal part of the curve in Fig. 2.37. It should be noted that this analysis gives no information about the extent of slag/metal interaction, since it is based on data for retained oxygen in the weld deposit. Consequently, because of the empirical nature and limited applicability of the basicity index, its role in the choice of welding fluxes for SA welding is a keenly debated question.
172
t
"i
g
c:
8
c Q)
0.02
Fig. 2.37. Correlation between retained oxygen and flux basicity in SA welding. Data from Eagar.47 2.10.2.2 Transient oxygen concentrations
In SA welding of C-Mn steels, the transient flux of oxygen passing through the weld pool can be estimated from the observed concentration displacements of silicon and manganese, which may be taken equal to the difference between absorbed and rejected Si and Mn, respectively:
[~%Si]
= [%Si]abs.
= [%Mn]abs.
- [%Si]rej.
(2-48)
and
[~%Mn] - [%Mn]rej.
(2-49)
If we assume that rejection of Si and Mn in the weld pool occurs as a result of MnSi03 microslag precipitation and subsequent phase separation, [%Mn]rej. is bound to [%Si]rej. through the following stoichiometric relationship:
[%Mn]rej.
(2-50)
Taking the ratio between absorbed Mn and Si in the weld metal equal to k, a combination of equations (2-48), (2-49) and (2-50) gives:
[%Si] .
rej.
= [~%Si]k
1.95 - k
- [~%Mn]
(2-51)
The value of k is difficult to evaluate in practice, but in view of the reported mass transfer coefficients for Mn and Si a reasonable estimate would be about 0.5 in the case of manganese silicate fluxes." Under such conditions the total oxygen absorption, [%O]abs. is given by:
173
[%O]abs.
= [L1%O] + [%O]rej.
= [L1%O]
+ ~([L1%Si]O.5
28.1
- [L1%Mn]) 1.45
(2-52)
where [L1%0] is the observed concentration displacement of oxygen in the weld metal, and [%O]rej- is the amount of oxygen rejected from the weld pool as a result of deoxidation reactions. Based on equation (2-52) it is possible to estimate the total oxygen absorption during SA welding of C-Mn steels from an analysis of measured concentration displacements of oxygen, silicon, and manganese in the weld metal. The results of such calculations are shown graphically in Fig. 2.38, using data from Indacochea et al. 44 It is evident from this plot that the total oxygen absorption during SA welding is much larger than that inferred from an analysis of retained oxygen in the weld deposit. The situation is thus quite similar to that observed experimentally in GMA welding (see Fig. 2.33). It should be noted that the calculated values for [%O]abs. in Fig. 2.38 may be encumbered by systematic errors due to the number of simplifying assumptions inherent in equation (2-52). However, this does not affect our main conclusion regarding the significance of the oxygen absorption, since more refined calculations give a pattern similar to that observed above (see Fig. 2.39). 2.10.3 Covered electrodes Chemical reactions during SMA welding have been studied by several investigators in the
t
E S
r::
Experimental MnO-FeO-Si02
fluxes
8
c: Q)
C')
0.30
g 0.20
-0Retained oxygen _ - -0-0- - .0 - _0-
0.10
[A
%
0.30
Fig. 2.38. Calculated total oxygen absorption in SA welding with experimental MnO-FeO-Si02 Data from Indacochea et al.44
fluxes.
174
Flux Type CS: Bead on Plate " -: Two Wires Filled Symbols: Flux FB
~ 0.6
E
$ c
<I>
8 c 0.4
C)
o
0.2
0.2
0.4
0.6
0.8
1.0
Silicon content. wt% ~ Fig. 2.39. Calculated total oxygen absorption in SA welding with commercial calcium silicate (CS) and fluoride-basic (FB) fluxes. Data from Christensen and Grong.f
past.47-51 Most of these investigators have interpreted their results as a high-temperature equilibrium between the slag and the weld metal, but a verified quantitative understanding of the transfer of elements during welding is lacking. This situation arises mainly from the lack of adequate thermodynamic data for the complex slag/metal systems involved. 2.10.3.1 Reaction model The reaction model presented here is restricted to welding with basic covered electrodes. During SMA welding gases are generated by decomposition of compounds present in the electrode coating. In the case of basic covered electrodes, the decomposition of limestone results in an atmosphere consisting predominantly of carbon monoxide and carbon dioxide, containing only small amounts of H2 and H20 (see data in Table 2.2). The characteristic high concentrations of CO and CO2 in the arc atmosphere would be expected to lead to extensive absorption of carbon and oxygen in the weld metal. Under the prevailing circumstances, it is reasonable to assume that these reactions approach a state of local pseudo-equilibrium during droplet transfer through the arc column. During the subsequent stage of cooling in the weld pool, a supersaturation with respect to the various deoxidation reactions is initially increasing, which is released when the conditions for nucleation of the respective reaction products are reached. Since carbon is a much stronger deoxidant than silicon and manganese at temperatures above about 1700C,23 it is reasonable to assume that carbon will be in control of the oxygen level during the initial stage of cooling *, in accordance with the reaction: ~ (dissolved) + Q (dissolved) = CO(gas) (2-53)
*Although gases such as CO and CO2 are widely dissociated and ionised in the arc column, from a thermodynamic standpoint, there is no objection to the choice of molecular species as components for the system, provided that equilibrium is maintained down to temperatures where such species are stable.
CHEMICAL
REACTIONS
IN
ARc WELDING
175
Carbon boil in the weld pool has been detected experimentally during welding with covered electrodes.i? which implies that heterogeneous nucleation of CO is kinetically feasible under the prevailing circumstances. Possible nucleation sites for CO are gas bubbles present in the macroscopic slag layer covering the metal, created by the vigorous stirring action of the arc plasma jet. It should be noted that this behaviour is in sharp contrast to experience with GMA welding, where little or no oxidation of carbon takes place in the weld pool, as shown previously in Section 2.10.1.2. It is tentatively suggested that the apparent difference between SMA and GMA welding regarding the possibilities for CO nucleation in the weld pool arises from the lack of a macroscopic slag layer in the latter case. Unlike carbon, the deoxidation capacity of silicon (and manganese) increases rapidly with decreasing metal temperatures (se Fig. 2.40), which means that carbon oxidation becomes gradually suppressed during cooling in the weld pool. Upon reaching the critical temperature indicated in Fig. 2.40, the carbon reaction is blocked, silicon and manganese now control the oxygen level. An unknown but significant fraction of the manganese silicate inclusions precipitated in the hot part of the weld pool beneath the root of the arc are brought by convection currents to the interface between the macroslag and the metal, where they are readily absorbed. The remaining fraction formed in the cold and un stirred part of the weld pool is trapped in the metal solidification front in the form of finely dispersed oxide particles. This results in a high and rather unpredictable weld metal oxygen content. The above reaction model has been tested experimentally against data obtained from a series of hyperbaric welding experiments carried out in a remotely controlled pressure chamber with basic covered electrodes containing various levels of ferrosilicon in the electrode coating (see Table 2.9). Welding under hyperbaric conditions offer the special advantage of assessing the reactions through variations in the ambient pressure without changing the composition of the electrode coating or the core wire. Consequently, if the proposed reaction model
t
....: c:
1.0
.... c o
Q)
-------0.1
o c:
Cl) C)
O.01~--~--~--~----~--~--~----~--~
1600 1700 1800
1900
2000
Temperature.tc ----Fig. 2.40. The break even equilibrium temperature for silicon control of oxygen level at 0.1 wt% C and 0.3 wt% Si. Data from Elliott et al.23
176
Table 2.9 Contents of ferrosilicon and iron powder in the electrode coating of experimental consumables used in the hyperbaric welding experiments.
Electrode R* A B C FeSi (76 wt% Si) 4.5 wt% 5.5 wt% 6.5 wt% 7.5 wt% type electrode). Iron Powder 31 wt% 30wt% 29wt% 28wt%
0.20
t
~ ~ ~
o Electrode R A A 0.15
<>
B C
0.10
~S o 0.05
Fig. 2.41. Carbon absorption in hyperbaric SMA welding. Data from Grong et al.51
is at least qualitatively correct would expect a correlation between the weld metal carbon content and the concentrations of oxygen, silicon and manganese, both under atmospheric and hyperbaric welding conditions. The main effect of pressure on weld metal chemistry is thus to suppress the carbon-oxygen reaction in the weld pool at the expense of intensified oxidation losses of silicon and manganese, as indicated by the thermodynamic data in Fig. 2.40. 2.10.3.2 Absorption of carbon and oxygen It is evident from the data presented Fig. 2.41 that the weld metal carbon content increases monotonically with pressure from 1 to 31 bar for all four electrodes involved. This indicates that the carbon oxidation in the weld pool is systematically suppressed under hyperbaric welding conditions. Moreover, Fig. 2.41 reveals a small but important effect of electrode deoxidation capacity on the weld metal carbon content. Since ferrosilicon itself is an insignificant source of carbon, the observed increase in the carbon concentrations with increasing additions of ferro silicon to the electrode coating is an indication that carbon oxidation in the weld pool is blocked at an earlier stage of the process at high silicon levels, according to the reaction:
CHEMICAL REACTIONS IN
177
Si (dissolved) + 2CO(gas)
= 2~
This interpretation is further supported by the results tained in Fig. 2.42. Although there is considerable scatter that the recorded enhancement of the weld metal carbon the electrode coating is accompanied by a corresponding tions.
from the oxygen determination conin the data in this figure, it is evident content at high ferro silicon levels in reduction in the oxygen concentra-
2.10.3.3 Losses of silicon and manganese Suppression of carbon oxidation in the weld pool at elevated pressures gives rise to intensified oxidation losses of silicon and manganese, as shown in Figs. 2.43 and 2.44. Moreover, it is apparent that increased additions of ferrosilicon to the electrode coating result in a corresponding increase in both the silicon and the manganese concentrations. This finding suggests that the final weld metal content of the deoxidants is controlled by the reaction: Si (dissolved)
(2-55)
pendent of pressure, equation (2-55) may be rewritten as: [%Mn] = C3 ~[%Si] (2-56)
In Fig. 2.45 the weld metal manganese content has been plotted versus the square root of the silicon content by inserting data from Figs. 2.43 and 2.44. As it appears from Fig. 2.45, the
1000
...: c: Q) ....
0 0
750 E Cc,
c:
500
c
Cl)
250
o Electrode R
A
A B
II
10
30
Fig. 2.42. Oxygen absorption in hyperbaric SMA welding. Data from Grong et a1.51
178
o Electrode R A II A
0.75
II
o~----~----~----~----~----~------~----~ o 10 20 30
Total pressure, bar
Fig. 2.43. Silicon oxidation in hyperbaric SMA welding. Data from Grong et al. 51
2.0
?fl
o Electrode R 11 A 1.5
o
<>
II
...;c: .m
c:
0
0
Q)
1.0
e en
C)
<>
c ctS ctS
~
0
~
0
c:
0.5
10
20
Total pressure, bar
30
Fig. 2.44. Manganese oxidation in hyperbaric SMA welding. Data from Grong et al.51
experimental data cluster around a straight line passing through the origin, which confirms that the silicon and manganese concentrations are balanced by a reaction according to equation (2-55).
179
2.0
~ ~ ...:
0
o Electrode
A
[J
II
u
II
B C
E
Q) C/) Q)
Q)
0 0
1.0
as
ctS
0)
c:
::E
1/2
----
Fig. 2.45. Correlation between weld metal manganese and silicon contents, Data from Grong et al.51
2.10.3.4 The product [%CJ [%OJ From steelmaking practice, the product [%C] [%0] is generally accepted as an adequate index of the interaction between carbon and oxygen during the refining stage. This product is related to the equilibrium content of dissolved carbon and oxygen in contact with carbon monoxide of a partial pressure Pco(2-57) m = [%C][%O] = K5 Neoptot. Here K5 is the equilibrium constant for reaction (2-53) (equal to about 2.0 X 10-3 at 1600C and 2.6 X 10-3 at 2000C),23 Neo is the mole fraction of CO in the reaction product (equal to the partial pressure of CO at 1 bar), and Ptot.is the total ambient pressure. During the initial stage of cooling in the weld pool, the oxygen content in an assumed equilibrium with carbon would be expected to be higher than the analytical values. This situation applies in particular to welds made under hyperbaric conditions, where significant quantities of oxygen clearly are removed from the weld pool in the form of oxide inclusions after the completion of the carbon oxidation. The concentration of dissolved oxygen at the break even temperature for silicon control of the oxygen level can be estimated from the measured concentration displacements of oxygen, silicon and manganese in the weld deposit with increasing pressures, relative to 1 bar (designated a[%O], a[%Si] and a[%Mn], respectively). If the total amount of oxygen which reacts with silicon and manganese at 1 bar, as a first approximation, is taken equal to the analytical weld metal oxygen content, the following balance is obtained:
[%O]eq
= [%O]anal
+ --
32 . ~[%Sl] 28.1
+ --
16 ~[%Mn] 54.9
(2-58)
180
Here [%O]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given pressure, and [%O]anal. is the analytical weld metal oxygen content. For this correlation, minor vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron have been neglected. In Fig. 2.46 the product m = [%C] [%0] is plotted vs the total ambient pressure. Calculations of m have been done both on the basis of [%O]anal. and [%O]eq. It can be seen from Fig. 2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a straight line passing through the origin, which should apply to a true equilibrium reaction. However, when proper corrections are made for the amount of oxygen removed from the weld pool after the completion of the carbon oxidation, such a correlation may be obtained as shown by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon content) on the product m = [%C] [%0] can be observed within the precision of measurements. This result is to be expected if the weld metal carbon content is controlled by a local equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspection of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product K5N co is about 1.14 X 10-3 under the prevailing circumstances. If a reasonable average value for the equilibrium constant K5 of 2.3 X 10-3 is assumed within the specific temperature range of the reaction, we get: Nco:::= 0.5 (2-59) The above calculations suggest that the controlling partial pressure of CO in the reaction product is significantly lower than the ambient pressure under hyperbaric welding conditions. This probably arises from an extensive infiltration of helium in the nucleating bubbles at the slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of carbon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding thus appear to be similar to OBM/Q-BOP steelmaking, where simultaneous injection of oxygen and inert gas from the bottom of the convertor during the decarburisation stage results in a steel carbon content which is typically below the value calculated for equilibrium between oxygen and carbon at 1 atm partial pressure of CO. 52
181
<>+
o Electrode R A.. II A c B
C
10
20
30
Fig. 2.46. The product [%C][%O] in hyperbaric SMA welding. Solid symbols: calculations based on [%O]eq .. Open symbols: calculations based on [%O]anai .. Data from Grong et al. 51
t
t
N "Cii
Q)
Q)
13
CO
O)
C-
en
e
~
Reaction time
Fig. 2.47. The three major consecutive steps in steel deoxidation (schematic).
182
Although rate phenomena in ladle refining of liquid steel are extensively investigated and reported in the literature,53-55 only recently attempts have been made to include such effects in an analysis of deoxidation reactions in arc welding. 1, 56,57 2.11.1 Nucleation of oxide inclusions During ladle-refining of liquid steel, it is well established that homogeneous nucleation of oxide inclusions may occur in certain regions of the melt where the supersaturation is sufficiently high. 55,58Over the composition range normally applicable to deoxidation of steel, the number of nuclei formed at the time of addition of deoxidisers is approximately 105 mm-3.55 However, in steel weld metals, the number of oxide nuclei formed during the initial stage of deoxidation must be considerably higher to account for the observed inclusion number density of about 107 mnr ' to 108 mnr '. 56,57This implies that the supersaturation established in the weld pool on cooling as a result of rapid temperature fluctuations (,.., 103 DC s') exceeds by far that obtained by additions of deoxidisers to a liquid steel melt under approximately isothermal conditions. There are several theories available for treating nucleation phenomena, but for the purpose of convenience, a simplified version of the model of Turpin and Elliott58 has been adopted here. Consider a steel melt which is brought to a state of supersaturation by first equilibrating it with pure MxOy at a high temperature T, and then rapidly cooling it to a lower temperature T2, as shown schematically in Fig. 2.48. It follows from the classical theory of homogeneous nucleation that the required temperature difference T1- T2 necessary to achieve spontaneous precipitation of MxOy is approximately given by the following relationship:
(2-60) where /lHo is the standard enthalpy of reaction, o is the oxide-steel interfacial energy (assumed to be constant and independent of temperature), and Vm is the molar volume of the nucleus. The derivation of equation (2-60) is shown in Appendix 2.2.
Example (2.9)
Assume that precipitation of manganese silicates in the weld pool occurs according to the following reaction: Si (dissolved) + Mn (dissolved) + 3 Q (dissolved) where
/lGO(J)
= MnSi03
(slag)
(2-61)
= 858620
+ 345T.
Based on equation (2-60), calculate the critical temperature interval of subcooling for homogeneous nucleation of MnSi03. Typical physical data for liquid steel and manganese silicate slags are given in Table 2.10.
183
Time
Table 2.10 Physical data for liquid steel and manganese silicate slags at 1600C. Data from Refs. 3 and 53.
Density (kg m-3) 6900 2300 Viscosity (kg rrr' S-I) 4.3 X 10-3
-
0.8
Solution
= (60.1 + 70.9)
1000 x 2300
By inserting the appropriate values for Vm, MID, and o in equation (2-60), we obtain:
from which
If we assume that the supersaturation is released at T2 = 1600C (1873K), the initial temperature of the liquid T, becomes:
Tl=[
184
Similar calculations can also be carried out for other types of oxide inclusions, e.g. FeO(l), SiD2(s), and A12D3(s). The results of such computations are presented graphically in Fig. 2.49, using data from Refs 55 and 58. It is evident from these plots that the critical temperature interval of subcooling depends on the interfacial energy, cr. Although data for oxide-steel interfacial energies are scarce, the following average values are frequently used in the literature,55,58i.e. cr(FeO-Fe) =:: 0.3 J m-2; cr(Si02-Fe) =:: 0.9 J m-2; and cr(AI2D3-Fe);:;; 1.5 J m-2. If these values are accepted, the results in Fig. 2.49 indicate that the critical temperature interval of subcooling for homogeneous nucleation ofFeO (1), SiD2(s), and AI203(s) is of the order of 200 to 300C. Considering the fact that the liquid weld metal spans a temperature range of about 2200 to 1500C,34 it is not surprising to find that nucleation of oxide inclusions occurs readily in the weld pool during cooling. It should be noted that the quoted data for o are representative of ladle-refined steel deoxidised at 1600C. At higher metal temperatures, in the presence of large amounts of dissolved oxygen, the oxide-steel interfacial energies would be expected to be significantly lower.59 Hence, it is reasonable to assume that the actual temperature interval of subcooling required for spontaneous oxide precipitation in a weld pool is well below 200C. 500 '-""
<I
U
0
t-
400 300
/ / /. / /
~q,/
/
O~/
~/
~ Q) .... .~ ~
Q)
(ij
N
....
(ij
0 :.;::;
::l
200
./
/'
;1'
E Q)
0-
./
100
,,-
./
'E::
""
0.5
00
1.5
2.0
Fig. 2.49. Critical temperature interval of subcooling for homogeneous nucleation of oxide inclusions in steel weld metals at 1600C as a function of the interfacial energy c.
2.11.2 Growth and separation of oxide inclusions In practice, there are three major growth processes in steel deoxidation."
185
(i) Collision (ii) Diffusion (iii) Ostwald ripening. From deoxidation and ladle refining of liquid steel it is well established that the flotation rate of the oxides generally depends on their growth rates, since large inclusions separate much more rapidly than small ones, in agreement with Stokes law.55 Growth of the oxides can proceed either through diffusion of reactants in the melt to the oxide nuclei or by collision and coalescence of ascending inclusions, and is therefore influenced by factors such as the number density of the nuclei, interfacial tensions, and the extent of melt stirring. 53-55 The last factor is of particular importance in welding, because the stirring action will increase the possibilities for collision and coalescence of inclusions and, depending on the direction of flow, can give rise to circulation of inclusion-laden metal to the surface. As a result, the separation of small oxide particles, i.e. microslag, is strongly favoured by the turbulent conditions existing in the hot part of the weld pool immediately beneath the root of the arc. 2.11.2.1 Buoyancy (Stokesjlotation) Assessment of the role of buoyancy (Stokes flotation) in the separation of deoxidation products from the weld pool can be done on the basis of the experimental data of Grong et al. 56 reproduced in Fig. 2.50. It is evident that a change in the welding position (i.e. from flat to overhead position) has no significant effect on the weld metal oxygen content. This shows that the buoyancy effect does not play an important role in the separation process of oxide inclusions during welding. 0.05
cf!.
~ ~ c 0.03 Q)
0.04
....
c:
0
Q)
9 ~
c:
~6/o;.
1'1)0..<11)0/
0.02
~ 0
0.01
45
90
135
180
Fig. 2.50. Effect of welding position on retained oxygen in GMA steel weld deposits. Data from Grong et al. 56
186
The above conclusion is also in agreement with predictions based on the Stokes law, which gives the terminal velocity of the ascending particles (u) relative to the liquid:55 (2-62) where g is the gravity constant, d; is the particle diameter, ~p is the difference in densities between the liquid steel and the inclusions, and J.L is the steel viscosity. Taking the Stokes parameter, g~p/18J.L, equal to 0.6 J.Lm-1 s-1for manganese silicate slags in steel (Table 2.10), equation (2-62) becomes: (2-63) Generally, the majority of non-metallic inclusions in steel weld metals are of a diameter below 2 J.Lm.565 , 7 According to equation (2-63), such particles have a relative velocity less than 2.4 J.Lms'.This implies that the buoyancy effect alone is far too insignificant to promote flotation of the inclusions out of the weld pool before solidification, when it is recognised that the average fluid flow velocity in the weld pool is four to five orders of magnitude higher (to be discussed later). Stokes law is based on the assumption that the inclusions are completely wetted by the liquid steel, i.e. there is no slip at the oxide/metal interface.P Normally, interfacial tension effects promote slip at the particle/metal interface, which, in tum, enhances the flotation rate of the ascending oxides (often referred to as the Plockinger effect).6o The concept of a wetting angle has been used in this context. But, the slip phenomenon is probably perceived better in terms of a secondary flow in the interfacial region between the liquid metal and the solid precipitate, which is produced by gradients in interfacial tension. In the rare case of no wetting (8 = 180), it has been shown that the average terminal velocity of the inclusions is approximately 50% higher than that given by equation (2-62).55 For silica slags, the wetting angle (in the presence of air) is close to 115 ,54 indicating that the correction in the particle terminal velocity due to the slip at the oxide/metal interface in less than about 30%. Consequently, interfacial tension effects between the slag and the steel do not significantly affect the flotation rate of the particles, and, therefore, can be ignored. 2.11.2.2 Fluidflow pattern It is evident from the above discussion that the separation of deoxidation products during arc welding is controlled by the fluid flow fields set up within the weld pool. The typical flow pattern in a submerged arc weld pool is shown in Fig. 2.51. It can be seen from the figure that a depression is formed at the forward edge of the pool, which forces the melted metal to flow underneathand on either side of the depression, following the arrows in Fig. 2.51. At the rear of the weld pool the flow direction is reversed, and the metal streams back along the pool surface. A similar flow pattern has also been observed in GMA weld pools.?' As a result, the inclusions precipitated in the turbulent part of the weld pool are rapidly brought to the upper surface behind the arc due to the high-velocity flow field created within the liquid metal, and disengaged by the surface tension effects in the pool. The drag force exerted on the particles because of the liquid flow-field velocity can be estimated from published data for weld pool flow velocities following electromagnetic fluid mechanics theory. Normally, for weld pools the Peclet number for heat transfer lies within the
187
range from 10 to 5000, which indicates that the heat flow is predominately convectional.s- The limits quoted above for the Peclet number correspond to a range in the average weld pool fluid velocity from about 0.025 to 0.4 m s-1 in the case of SA welding.63,64 Unfortunately, specific information about the fluid flow velocity gradients in the weld pool are lacking, which prevents a more complete analysis of variations in the flow pattern with increasing distance from the root of the arc. Generally, the drag force, Fd, acting on a spherical particle in relative motion to a fluid can be expressed as:3 (2-64) where Cd is the drag coefficient, p is the fluid density, u' is the bulk velocity of the fluid relative to the particle, and d; is the particle diameter. The drag coefficient is, in tum, a function of the particle Reynolds number, NRe:3 (2-65) where f..L is the fluid viscosity. Typical physical-property data for liquid steel are given in Table 2.10. By inserting the values for p and f.L from Table 2.10, then using the relative velocity calculated in equation (2-63) (2.4 urn S-I) as one extreme and the fluid flow velocity (0.4 m S-I) as the other (i.e. assuming a stationary particle), one can demonstrate that the particle Reynolds numbers obtained never exceed 1.3 for most weld metal inclusions. The limit calculated above for the Reynolds number is within the so-called creeping flow region where Stokes law is indeed valid. Under such conditions, the ratio between the particle drag force Fd and the corresponding gravity force Fg is given by the expression.' (2-66) Note that equation (2-66) is the basis for obtaining Stokes law (equation (2-62. For steady motion of the particle, F d and F g are, of course, equal. However, the significance of Fd can also be interpreted in a transient sense, by considering the limiting case where the particle is stationary and its instantaneous motion is governed by the forces acting on it. In this case, the ratio of Fd (max) acting on the particle relative to the net gravity force (given by
Electrode
I;;Base plate
Pulsating cavity
Fig. 2.51. Typical flow pattern in SA weld pools. For clarity, the arc, slag and flux have been omitted. The sketch is based on the ideas of Lancaster. 32
188
equation (2-66)) can determine the dominant force responsible for the particle's trajectory. In Fig. 2.52 the above ratio has been calculated and plotted against the particle diameter, d.; for values of u' equal to 0.025 and 0.4 m s-l, which represent the typical velocities reported for SA steel weld pools.63, 64 It can be seen from the figure that the drag force is always several orders of magnitude greater than the gravity force for particles within the typical size range of weld metal inclusions (i.e. less than 2 urn in diameter). This result is to be expected, since the relative velocity, based on Stokes law and calculated in equation (2-63), is negligible compared with the liquid velocities in the weld pool (2.4 urn S-1 vs 0.02 to 0.4 m S-I). This calculation thus supports our previous conclusion that the separation of the oxide inclusions is controlled solely by the fluid flow behaviour in the weld pool. The fact that the phase separation proceeds under strongly turbulent conditions is also evidenced by the large number of iron droplets being mechanically dispersed in the top bead slag of GMA steel welds, as shown by the optical micrographs in Fig. 2.53. In the case of GMA welding, the non-metallic inclusions that are brought to the upper surface behind the arc coalesce rapidly to form large slag clusters that float on the top of the bead. Generally, re-entrapment of the slag does not occur owing to the decrease in the total surface free energy of the system, which is caused by the emergence of the inclusions from the weld metal. 54 Consequently, the slag will remain floating on the top of the bead even when welding is performed in the overhead position, as shown previously in Fig. 2.50. 2.11.2.3 Separation model Based mainly on experience with the GMA welding process, a simple model for the assessment of the sequence of deoxidation reactions in arc welding has been proposed,' and is shown schematically in Fig. 2.54. The model is based on the assumption that equilibrium between the reacting elements and precipitated slag is maintained down to low metal temperatures, and divides the weld pool into the following two main reaction zones:
1010~------------~---------------~------------~
100
0.1
~
10 Inclusion diameter, 11m ~
~
100
Fig. 2.52. The ratio between maximum particle drag force, Fd(max) and the corresponding gravity force, Fg, vs the inclusion diameter at two different flow velocities in the weld pool. Calculations are based on equation (2-66).
189
(a)
(b)
Fig. 2.53. Optical micrographs showing characteristic "comet's tails" of trapped iron droplets (light areas) in collected top bead slags of GMA steel welds; (a) Ar + 5vol%02, (b) Ar + 5vol%C02.
Electrode
(a)
Base plate
(b)
Sec t ion
A-A
Hot part of weld pool: Deoxidation/phase separation Final weld metal oxygen concentration
Oxygen content
Fig. 2.54. Schematic diagrams showing the sequence of reactions occurring during weld metal deoxidation; (a) Longitudinal section of weld pool, (b) Cross section of weld pool along A-A. The diagrams are based on the ideas of Grong and Christensen.'
190
(i) The hot part of the weld pool, characterised by simultaneous oxidation and deoxidation of the metal, where the separation of microslag takes place continuously as a result of highly turbulent flow conditions. (ii) The cold part of the weld pool, where precipitated slag will largely remain in the metal as finely dispersed particles as a result of inadequate melt stirring. Under such conditions equation (2-23) predicts a direct correlation between absorbed and retained (analytical) oxygen in the weld metal, i.e.:
= C4[%0]tot.
(2-67)
in agreement with experimental observations (see plots in Figs. 2.33, 2.38, and 2.39). Typically, the proportionality constant C4 varies between 0.1 to 0.2, which corresponds to a range in the t/t; ratio from 2.3 to 1.6. This shows that the boundary between 'hot' and 'cold' parts of the weld pool is not well defined, but depends on the welding system under consideration as well as on the operational conditions applied. 2.11.3 Predictions of retained oxygen in the weld metal Although the weld metal oxygen content is controlled by a transport mechanism in the weld pool, the concept of pseudo-equilibrium can still be used for an assessment of slag/metal reactions in arc welding. 2.11.3.1 Thermodynamic model In the case of silicon deoxidation of steel weld metals, we may write: Si (dissolved) + 2 Q (dissolved) = Si02 (slag) On introduction of the equilibrium constant for the reaction, we obtain: (2-69) To allow for the decrease in the silica activity with increasing manganese-to-silicon ratios, it is essential to establish a correlation that links the activity of silica to the concentrations of the deoxidation elements in the weld metal. A semi-empirical correlation of this kind has been presented by Walsh and Ramachandran.P derived from a re-analysis of activity data for silica in the Fe-Mn-Si-O system previously published by Hilty and Crafts.P" Within the temperaturerange from 1550 to 1650C, they showed that the silica activity in the deoxidation product can be approximately expressed as: (2-70) where K7 represents the manganese-to-silicon ratio at which the activity of silica becomes (2-68)
191
unity for a given temperature. A check of this equation against more recent data for silica activities in MnO-Si02 slags reported by 'Iurkdogan"? supports the findings of Walsh and Ramachandran'< that the activity of silica is approximately given by equation (2-70) for a wide range in the steel manganese-to-silicon ratio (i.e. from 0.1 to 50). By combining equations (269) and (2-70), it is possible to obtain an expression for the equilibrium oxygen content, solely in terms of the silicon and manganese concentrations: [%0]
= K8 ([%Si]
[%Mn])-O25
(2-71)
where K8 is a temperature-dependent parameter equal to (K7/ K6)o.5. The temperature dependence of the Si-O reaction (equation (2-68)) is well established and is approximately given by the relationship+' I og K 6 = +31035 T 12.02 (2-72)
when pure Si02 is used as the standard state for the silica activity. For a rough estimate of the temperature dependence of equation (2-70), the results of Turkdogan+ can be used. It should be noted that Walsh and Ramachandran'f did calculate the temperature dependence of K7 within the range from 1550 to 1650C. However, because equation (2-70) is empirical, the function cannot be extrapolated beyond these temperature limits. The data quoted in Ref. 55 are derived directly from the Si-Mn reaction (equation (2-55)) and activity data for MnO at silica saturation. On introduction of the equilibrium constant for equation (2-55), we obtain: (2-73) By using data from Ref.55, the initial [%Mn]2/[%Si] ratio for precipitation of silica saturated slags (equal to K9(aMnO)2) at 1500, 1550, 1600 and 1650C has been recorded and replotted against temperature, as shown in Fig. 2.55. The figure shows that the critical [%Mn]2/[%Si] ratio for precipitation of silica saturated slags is temperature dependent and decreases from about 5 at 1500C to below 1.5 at 1650C. On the basis of a crude extrapolation of the data to higher metal temperatures, it can, however, be argued that the ratio would approach a constant value of approximately 0.75 at temperatures beyond about 1750C (indicated by the broken horizontal line in Fig. 2.55). The above observation reflects the fact that the Si-Mn-O reaction equilibrium (equation (2-55)) is not very sensitive to a change in temperature (i.e. the enthalpy for the reaction is small). Over the composition range normally applicable to Si-Mn deoxidation of steel weld metals the observed threshold for the critical [%Mn]2/[%Si] ratio for precipitation of silica-saturated slags in Fig. 2.55 corresponds to a manganese-to-silicon ratio closely equal to unity. Consequently, at temperatures higher than about 1750C, K7 can, as a first approximation, be taken constant and independent of temperature (i.e. K7 = 1). The temperature dependence of equation (2-71) is thus simply (I/K6)o.5 or: - -15518 I og K 8 T
+.
6 01
(2-74)
192
5~--------------------------------------------~
t
0.75 - ---------~ .:--' --=1500
Fig. 2.55. The critical [%Mn]2/ [%Si] ratio for precipitation temperature. Data from Turkdogan.P
2.11.3.2 Implications of model Figure 2.56 shows plots of retained (analytical) oxygen in GMA and SMA steel weld metals vs the theoretical deoxidation parameter ([%Si][%Mn])-O.25, using relevant literature data.56,68 This parameter allows for the inherent decrease in the silica activity with increasing weld metal Mn to Si ratios. A closer inspection of the slopes of the curves reveals that the effective reaction temperature falls within the range calculated for chemical equilibrium between silicon, manganese, and oxygen at 1800 to 1900C. Although the boundary between 'hot' and 'cold' parts of the weld pool for possible inclusion removal in practice is not sharp, the results in Fig. 2.56 literally suggest that all oxides precipitated above 1800 to 1900C are simultaneously removed from the weld pool under the prevailing turbulent flow conditions. At lower temperatures, the degree of melt stirring is too low to promote separation of the deoxidation products out of the pool before solidification and hence, they are trapped in the metal solidification front in the form of finely dispersed inclusions. Moreover, the observed difference in the effective reaction temperature between GMA and SMA welding supports our previous conclusion that any change in welding parameters, flux or shielding gas composition, which alters the fluid flow pattern in the weld pool, will often have a stronger influence on the weld metal oxygen content than variations in the deoxidation practice. This implies that control of the weld metal oxygen level through additions of deoxidants, in practice, is difficult to achieve.
2.12 Non-Metallic
Inclusions commonly found in steel weldments will either be exogenous or indigenous, dependent on their origin. The first type arises from entrapment of welding slags and surface scale, while indigenous inclusions are formed within the system as a result of deoxidation reactions (oxides) or precipitation reactions (nitrides, sulphides). The latter group is almost always seen to be heterogeneous in nature both with respect to chemistry (multiphase parti-
193
~ ~
0
t
r:::
Q)
0.10
GMA Welding
0.08 0.06
o SMA Welding
..:(J
Q)
'E
r:::
0.02
0.4
0.8
([%Si] [%Mn])-O.25
1.2
1.6
Fig. 2.56. Examples of pseudo-equilibrium in GMA and SMA welding of C-Mn steels. The solid lines in the graph represent thermodynamical calculations at indicated temperatures. Data from Refs. 56 and 68.
cles), shape (angular or spherical particles), and crystallographic properties as a result of the complex alloying systems involved''? (see Fig. 2.57). An exception may be C-Mn steel welds, where the oxide inclusions will be predominately glassy, spherical, manganese silicates.' A survey of important weld metal inclusion characteristics is given in Table 2.11. 2.12.1 Volume fraction of inclusions It is evident from Table 2.11 that the volume fraction of non-metallic inclusions in steel weld metals normally falls within the range from 2 X 10-3 to 8 X 10-3, depending on the type of weld under consideration. Based on simple stoichiometric calculations it is possible to convert the analytical weld metal oxygen and sulphur contents to an equivalent inclusion volume fraction when the chemical composition of the reaction products is known. This is shown below.
Table 2.11 Summary of weld metal inclusion characteristics. The data are compiled from miscellaneous sources.
Size Distribution* Type of Weld Vv X 10-3 C-Mn steel weld metals Chemical Composition
a,
0.3-0.6
n, X
107
s,
10-50
Constituent elements Si,Mn,O,S (traces of AI, Ti, and Cu) AI,Ti,Si, Mn,O,S,N (Cu)
Reported phases Si02, MnOSi02, MnS, (CuxS) MnOAI203, 'Y-AI203' TiN, MnOSi02, Si02, a-MnS, (3-MnS,
(CuxS)
3-8
1-10
2-6
0.3-0.7
0.5-5
10-40
*Vv: volume fraction; dv: arithmetic mean (3-D) particle diameter (urn); Nv: number of particles per unit volume (mm='): Sv: total particle surface area per unit volume (mm? per mm-).
194
Fig. 2.57. Digital STEM brightfield image and Si, AI, S, Mn and Ti X-ray images of a multiphase weld metal inclusion. After Kluken and Grong.V Example (2.10)
Consider a GMA C-Mn steel weld which contains 0.040 wt% oxygen and 0.009wt% sulphur. Calculate the total volume fraction of non-metallic inclusions in the weld metal under the assumption that all oxygen and sulphur are tied up as MnOSi02 and MnS, respectively. Relevant physical data are given below: Density of steel : 7.8 g em>' Density of MnOSi02: 4.1 g cm' Density of MnS : 4.0 g cm'
CHEMICAL
REACTIONS
IN ARC
WELDING
195
Solution
First we calculate the total weight of retained MnOSi02 per 100 g weld metal:
_ 1 - -3 X
2
mMnOSiO
This corresponds to an equivalent volume fraction of: V (MnOSiO )= 0.11x7.8 =2.1x10-3 v 2 4.1 x 100 Similarly, we can calculate the weight and volume fraction of MnS in the weld metal: mMnS = 87.0 (0.009) g;::: 0.024 g 32.1 and V (MnS) = 0.024x7.8 =:0.5x10-3 v 4.0 x 100
The total volume fraction of MnOSi02 and MnS is thus: Vv = (2.1+0.5)x10-3 = 2.6x10-3
In practice, the stoichiometric conversion factors for oxygen and sulphur are virtually constant for a wide spectrum of inclusions 70 and hence, they can be regarded as independent of composition. Taking the solubility of sulphur in solid steel equal to 0.003 wt%, the following relationship is obtained for steel weld metalsr'": 70 Vv (cal) = 10-2 [5. O(wt % 0) + 5. 4(wt % s- 0.003)] (2-75)
The validity of equation (2-75) has been confirmed experimentally by comparison with microscopic assessment methods-"-" In steel weld metals the majority of the inclusions will be in the submicroscopic range owing to the limited time available for growth of the oxides. From the histogram in Fig. 2.58 it is seen that particles with diameters between 0.3 to 0.8 urn contribute to nearly 50 percent of the total inclusion volume fraction. This trend is not significantly changed by additions of strong deoxidisers, such .as aluminium and titanium, or by a moderate increase/decrease in the heat input. 57 2.12.2 Size distribution of inclusions As shown in Fig. 2.59 the majority of the three-dimensional (3-D) inclusion diameters fall within the range of 0.05 to 1.5 um, with a characteristic peak in the particle frequency at about 0.4 to 0.5 urn. These data obey the log-normal law, since a plot of the frequencies against the logarithms of the diameters approximately gives a symmetrical curve. Considering specific inclusion size classes, deoxidation with aluminium generally results in a higher fraction of coarse particles (> 1 urn) due to incipient clustering of AI203.57, 69 However, the observed particle clustering has no significant influence on the arithmetic mean 3-D inclusion diameter, as shown by the data in Fig. 2.59.
196
20 ~
"f0-
~0 0 ~
CI)
c:
g
Q)
E 10
~ Cii a:
>
2.0
Particle diameter (Jlm) ~
Fig. 2.58. Percental contribution of different size classes to the total volume fraction of non-metallic inclusions in a low-alloy steel weld metal. Data from Kluken and Grong.V
197
(a)
30r----------r--------~----------~--------_
t
~ c:
Q)
::::J
~ 10 u.
C"
00
0.5
1.0
Particle diameter (j.J.m)
--...
1.5
2.0
(b)
30
.....
~ ~ ~ c:
Q)
::s
0"
u,
10
Fig. 2.59. Three-dimensional (3-D) size distribution of non-metallic inclusions in two different lowalloy steel weld metals; (a) Low weld metal aluminium and titanium levels, (b) High weld metal aluminium and titanium levels. Data from Kluken and Grong.F
(2-76)
Here do is the initial particle diameter, o is the oxide-steel interfacial energy, Dm is the element diffusivity, em is the element bulk concentration, V~ is the molar volume of the oxide per mole of the diffusate, and t is the retention time.
198
30
~ ~ c:
Q)
t
~
20
::s
C'"
u,
10
0.5
1.0
Particle diameter, Jl m ~
1.5
2.0
Fig. 2.60. Effect of heat input on the 3-D inclusion size distribution in low-alloy steel weld metals. Data from Kluken and Grong.I?
t
~
E
Q)
....:
Q)
:::L
..---"----dvaE
1/3
E
tU
Q)
13
~ I n,
10
Fig. 2.61. Variation of arithmetic mean 3-D inclusion diameter with heat input during SA welding. Data from Kluken and Grong.F
For welding of thick plates, the time available for growth of particles in the 'cold' part of the weld pool can be estimated from" the Rosenthal equation, i.e. equation (1-45) in Section 1.10.2 (Chapter 1). If the characteristic length of the cooling zone is taken equal to the weld ripple lag (defined in Fig. 2.63), the retention time t is approximately given by the following relationship:
199
1800
o
~
E
2.0
6.0
Fig. 2.62. Measured temperature level in the trailing edge of the weld pool during GMA welding. Data from Kluken and Grong.F
= 0)
xm =V
Xm,
t(s) ~ 2. 1311E
(2-77)
where n is the arc efficiency (equal to about 0.95 for SAW and 0.80 for GMAW /SMAW), and E is the gross heat input (kJ mm '). Note that equation (2-77) assumes constant values for the steel thermal diffusivity and volume heat capacity (5 mm? S-1 and 0.0063 J mm " C-1, respectively), and no preheating (To * = 20C). In Fig. 2.64 the Ostwald ripening theory has been tested against relevant literature data, which may be considered representative of the 3-D particle size distribution. Although there is some scatter in the data, the observed inclusion growth rates fall within the range calculated for oxygen diffusion-controlled coarsening of Si02 and Al203 at 1550C, using the Wagner equation. In these calculations, a reasonable average value for the bulk diffusivity of oxygen has been assumed (i.e. 10-2 rnm'' S-I).55 If the effective inclusion growth rate constant for lowalloy steel weld metals is taken equal to the slope of the curve in Fig. 2.64, the following relationship is achieved:
200
t
:::t
1.0
GMAW
0.8
SAW
SMAW
E 0.6
3.0
Fig. 2.64. Relation between arithmetic mean (3-D) inclusion diameter d; and retention time t for different arc welding processes. Data compiled by Kluken and Grong."
-1900-1800C
Gas/metalslag/metal reactions
Ostwald ripening
Fig. 2.65. Proposed deoxidation model for steel weld metals (schematic). The diagram is based on the ideas of Kluken and Grong.l?
(2-78) By substituting equation (2-77) into equation (2-78), we obtain a direct correlation between the arithmetic mean 3-D inclusion diameter d; and the net heat input 11E:
d; (urn) ;:::: O.35(l1E)1/3 (2-79)
201
Equation (2-79) predicts that d; is a simple cube root function of E, in agreement with the experimental data in Fig. 2.61. It should be noted that the measured shape of the particle distributions (see Figs. 2.59 and 2.60) deviates somewhat from that required by the Wagner equation, which assumes a quasistationary distribution curve, and that the maximum stable particle diameter is about 1.5 times the mean diameter of the system.P Although the origin of this discrepancy remains to be resolved, this suggests that particle clustering is also a significant process in steel weld metals as it is in other metallurgical systems. In fact, such effects would be expected to be most pronounced at high aluminium levels because of a large interfacial energy between Al203 and liquid steel, in agreement with experimental observations. 57, 69 2.12.2.3 Proposed deoxidation model Referring to Fig. 2.65, the sequence of reactions occurring during weld metal deoxidation can be summarised as follows. In general, nucleation of oxide inclusions occurs homogeneously as a result of the supersaturation established during cooling in the weld pool. The diffusioncontrolled deoxidation reactions (i.e. diffusion of reactants to the oxide nuclei) will be essentially complete when the liquid metal temperature attains a constant level of about 1550C at some distance behind the root of the arc. Growth of the particles may then proceed under approximately isothermal conditions at a rate controlled by the Wagner equation until the temperature reaches the melting point of the steel. Since retention times in welding generally depend on the heat input, it follows that choice of operating parameters will finally determine the degree of particle coarsening to be achieved.
Example (2.11)
Consider SA welding with a basic flux on a thick plate of low-alloy steel under the following conditions: 1= 500A, U = 31V, V = 6mm s',11= 0.95, To = 20C Previous experience has shown that this steel/flux combination gives a weld metal oxygen and sulphur content of 0.035 and 0.008 weight percent, respectively. Based on the stereometric relationships given below, calculate the total number of particles per unit volume Ni; the total number of particles per unit area Na, the total particle surface area per unit volume Sv, and the mean particle centre to centre volume spacing Av in the weld deposit: 74, 75
N=~
v
n(dv)3
v
(2-80) (2-81)
N a =Nd v
Sv
= nN v (d v )2
(2-82)
Av = 0.554 :v
( f3
(2-83)
202
Solution
First we calculate the total volume fraction of oxide and sulphide inclusions from equation (2-75): v, = 10-2 [5.0(0.035) + 5.4(0.008 -0.003)] = 2.0 X 10-3 The arithmetic mean (3-D) inclusion diameter can then be evaluated from equation (2-79): d This gives: Nv
= 0.35 ( 0.95
= 0.471lm
N; = 3.68 x 107 (0. 47 X 10-3) particles per mm2 = 1.73 X 104 particles per mm2
A =0.554 ( v 3.68xl0
1.7Jlm
2.12.3 Constituent elements and phases in inclusions It is evident from Table 2.11 that non-metallic inclusions commonly found in low-alloy steel weld metals may contain a considerable number of constituent elements and phases. 2.12.3.1 Aluminium, silicon and manganese contents Figure 2.66 shows examples of measured X -ray intensity histograms for silicon, manganese, and aluminium in inclusions extracted from a low-alloy steel weld metal. These results have been converted into arbitrary elemental weight concentrations by normalising the collected Xray data for all preselected elements (i.e. AI, Si, Mn, Ti, Cu, and S) to 100%. The characteristic normal distribution curves recorded for silicon, manganese, and aluminium show that the content of each oxide phase may vary significantly from one particle to another. This observation is not surprising, considering the complex chemical nature of the weld metal inclusions (see Fig. 2.57). 2.12.3.2 Copper and sulphur contents In addition, the inclusions may contain significant levels of both copper and sulphur in addition to aluminium, silicon, manganese and titanium. Sulphide shells around extracted inclusions have frequently been observed in SA and SMA steel welds, often in combination with copper. This has been taken as an indication of copper sulphide formation. However, based on the wavelength dispersive X-ray (WDX) data reported by Kluken and Grong.V it can be argued that copper sulphide is a rather unlikely reaction product in steel weld metals as it is in ordinary ladle-refined steel." From their data it is evident that the total
203
Silicon
Manganese
Aluminium
extracted from a low-alloy steel weld metal. Arrows indicate average composition. Data from Kluken and Grong.S7
number of counts recorded for copper in discrete particles is not significantly higher than the corresponding matrix value, which shows that the copper content of the inclusions is low. Since these measurements were carried out on mechanically polished specimens and not on carbon extraction replicas, as done in the EDX analysis, the indications are that the higher inclusion copper level observed in the latter case mainly results from surface copper contamination inherent from the extraction process. In contrast, the WDX analysis of sulphur revealed evidence of sulphur enrichment in most of the particles. Considering the fact that these particles also contained significant amounts of manganese, it is reasonable to assume that most of the sulphur is present in the form of MnS (possibly with some copper in solid solutionj." 2.12.3.3 Titanium and nitrogen contents From the literature reviewed, it is apparent that conflicting views are held about the role of titanium in weld metal deoxidation. From a thermodynamic standpoint, Ti203 is the stable reaction product in titanium deoxidation." but this phase has not yet been detected experimentally in steel weldments (only in continuous cast steel)."? However several authors have reported the presence of crystalline patches containing titanium toward the edges of inclusions.78-8o This constituent has a cubic crystal structure with a lattice parameter close to 0.42 nrn, conforming to either 1- TiO, TiN, or TiC (note that the 1- TiO phase is the hightemperature modification of the titanium monoxide). 81 In general, formation of titanium monoxide requires strongly reducing conditions, which
204
implies that the "I-TiO constituent is not an equilibrium reaction product in steel deoxidation.Pt '" Hence, it is reasonable to assume that the titanium-rich crystalline patches observed toward the edges of weld metal inclusions are titanium nitride. This conclusion has later been confirmed experimentally by Kluken and Grong."? 2.12.3.4 Constituent phases Based on the above discussion, it is possible to rationalise the formation of primary and secondary reaction products (i.e. oxides, sulphides and nitrides) during cooling in the weld pool, as shown in Fig. 2.67. In general, the inclusions will consist of an oxide core which is formed during the primary deoxidation stage. The chemical composition of the deoxidation product can vary within wide limits, depending on the activities of AI, Ti, Si, Mn, and 0 in the weld metal. The surface of the oxides will partly be covered by MnS and TiN (see also STEM micrograph in Fig. 2.57). Precipitation of these phases occurs after the completion of the weld metal deoxidation, probably during solidification, where the reactions are favoured by solute enrichment in the interdendritic liquid. Additional precipitation of TiN may occur in the solid state as a result of diffusion of titanium and nitrogen to the surface of the inclusions. 2.12.4 Prediction of inclusion composition Since the diffusion-controlled deoxidation reactions are completed within a fraction of a second when the number of nuclei is of the order of 107 mm=' or higher,55 the average chemical composition of the inclusions should be compatible with that calculated for chemical equilibrium at temperatures close to the melting point (e.g. 1550C). 2.12.4.1 C-Mn steel weld metals Over the composition range normally applicable to silicon-manganese deoxidation of steel weld metals (i.e. between 0.4 to 0.7 wt% Si and 0.8 to 1.5 wt% Mn) the equilibrium reaction product at 1550C should be silica-saturated slags with a mole fraction of Si02 close to 0.55. 55,65 Since the two other slag components are MnO and FeO, we may write: (2-84) The activity coefficients for MnO and FeO in the ternary system Si02-MnO-FeO computed from the equations presented by Sommerville et al.: 82 can be
(2-85) and (2-86) For the specific case of silica-saturated slags, we obtain: logy and log y FeO
MnO
= -2.2(0.55)2
-1.0(0.55)(N
FeO)
= 0.55N
= O.55N
MnO -
0.91
(2-87)
= -1.
MnO)
MnO -
0.36
(2-88)
205
Oxide Core (primary reaction product' MnS (secondary reaction product) Fig. 2.67. Schematic diagram showing the presence of primary and secondary phases in weld metal inclusions.
Under such conditions the activity ratio aMnc/aFeO in the slag is given by: aMnO aFeO
(2-89)
The activity ratio aMnc/aFeO can also be estimated from the equilibrium constant for the FeMn reaction at 1550C,23 i.e.: a MnO ~ 5.93[%Mn] (2-90) aFeO The corresponding mole fractions of MnO and FeO in the slag phase are then obtained by combining equations (2-89) and (2-90): N MnO and N FeO
=
9. 53[%Mn]
(1 + 21. 18[%MnD
(2-91)
0.45
(1 + 21. 18[%Mn])
(2-92)
Equations (2-91) and (2-92) provide a basis for calculating the chemical composition of the inclusions under different deoxidation conditions. A requirement is, however, that the weld metal Si to Mn ratio is sufficiently high to promote precipitation of silica-saturated slags at 1550C.
Example (2.12)
Consider a GMA steel weld which contains 0.6 wt% Si and 1.3 wt% Mn. Based on equations (2-91) and (2-92), calculate the chemical composition of the weld metal oxide inclusions at 1550C.
Solution
MnO
206 and
N
FeO
This gives the following chemical composition of the inclusions (in wt%):
wt%SiO
2
60.1xO.55 100%=51% (60.1 x 0.55) + (70.9 x 0.43) + (71.8 x 0.02) 70.9xO.43 100%=47% (60.1 x 0.55) + (70.9 x 0.43) + (71.8 x 0.02) 71.9xO.02 100%=2% (60.1 x 0.55) + (70.9 x 0.43) + (71.8 x 0.02)
wt%MnO=
wt%FeO=
The above calculations should be compared with the composition data presented in Fig. 2.68. It is evident from this plot that the agreement between predictions and experiments is reasonably good both at high and low weld metal manganese levels. This justifies the simplifications made in deriving equations (2-91) and (2-92). 2.12.4.2 Low-alloy steel weld metals In principle, a procedure similar to that described above could also be used to establish a theoretical basis for predicting the chemical composition of the inclusions in the case lowalloy steel weld metals. Unfortunately, adequate activity data for the Fe-AI- Ti-Si-Mn-O system are not available in the literature. An alternative approach would be to calculate the average inclusion composition from simple mass balances, assuming that all oxygen combines stoichiometrically with the various deoxidation elements to form stable oxides in the order A1203, Ti203, Si02, and MnO, according to their oxygen affinity in liquid steel (see Fig. 2.69). If reasonable average values for the inclusion and steel densities are used (i.e. 4.2 and 7.8 g cnr', respectively), the following set of equations* can be derived from a balance of 0, AI, Ti, S, Si, and Mn and data for acid soluble aluminium and titanium in the weld metal.F
Aluminium
The average aluminium content of the inclusions, [%Al]incz., can be estimated from the measured difference between total and acid soluble aluminium in the weld metal, (Ll %Al)weld. This difference is, in turn, equal to the total mass of aluminium in the inclusions: (~%Al) weld
= minel. [%Al]inel. = V (cal) 4.2 [%Al].
100
7.8
mel.
(2-93)
where mind. and Vv (cal) are the total mass and volume fraction of non-metallic inclusions in the weld deposit, respectively.
*The numerical constants in the constitutive equations given below could alternatively be expressed in terms of atomic weights etc. to bring out more clearly their physical significance (e.g. see the treatment of Bhadeshia and Svensson'").
207
Ar .. 02 gas mixtures
o 0.6
(1)
r---------------~~~~~--~~~
(5
::?!
0.2
0.4
0.6
1.0
1.2
1.4
Manganese content, wt% --Fig. 2.68. Comparison between measured and predicted microslag composition in GMA welding of C-Mn steels. Solid lines represent theoretical calculations based on equations (2-91) and (2-92). Data from Grong and Christensen.'
104--~~~~--~~--~~~--~--~~~--~~~ 2 1'0-3 10 ..
1 10 ..
= 1.86 (A%Al)weld
Vv(cal)
(2-94)
208
However, since data for acid soluble aluminium (and titanium), in practice, may contain large inherent errors, the following restriction applies: (~%Al)weld ~ 1. 13[%O]anal. (2-95)
Similarly, the average titanium content of the inclusions, [%Ti]inel., can be estimated from the measured difference between total and acid soluble titanium, (a%Ti)weld' However, since TiN dissolves readily in strong acid, it is necessary to include an empirical correction for the amount of titanium nitride which simultaneously forms at the surface of the inclusions during solidification. This can be done on the basis of published data for the solubility product of TiN in liquid steel. 84 If we assume that the nitrogen content of the inclusions is proportional to the calculated difference between total and dissolved nitrogen at the melting point (1520C), the following relationship is obtained:
(~%Ti).
mel.
=~(
=
4.2
1 Vv(cal)
J[(~%Ti)
weld
+ 47.9 ([%N]
14.0
anal.
1.86 Vv(cal)
[(~%Ti)
weld
+ 3. 42([%N]
anal.
_ 7.21XIO- Jl [%Ti]sol.
where [%N]anal. is the analytical weld metal nitrogen content, and [%Ti]sol. is acid soluble titanium. Note that the correction term forTiN, in practice, neither can be negative nor exceed [%Ti]sol. In this case, the maximum amount of titanium which can be bound as Ti203 is determined by the overall oxygen balance: (tl%Ti)weld
Sulphur
~ 2. O([%O]anal. - O. 89(~%AI)weld)
(2-97)
If the solubility of sulphur in solid steel is taken equal to 0.003 wt%,70 the average sulphur content of the inclusions, [%S]incl. is given by: [%Slnel. 1.86 ([%S]anal. - 0.003) Vv (cal)
(2-98)
From the experimental data of Saggese et al. 85 reproduced in Fig. 2.70, it is evident that the mass ratio between Si02 and MnO in the oxide phase may be considered constant and virtually independent of composition (Le. equal to about 0.94). This implies that the average silicon content of the inclusions, [%Si];nel., can be calculated from a balance of oxygen:
209
0\0
....', ......
....
40
MnO "----2
Fig. 2.70. Measured inclusion compositions in low-alloy steel weld metals. Data from Saggese et al.85
[%Si]incl. ==
V:C ~:l)
Considering manganese, proper adjustments should also be made for the amount of MnS formed at the surface of the inclusions during solidification. Hence, the average manganese content of the inclusions, [%MnJ;ncl., is given by the sum of the oxygen and the sulphur contributions:
+ 3.18
Vv (cal)
([%S]
anal.
- 0.003)
(2-100)
In Fig. 2.71, the accuracy of the model has been tested against the experimental SAIGMA inclusion data reported by Kluken and Grong.V taking the sum (%AI + %Ti + %S + %Si + %Mn) equal to 100%. A closer inspection of the graphs reveals a reasonable agreement between calculated and measured average inclusion compositions in all cases, which confirms that the model is sound. Included in Fig. 2.71 is also a collection of data for SMA low-alloy steel weld metals (3 kJ mm' - basic electrodes). Since these results follow the same pattern, it implies that the model is generally applicable and, therefore, can be adopted to all relevant arc welding processes.
210
t
~ ~
0
100
80
c .. c. E
0
~0 60
-0
-c
CJ)
~ :::I
co
CD
40
t
~ ~
C
0
E Ul 60 0 a.
"'0
8
~ :::I
CJ)
40
):(
);(
CD
Fig. 2.71. Comparison between measured and predicted inclusion compositions; sulphur, (b) Titanium, manganese and silicon. Data from Kluken and Grong-?: 86
(a) Aluminium
and
Implications of model
It can be inferred from equations (2-94) and (2-95) that the chemical composition of the inclusion oxide core is directly related to the aluminium to oxygen ratio in the weld metal. Referring to Fig. 2.72', the fraction of MnOAl203 and )'-AI203 in the inclusions is seen to increase steadily with increasing (Ll %Al)weli[% O]anal. ratios until the stoichiometric composition for precipitation of aluminium oxide is reached at 1.13. At higher ratios, the deoxidation product will be pure A1203, since aluminium is present in an over-stoichiometric amount with respect to oxygen. When titanium is added to the weld metal, titanium oxide (in the form of Ti203) may also enter the reaction product. At the same time both TiN and a-MnS form epitaxially on the surface of the inclusions during solidification. Consequently, in AI-Ti-Si-Mn deoxidised steel weld metals the total number of constituent phases within the inclusions may approach six. The kinetics of inclusion formation are further discussed in Ref. 87.
211
Si02
(+ Ti203 + TiN + a-MnS)
Fig. 2.72. Coexisting phases in inclusions at different weld metal aluminium-to-oxygen ratios. Shaded region indicates the approximate composition range for the oxide phase. The diagram is constructed on the basis of the model of Kluken and Grong-? and relevant literature data.
Example (2.13)
Consider SA welding of low-alloy steel with two different basic fluxes. Data for the weld metal chemical compositions are given in Table 2.12. Based on Fig. 2.72, estimate the total number of constituent phases in the inclusions in each case.
Solution
It follows from Fig. 2.72 that the chemical composition of the deoxidation product is determined by the aluminium to oxygen ratio in the weld metal. For weld A, we have:
(Ll %Al)weld
= 1.19
[%O]anal.
Since this ratio is higher than the stoichiometric factor of 1.13, all oxygen is probably tied up as aluminium oxide .. In addition, weld A contains small amounts of titanium and sulphur, which may give rise to precipitation of TiN and MnS at the surface of the inclusions during solidification. Hence, the three major constituent phases in the weld metal inclusions are 1'-AI203, TiN, and a-MnS.
212
Table 2.12 Chemical composition of SA steel weld metals considered in Example 2.13.
Weld No. A C (wt%) 0.09 0 (wt%) 0.021 Si (wt%) 0.45 Mn (wt%) 1.52 S (wt%) 0.01 N (wt%) 0.006 AI* (wt%) 0.028 (0.003) 0.028 (0.003) Ti* (wt%) 0.010 (0.009) 0.028 (0.010)
0.09
0.045
0.45
1.52
0.01
0.006
*Data for acid soluble Al and Ti are given by the values in brackets.
In the case of weld B the situation is much more complex due to a higher content of oxygen and titanium. From Table 2.12, we have: (L1%AI)weld [%O]anal. and [%O]excess
= (0.028-0.003) = 0.56
0.045
= [%O]anal.
- 0.89(L1%AI)weld
- 0.5(L1%Ti)weld = 0.014
= 0.045-0.89(0.025)-0.50(0.018)
These calculations show that Al and Ti are not present in sufficient amounts to tie-up all oxygen. Under such conditions Fig. 2.72 predicts that the total number of constituent phases in the inclusions is six, i.e.: Si02, MnOAI203, ,),-AI203, Ti203, a-MnS and TiN.
References
0. Grong and N. Christensen: Scand. 1. Metall., 1983,12,155-165. J.F. Lancaster: Metallurgy of Welding, 3rd Edn, 1980, London, George Allen & Unwin Ltd. J. Szekely and N.J. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John Wiley & Sons, Inc. N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Norwegian Institute of Technology, Trondheim, Norway. E.T. Turkdogan: Physical Chemistry of Oxygen Steelmaking, Thermochemistry and Thermodynamics, 1970, United States Steel Corporation. ER. Coe: Welding Steels without Hydrogen Cracking, 1973, Abington (Cambridge), The Welding Institute. Doc. IISIIIW-532-77: Weld. World, 1977,15,69-72. B. Chew and R.A. Willgoss: Proc. Int. Conf. Weld Pool Chemistry and Metallurgy, London, April 1980, Paper 25,155-165. Publ. The Welding Institute (England). F. Matsuda, H. Nakagawa, K. Shinozaki and H. Kihara: Trans. fWRI, 1978,7,135-137. B. Chew: Weld. i.. 1973,52, 386s-391s. G.M. Evans: Schweissmitteilungen (Oerlikon), 1977, No. 79,4-8. G.M. Evans and H. Bach: IIW Doc. IIA-363-74, 1974.
213
13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.
R. Roux: Rev. Met. 1954,192-209. E.M. Skjolberg: M.Sc. thesis, 1980, University ofTrondheim, The Norwegian Institute of Technology, Trondheim, Norway. D.G. Howden and D.R. Milner: Brit. Weld. t; 1963,10,304-316. T. Kobayashi, T. Kuwana and T. Kikuchi: IIW Doc. XII-265-65, 1965. J.E. O'Brien and M.F. Jordan: Met. Constr. Brit. Weld. 1971,3,299-303. M. Uda and S. Ohno: Trans. Nat. Res. Inst.for Metals, 1973,15,20-28. o. Morigaki, T. Tanigaki, M. Kuwabara and K. Fujibayashi: IIW Doc. 11-746-75, 1975. H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: f. Mater. Sci. Lett., 1988,7,610-612. E. Halmey: Proc. Int. Conf. on Arc Physics and Weld Pool Behaviour, London, May, 1979,4957. Publ. The Welding Institute (England). D.J.H. Corderoy, B. Wills and G.R. Wallwork: Proc. Int. Con! on Weld Pool Chemistry and Metallurgy, London, April 1980, 147-153. Publ. The Welding Institute (England). J.F. Elliott, M.Gleiser and V. Ramakrishna: Thermochemistry for Steelmaking (Vol. 11),1963, London, Addison-Wesley Publ. Company (Pergamon Press). Doc. IISIIIW-343-70: Weld. World, 1970, 8, 28-35. I. Barin and o. Knacke: The Thermochemical Properties of Inorganic Substances, 1973, Berlin, Springer Verlag. Janaf, Thermochemical Tables - 2nd Edn (NSRDS); 1971, New York, National Bureau of Standards. R.F. Heile and D.C. Hill: Weld. i.. 1975, 54, 201s-210s. B. Ozturk and R.I. Fruehan: Metall. Trans. B, 1985,16B, 801-806. G. Jelmorini, G.W. Tichelaar and G.J.P.M. Van den Heuvel: IIW Doc. 212-411-77, 1977. O. Kubaschewski and C.B. Alcock: Metallurgical Thermochemistry, 5th Edn, 1979, Oxford, Pergamon Press. P.A. Distin, S.G. Whiteway and C.R. Masson: Can. Metall. Quart., 1971, 10, 13-18. J.F. Lancaster: The Physics of Welding, 1984, Oxford, IIW-Pergamon Press, 204-267. W.A. Fischer and J.F. Schumacher: Arch. Eisenhiittenwesen, 1978,49,431-435. N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. i; 1965, 12,54-75. G.K. Sigworth and IF. Elliott: Metall. Trans., 1973,4,105-113. D.N. Shackleton and A.A. Smith: IIW Doc. XII-832-84, 1984. 0. Grong, N.H. Rodi and N. Christensen: Scand. f. Metall., 1984,13,154-156. A.A. Smith et al.: Weld. World., 1970,8,28-35. A.A. Smith: C02-Shielded Consumable Electrode Arc Welding, 2nd Edn, 1965, Brit. Weld. Res. Assoc. U. Lindborg: Met. Constr. and Brit. Weld. L, 1972,4,52-55. T.H. North, H.B. Bell, A. Nowicki and I. Craig: Weld. i; 1978,57, 63s-71s. C.S. Chai and T. W. Eagar: Metall. Trans., 1981, 12B, 539-547. U. Mitra and T.W. Eagar: Metall. Trans., 1984, 15A, 217-227. J.E. Indacochea, M. Blander, N. Christensen and D.L. Olson: Me ta ll. Trans., 1985, 16B, 237245. N. Christensen and 0. Grong: Scand. J. Metall., 1986, 15, 30-40. U. Mitra and T.W. Eagar: Metall. Trans., 1991, 22B, 65-100. T.W. Eagar: Weld. i.. 1978,57, 76s-80s. N. Christensen and J. Chipman: Weld. Res. Coun. Bull. Series, 1953, No. 15, New York, Welding Research Council (USA). A. Apold: Carbon Oxidation in the Weld Pool, 1962, Oxford, Pergamon Press. D.A. Wolstenholme: Proc. Int. Con! on Trends in Steel and Consumables for Welding, London, 1978,123-134. Publ. The Welding Institute (England). 0. Grong, D.L. Olson and N. Christensen: Metal Constr., 1985, 17, 810R-814R.
i:
214
52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85.
86. 87.
P. Nilles, P. Dauby and J. Claes: Proc. Int. Con! Basic Oxygen Steelmaking -A New Technology Emerges, London, 1978,60-72, The Metals Society (England). U. Lindborg and K. Torsell: Trans. TMS-AIME, 1968,242,94-102. N.F. Grevillius: Jernkont. Ann., 1969,153, 547-572. E.T. Turkdogan: Proc. Int. Con! on Chem. Me ta ll. of Iron and Steel, Sheffield, July 1971,153170, Publ. The Iron and Steel Institute (England). 0. Grong. T.A. Siewert, G.P. Martins and D.L. Olson: Metall. Trans. A, 1986,17A, 1797-1807. A.O. Kluken and 0. Grong: Me tall. Trans. A, 1989, 20A, 1335-1349. M.L. Turpin and J.E Elliott: J. Iron Steel Inst., 1966,204,217-225. E.T. Turkdogan: Physicochemical Properties of Molten Slags and Glasses, 1983, London, The Metals Society. E. Plockinger and M. Wahlster: Stahl und Eisen, 1960,80,659-669. B.J. Bradstreet: IIW Doc. 212-138-68, 1968. J .F. Lancaster: Phys. Technol., 1984, 15, 73-79 F. Eickhom and A. Engel: IIW Doc. 212-201-70,1970. N. Mori and Y. Horii: IIW Doc. 212-188-70, 1970. , R.A. Walsh and S. Ramachandran: Trans. TMS-AIME, 1963,227,560-562. D.C. Hilty and W. Crafts: Trans.-AIME, 1950,188,425-436. E.T. Turkdogan: Trans. TMS-AIME, 1965,233,2100-2112. G.M. Evans: IIW Doc. IIA-630-84, 1984. 0. Grong and D.K Matlock: Int. Met. Rev., 1986,31, 27-48. A.G. Franklin: J. Iron Steel Inst., 1969,207, 181-186. A.O. Kluken, 0. Grong and J. Hjelen: Mat. Sci. Technol., 1988,4,649-654. L.M. Hocking: Quart. J. Royal Meterol. Soc., 1959, 85, 44-50. C. Wagner: Z. Electrochemie, 1961,65,581-591. E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co. M.F. Ashby and R. Ebeling: Trans. TMS-AIME, 1966,236,1396-1404. R. Kiessling: Non-Metallic Inclusions in Steel, 1978, London, The Metals Society (TMS). H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. i.. 1987,66, 301s-309s. G. Thewlis: IIW Doc. IIA-736-88, 1988. J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans. A, 1986, 17A, 1611-1623. G.M. Evans: Metal Constr., 1986,18, 631R-636R. J.L. Murray and H.A. Wriedt: Bull. Alloy Phase Diagr., 1987,8,148-165. I.D. Sommerville, I. Ivanchev and H.B. Bell: Proc. Int. Con! Chem. Metall. of Iron and Steel, Sheffield, July 1971,23-25, Publ. The Iron and Steel Inst. (1973). H.K.D.H. Bhadeshia and L.E. Svensson: Mathematical Modelling of Weld Phenomena (Eds H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, 109-180. S. Matsuda and N. Okumura: Trans. ISIJ, 1978, 18,198-205. M.E. Saggese, A.R. Bhatti, D.N. Hawkins and J.A. Whiteman: Proc. Int. Con! on the Effect of Residual, Impurity and Micro-Alloying Elements on Weldability and Weld Properties, London, Nov. 1983, Paper 15, Publ. The Welding Institute (England). A.O. Kluken and 0. Grong: Report No. STF34 F87125, 1987, Sintef, Trondheim, Norway. S.S. Babu, S.A. David, 1M. Vitek, K. Mundra and T. DebRoy: Mater. Sci. Technol., 1995,11,186199.
215
Ix
g
activity coefficient of element X gravity constant (9.81 m s-2) standard free energy of reaction (J mol-lor kJ mot+) energy barrier for homogeneous nucleation (kJ mol:') driving force for precipitation of oxide inclusions (J m=')
8Go
aMxOy
8G*
B.I.
CI... -C5
HDM
Dm
E
FCAW Fd
Fg
HJIS
FFR
216 1
nucleation rate of oxide inclusions in the weld pool (nuclei m-3 s-l) constant in equation (A2-1) (nuclei m-3s-1) ratio between absorbed Si and Mn in the weld metal coefficient of weld metal deposition (g A-IS-I) mass transfer coefficient for gas absorption (mm S-I) mass transfer coefficient for gas desorption (mm s')
total mass of liquid metal leaving/entering the reaction zone per unit time (g s') arbitrary flux or slag component concentration displacement of manganese referred to different standard states (wt%) total manganese absorption in the weld metal due to slag/metal interactions (wt%) rejected manganese during cooling in the weld pool (wt%) oxidation loss of manganese at electrode tip (wt%) atomic weight of element X dimensionless parameter operating
10
k
,
d[%Mn]
k'
ka
[%Mn]abs.
k'd
[%Mn] kd overall mass transfer coefficient (mm S-I) equilibrium constants the product [%C] [%0] Mx
me
rej.
K1K9
m
[%Mn]
ox.
mass of CO2 per 100 g of electrode coating (g or wt%) total mass of inclusions per 100 g weld deposit (g) mass of H20 per 100 g of electrode coating (g or wt%) total weight of retained MnOSi02 (g per 100 g weld deposit) total weight of precipitated microslag (in g per 100 g weld deposit) total weight of precipitated MnS (g per 100 g weld deposit)
n3
mind.
nI
total number of moles of component! absorption/desorption rate of element X in the weld pool (mol s') total number of moles of component X total number of moles of component X2 number of particles per unit area (mm-) Avogadro constant (6.022 X 1023 mol ")
mw
nx
nx
mMnOSi02
ms1ag
nX2
.
:
mMnS
217
combined partial pressure of H2 and H20 (atm) post weld heat treatment parts per million (g per ton) net arc energy (W) universal gas constant (J K-I mol") relative humidity (%) analytical weld metal sulphur content (wt%) standard entropy of reaction (J K-I mor-') concentration displacement of silicon referred to different standard states (wt%) total silicon absorption in the weld metal due to slag/ metal interactions (wt%) rejected silicon during cooling in the weld pool (wt%) oxidation loss of silicon at electrode tip (wt%) solubility of element X at 1 atm total pressure (ml per 1~Og, ppm or wt% ) total particle surface area per unit volume (mm/ per mm-') modified solubility of hydrogen at 1 atm total pressure (ml per 100 g, ppm or wt%)
NMO
mole fraction of arbitrary slag component analytical weld metal nitrogen content (wt%) mole fraction of CO in gas phase Reynolds number number of particles per unit volume (No. per mrrr") calculated oxygen absorption in electrode tip or falling droplets (wt%) analytical weld metal oxygen content (wt%)
Pw
PWHT ppm
[%N]anal.
Nco
qo R
NRe Nv
R.A. [%S]anal.
[%O]abs.
ilSo
[%O]anal.
il[%Si] [%O]eq. equilibrium weld metal oxygen content (wt%) . rejected oxygen during cooling in the weld pool
(wt%)
[%0]
rej.
[%Si]abs.
[%O]tot.
total oxygen absorption in hot part of weld pool (wt%) total pressure (atm or bar)
[%Si]
rej.
:
Px
[%Sijox. partial pressure of component X in bulk gas phase (atm) equilibrium partial pressure of component X at gas/ metal interface (atm) partial pressure of component X2 in bulk gas phase (atm) Sw
Sx
P~
s,
PX2
pO x,
218
SAW SMAW STEM
Vv
volume fraction of inclusions concentration of component I (vol%) concentration of component X2 (vol%) concentration of component YX2 (vol %) wavelength dispersive Xray analysis wire feed rate (mm S-1 or m min-I) arbitrary element or gaseous species weld ripple lag (mm) concentration of element X in the weld metal (wt%) concentration displacement of element X referred to nominal weld metal composition (wt%) equilibrium concentration of element X in the weld pool (wt%) initial concentration of element X in the weld pool (wt%) average content of element X in inclusions (wt%) nominal concentration of element X in weld metal (wt%) concentration of element X in base plate (wt%)
vol% I scanning transmission electron microscope vol% X2 time (s) to [%Ti]sol. time constant (s) acid soluble titanium in weld metal (wt%) difference between total and acid .soluble Ti in weld metal (wt%) temperature (OC or K) ambient temperature (OC or K) preheating temperature (OC orK) reference temperatures (OC orK) rising velocity of ascending particles relative to the liquid (urn s-I) bulk velocity of the fluid relative to the particles (m S-I) voltage (V) welding (travel) speed (mm s') volume of melt (mm") molar volume of nucleus (m3 mor ') [%X]plate molar volume of oxide per mole of the diffusate (m3 mol ") [%X]incl. vol% YX2
WDX
(Ll%Ti)weld
WFR
T T0
xm [%X]
T* 0
TI, T2
[~%X]
[%X] eq.
u'
[%X]i
V Vm
[%X]nom.
V' m
219
thickness of stagnant gaseous boundary layer (mm) wetting angle
[%X]wire
concentration of element X in filler wire (wt%) arbitrary element or gaseous species density (kg g mm=')
e
A.
v
p Llp
m' or
mean particle centre to centre volume spacing (urn) dimensionless y-coordinate at maximum width of isotherm activity coefficient for MnO in slag phase activity coefficient for Si02 in slag phase
difference in density between liquid steel and inclusion (kg m-3 or g mm=') slag/metal interfacial energy (J m-2)
"'m
'YMnO
1")
'YSiOz
f.L
where 10 is a constant (with the unit nuclei per m3 and s) and ~G* is the energy barrier for nucleation. By rearranging equation (A2-1) and inserting reasonable values for 1 and 10 for the specific case of homogeneous nucleation of oxide inclusions in liquid steel, 55 we obtain:
(A2-2)
(A2-3) where NA is the Avogadro constant, 0' is the interfacial energy between the nucleus and the liquid (in J m-2) and ~Gv is the driving force for the precipitation reaction (in J nr-').
220
(A2-4)
where IlHo and IlSo are the standard entalpy and entropy of the precipitation reaction, respectively and Vm is the molar volume of the nucleus (in m3 mol:'). It is evident from Fig. 2.48 that IlGv = 0 when T= T1, which gives IlSo = IlHo/ Ti. Hence, equation (A2-4) may be rewritten as: I1G '\) 1li!0 = -V
m
[ 1- ---2 T
1
T]
(A2-5)
By combining equations (A2-2), (A2-3) and (A2-5), we obtain the following relationship between r. and T2:
(A2- 6)
3.1 Introduction
Inherent to the welding process is the formation of a pool of molten metal directly below the heat source. The shape of this molten pool is influenced by the flow of both heat and metal, with melting occurring ahead of the heat source and solidification behind it. The heat input determines the volume of molten metal and, hence, dilution 'and weld metal composition, as well as the thermal conditions under which solidification takes place. Also important to solidification is the crystal growth rate, which is geometrically related to weld travel speed and weld pool shape. Hence, weld pool shape, weld metal composition, cooling rate, and growth rate are all factors interrelated to heat input which will affect the solidification microstructure. Some important points regarding interpretation of weld metal microstructure in terms of these four factors will be discussed below. Since the properties and integrity of the weld metal depend on the solidification microstructure, a verified quantitative understanding of the weld pool solidification behaviour is essential. At present, our knowledge of the chemical and physical reactions occurring during solidification of fusion welds is limited. This situation arises mainly from a complex sequence of reactions caused by the interplay between a number of variables which cannot readily be accounted for in a mathematical simulation of the process. Nevertheless, the present treatment will show that it is possible to rationalise the development of the weld metal solidification microstructure with models based on well established concepts from casting and homogenising treatment of metals and alloys.
222
Chill zone
zone zone
Fig. 3.1. Transverse section of an ingot showing the chill zone, the columnar zone and the equiaxed zone (schematic) .
grains will change continuously from the fusion line towards the centre of the weld due to a corresponding shift in the direction of the maximum temperature gradient in the weld pool. This change in orientation may result in a curvature of the columnar grains (Fig. 3.2(a)). Alternatively, new grains can nucleate and grow in a columnar manner, producing a so-called 'stray' structure as shown schematically in Fig. 3.2(b). Finally, if the conditions for nucleation of new grains are favourable, an equiaxed zone will form near the weld centreline similar to that observed in ingots or castings (see Fig. 3.2(b)). Although the process of weld pool solidification is frequently compared with that of an ingot in 'miniature', a number of basic differences, already mentioned, exist which strongly influence the microstructure and properties of the weld metal. Of particular importance is also the disparity in cooling rate between a fusion weld and an ingot (see Fig. 3.3). For conventional processes such as shielded metal arc (SMA), gas metal arc (GMA), submerged arc (SA) or gas tungsten arc (GTA) welding the cooling rate may vary from 10 to 103oC s", while for modem high energy beam processes such as electron beam (EB) and laser welding the cooling 4 Consequentl y, to appreciate fully the implirate is typically of the order of 103 to 106oC s'. cations of these differences in general solidification behaviour between a weld pool and an ingot, it is necessary to consider in detail the sequence of events taking place in the solidifying weld metal beginning with the initiation of crystal growth at the fusion boundary.
223
(a)
(b)
Columnar zone --. Equiaxed zone Columnar zone -..
grains,
10
7 6 5
10
t
0 ,~
10 10 10
C/)
3 2
N
C)
cD
10 10
15
c: '=0
C)
g>
0 C,)
(5
c: .s::::
0 .m 0 ;
(i5 3:
c:
10 10.1
~~
..~
~(!l
.032
EO) C'CSc:
0)'03: ~ ~
co
"0 -c '0..
C/)
cO)
s:~
CI)(!J
~~
om mj
Process ~
c:
co
Fig. 3.3. Disparity in cooling conditions between casting, welding and rapid solidification.
224
t=~
Fig. 3.4. Schematic illustration showing epitaxial growth of columnar grains from partly melted base metal grains at fusion boundary.
,
,
,
,,
,
Substrate (8)
,
,
1
1 /
1
/1
r~,
,
\ \
,,
,
/ / /
,
, , V 1
I
/ I
of heterogeneous
nucleation.
225
3.3.1 Energy barrier to nucleation During epitaxial solidification, a solid embryo (nucleus) of the weld metal forms at the meltedback surface of the base metal grain. Assuming that the interfacial energy between the embryo and the liquid is isotropic, it can be shown, for a given volume of the embryo, that the interfacial energy of the whole system is minimised if the embryo has the shape of a spherical cap. Under such conditions, the following relationship exists between the interfacial energies (see Fig. 3.5): (3-1) where f3is the wetting angle. The change in free energy, configuration is given by:5
~Ghet'
(3-2)
_ 43 - f(p)A[ -31trs~Gv
+ 41trs 'Y EL
2]
where V E is the volume of the solid embryo, ~Gv is the free energy change associated with the embryo formation.Ag. andAEsare the areas of the embryo-liquid and embryo-substrate interfaces, respectively, and.f(f3) is the so-called shape factor, defined as: f(~)
= (2 + cos ~)(14
cos ~)2
(3-3)
The critical radius of the stable nucleus, rs *, is found by differentiating equation (3-2) with respect to rs and equating to zero: * 2'YEL r =--s
~Gv
(3-4)
By substituting equation (3-4) into equation (3-2), we obtain the following expression for the energy barrier to heterogeneous nucleation ( 8G~et):
where dB m is the latent heat of melting, T m is the melting point, and ~T is the undercooling. It is easy to verify that the first term in equation (3-5) is equal to the energy barrier to homogeneous nucleation, 8G;om. Hence, we may write: (3-6) Equation (3-6) shows that 8G~et is a simple function of the wetting angle (B). Since the
226
chemical composition and the crystal structure of the two solid phases are usually very similar, we have:"
y SL
Y EL and Y ES
Under such conditions equation (3-1) predicts that the wetting angle J3 ~ 0 (cosJ3 ~ 1), which implies that there is a negligible energy barrier to solidification of the weld metal ( dGZet ~ 0), i.e. no undercooling of the melt is needed, and solidification occurs uniformly over the whole grain of the base metal. This is in sharp contrast to conventional casting of metals and alloys where some undercooling of the melt is always required to overcome the inherent energy barrier to solidification (see Fig. 3.6). 3.3.2 Implications of epitaxial solidification Since the initial size of the weld metal columnar grains is inherited directly from the grain growth zone adjacent to the fusion boundary, the solidification microstructure depends on the grain coarsening behaviour of the base material. This is particularly a problem in high energy processes such as submerged arc and gas metal arc welding, where grain growth of the base metal can be considerable. In such cases the size of the columnar grains at the fusion boundary will be correspondingly coarse, as indicated by the data in Fig. 3.7. Moreover, during multipass welding the columnar grains can renucleate at the boundary between for instance the first and the second weld pass and subsequently grow across the entire fusion zone, as illustrated in Fig. 3.8. This type of behaviour is usually observed in weldments which do not undergo transformations in the solid state (e.g. aluminium, certain titanium alloys, stainless steel etc.). In practice, the problem can be eliminated by additions of inoculants via the filler wire, which facilitates a refinement of the columnar grain structure through heterogeneous nucleation of new (equiaxed) grains ahead of the advancing interface (to be discussed later).
.,,-
" ,,
\ \ \ \ \
Il~*
nom
Fig. 3.6. The free energy change associated with heterogeneous nucleation during casting and weld metal solidification, respectively (schematic). The corresponding free energy change associated with homogeneous nucleation is indicated by the broken curve in the graph.
227
140
E
~
Q)
N
120 100 80 60
40 20 GMAW (low-alloy steel)
-'w Sc Q)'-
Ee "CO>
-(1)
~c:Q)
Q):=:
c
Co
1ii :J co '0
20
40
60
80
100
120
140
HAZ prior austenite grain size (urn) ----Fig. 3.7. Correlation between HAZ prior austenite grain size at the fusion boundary and the corresponding weld metal prior austenite grain size. Data from Grong et al. 7
2. pass
1. pass HAZ
Base metal
Fig. 3.S. Optical micrograph showing renucleation of columnar grains during multipass GMA welding of a J3-titanium alloy.
228
qov
(3-7)
where q 0 is the net arc power, v is the welding speed, a is the thermal diffusi vity of the base plate, and Hm-Ho is the heat content per unit volume at the melting point. As shown in Fig. 3.9(a), a tear-shaped weld pool is favoured by a high n3 value, which is characteristic of fast moving high power sources. In contrast, at a low arc power and a low welding speed the shape of the weld pool becomes more elliptical because of a shift in the mode of heat flow (see Fig. 3.9(b). Note, however, that the thermal properties of the base metal is also of importance in this respect, since the n3 parameter is a function of both a and Hm-Ho. Consequently, a tear-shaped weld pool is usually observed in weldments of a low thermal diffusivity (e.g. austenitic stainless steel), whereas an elliptical or spherical weld pool is more likely to form during aluminium welding owing to the resulting higher thermal diffusivity of the base metal. In addition to the factors mentioned above, the geometry of the weld pool is also affected by convectional heat transfer due to the presence of buoyancy, electromagnetic or suface tension gradient forces. Recently, attempts have been made to include such effects in heat flow models for welding.r"! Referring to Fig. 3.10(a) the buoyancy force will promote the formation of a shallow, wide weld pool because of transport of 'hot' metal to the surface and 'cold' metal to the bottom of the pool. In the presence of the electromagnetic force the flow pattern is reversed, since the latter force will tend to push the liquid metal in the central part of the pool downward to the root of the weld. This makes the weld pool deeper and more narrow, as shown in Fig. 3.10(b). Moreover, it is generally accepted that surface tension gradients can promote circulation of liquid metal within the weld pool from the region of low surface tension to the region of higher surface tension." In the absence of surface active elements such as oxygen and sulphur, the surface tension decreases with increasing temperature as illustrated in Fig. 3.1O(c), which forces the metal to flow outwards towards the fusion boundary. This results in the formation of a relatively wide and shallow weld pool. However, if oxygen or sulphur is present in sufficient quantities a positive temperature coefficient of the surface tension may develop, which facilitates an inward fluid flow pattern and an increased weld penetration (see Fig. 3.10(d)). The important influence of surface active elements on the resulting bead morphology is well docu-
229
(b) Fig. 3.9. Theoretical shape of fusion boundary and neighbouring isotherms under different operational conditions; (a) High n3 values, (b) Low n3-values.
mented for ordinary GTA austenitic stainless steel welds.12,13 The indications are that such effects become even more important under hyperbaric welding conditions.!" 3.4.2 Columnar grain morphology It is evident from the above discussion that a change in the weld pool geometry, caused by variations in the operational conditions, may strongly alter the weld metal solidification microstructure. In fact, more than nine different grain morphologies have been observed during fusion welding. 15 The two most important are shown in Fig. 3.11. Referring to Fig. 3.11(a) a spherical or elliptical weld pool will reveal curved and tapered columnar grains owing to a shift in the direction of the maximum thermal gradient in the liquid from the fusion boundary towards the weld centre-line. In contrast, a tear-shaped weld pool yields straight and broad
230
(a)
(b)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool; (a) Buoyancy force (b) Electromagnetic force.
columnar grains as shown in Fig. 3.11(b), since the direction of the maximum temperature gradient in the melt does not change significantly during the solidification process. The latter condition is known to promote formation of centre-line cracking because of mechanical entrapment of inclusions and enrichment of eutectic liquid at the trailing edge of the weld pool. 3.4.3 Growth rate of columnar grains The growth rate of the columnar grains is geometrically related to the weld travel speed and the weld pool shape. 3.4.3.1 Nominal crystal growth rate Since the shape of the weld pool remains constant during steady state welding, the growth rate of the columnar grains must vary with position along the fusion boundary. This point is more clearly illustrated in Fig. 3.12 which shows a sketch of a single columnar grain growing parallel with the steepest temperature gradient in the weld pool. Taking the angle between the
231
"00 c:
o
Q)
c:
o cu 't:
en
Temperature
::::s
Weld pool
(c)
o "00
c
Q)
o
't: ::::s
cu
en
Temperature
Weld pool
(d)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool (continued); (c) Surface tension gradient force (negative gradient); (d) Surface tension gradient force (positive gradient).
232
(a)
(b)
Fig. 3.11. Schematic comparison of columnar grain structures obtained under different welding conditions; (a) Elliptical weld pool (low n3 values), (b) Tear-shaped weld pool (high n3 values). Open arrows indicate the direction of the maximum temperature gradient in the weld pool.
Fusion boundary
Crystal
Fig. 3.12. Definition of the nominal crystal growth rate RN.
growth direction and the welding direction equal to a, the steady state growth rate, RN, becomes:
RN = v cos
o;
(3-8)
where v is the welding speed. Considering spherical or elliptical weld pools, the nominal crystal growth rate is lowest at the edge of the weld pool (a~90, coso-e-O) and highest at the weld centre-line where RN approaches v (a~O, cosa~ 1). In contrast, columnar grains trailing behind a tear-shaped weld pool will grow at an approximately constant rate which is significantly lower than the actual welding speed (a 0), since the direction of the maximum temperature gradient in the weld pool does not change during the solidification process. This is also in agreement with practical experience (see Fig. 3.13).
233
(a)
t
E
E
C)
16
12 10 8 6
e .c
~.,;..
(ij
.e
E
z
0
c:
20
40
60
80
100
(b)
t
'0)
c:
16 14
E E 12 10 8
6
Z a:
$ .c
~ ~ ~
~ J:::::::::~
ov =1.7mm/s
OO/C
1 O~%
G2?10~% ,
ev = 5.3 mm/s 00lc
IQ)
Ie:
10 IN 1"0 IQ)
(ij
'E 0
z
2
RN (max)
= 2.7 mm/s
1)(
ICU
1'5
IUJ
10'"
20
40
60
Fig. 3.13. Measured crystal growth rates in thin sheet electron beam welding; (a) Niobium, (b) Stainless steel. Data from Senda et ale 16
234
Example (3.1)
Consider electron beam (EB) welding of a 1mm thin sheet of austenitic stainless steel under the following conditions: qo
590W, v
5.3mm s -1, To
200e
Estimate on the basis of the Rosenthal thin plate solution (equation 1-83) the steady state growth rate of the columnar grains trailing the weld pool.
Solution
The contour of the fusion boundary can be calculated from the Rosenthal thin plate solution according to the procedure shown in Example (1.10). If we include a correction for the latent heat of melting, the 681n3 ratio at the melting point becomes: 8mo
n3
= 21tad(Hm
qo
- Ho)
Substitution of the above value into equation (1-83) gives the fusion boundary contour shown in Fig. 3.14. It is evident from Fig. 3.14 that the weld pool is very elongated under the prevailing circumstances due to a constrained heat flow in the x-direction. This implies that the angle <X will not change significantly during the solidification process. Taking <X as an average, equal to about 70, the steady-state crystal growth rate RN becomes: RN
= v cos a =
5.3 cos(700)mm s -1
1.8mm
S-1
This value is in reasonable agreement with the measured crystal growth rates in Fig. 3.13(b). 3.4.3.2 Local crystal growth rate Equation (3-8) does not take into account the inherent anisotropy of crystal growth. For faceted materials the dendrite growth directions are always those that are 'capped' by relatively slow-growing (usually low-index) crystallographic planes.' Figure 3.15 shows examples of faceted cubic crystals delimited by {100} and {Ill} planes, respectively. If the {Ill} planes are the slowest growing ones, the {I OO}planes will grow out, leaving the {Ill} facets and a new crystal growing in the <100> directions as shown schematically in Fig. 3.15(b). Although most metals and alloys do not foim faceted dendrites, the anisotropy of crystal growth is still maintained during solidification.? In fact, experience has shown that the major dendrite growth direction is normally the axis of a pyramid whose sides are the most closely packed planes with which a pyramid can be formed. 1 These directions are thus < 100> for body- and face-centred cubic structures, < 1010 > for hexagonal close-packed structures, and < 110> for body-centred tetragonal structures. Because of the existence of preferred growth directions, the local growth rate of the crystals RL will always be higher than the nominal growth rate RN defined in equation (3-8). Consider now a cubic crystal which grows along the steepest temperature gradient in the weld pool, as shown schematically in Fig. 3.16. If ~ denotes the angle between the interface normal and the < 100> direction, the following relationship exists between R Nand R L:
235
E
~
E
Columnar zone
-5 Columnar zone
Fusion boundary
Fig. 3.14. Predicted shape of fusion boundary during electron beam welding of austenitic stainless steel (Example (3.1.
(a)
(b)
Fig. 3.15.Examples of faceted cubic crystals; (a) Crystal delimited by {100} planes, (b) Crystal delimited by {Ill} planes.
236
1491~~--~~~--~~--~~----~~~
U o
1489
Fe - 15 Ni - 15 Cr GL=105 KIm
1483 1481~~--~~~~--~~~~--~~
4 6 10 20 40 60 100
Fig. 3.17. Calculated dendrite tip temperature vs dendrite growth velocity for an Fe-15Ni-15Cr alloy. The undercooling of the dendrite tip is given by the difference between the liquidus temperature and the solid curve in the graph. Data from Rappaz et al. 17
RN which gives:
= RL cos <I> = v
R = v cosu L cos <I>
COSU
(3-9) (3-10)
Equation (3-10) shows that the local growth rate increases with increasing misalignment of
237
the crystal with respect to the direction of the maximum temperature gradient in the weld pool. Since such crystals cannot advance without a corresponding increase in the undercooling ahead of the solidlliquid interface (see Fig. 3.17), they will soon be outgrowed by other grains which have a more favourable orientation. Fusion welds of the fcc and bee type will therefore develop a sharp <100> solidification texture in the columnar grain region, similar to that documented for ingots and castings. The weld metal columnar grains may nevertheless be separated by 'high-angle' boundaries, as shown in Fig. 3.1S, due to a possible rotation of the grains in the plane perpendicular to their < 100> length axes.
Example (3.2)
Consider electron beam welding of a 2mm thick single crystal disk of Fe-15Ni-15Cr the following conditions:
under
The orientation of the disk with respect to the beam travel direction is shown in Fig. 3.19. Calculate on the basis of the minimum velocity (undercooling) criterion the growth rate of the dendrites trailing the weld pool under steady state welding conditions (assume 2-D heat flow). Make also schematic drawings of the solidification microstructure in different sections of the weld. Relevant thermal properties for the Fe-15Ni-15Cr single crystal are given below:
Solution
Since the base metal is a single crystal, separate columnar grains will not develop. Nevertheless, under 2-D heat flow conditions growth of the dendrites can occur both in the [100] and the [010] (alternatively the [010]) direction. Referring to Fig. 3.20 the growth rate of the [100] and the [010] dendrites is given by: R100 = V and
v coso.
RoIO = cos(900-
a)
238
[100]
Fig. 3.20. Schematic diagram showing the pertinent orientation relations between the fusion boundary interface normal and the dendrite growth directions (Example (3.2.
From this it is seen that the velocity of the [100] dendrites is always equal to that of the heat source v. In contrast, the growth rate of the [010] dendrites depends both on v and o, and will therefore vary with position along the fusion boundary. It follows from minimum velocity criterion that the [100] dendrites will be selected when the interface normal angle u is less than 45, while the [010] dendrites will develop at larger angles. This is shown graphically in Fig. 3.21. At pseudo-steady state the fusion boundary can be calculated from the Rosenthal thin plate solution (equation (1-83 according to the procedure shown in Example (1.10). If we include
239
2.0
[010]
dendrites
> ......
a:
~ 1.0
0.5
20
40 ex,
degrees
60
80
Fig. 3.21. Normalised minimum dendrite tip velocity vs interface normal angle ex (Example 3.2.
a correction for the latent heat of melting, the 88/n3 ratio at the melting point becomes:
Substitution of this value into equation (1-83) gives the fusion boundary contour shown in Fig.3.22(a). Included in Fig. 3.22 are also schematic drawings of the predicted solidification microstructure in different sections of the weld. The results in Fig. 3.22 should be compared with the reconstructed 3-D image of the solidification microstructure in Fig. 3.23, taken form Rappaz et al.17 Due to partial heat flow in the z-direction, [001] dendrite trunks will also develop. Nevertheless, these data confirm the general validity of equations (3-8) and (3-10) relating crystal growth rate to welding speed and weld pool shape. 3.4.4 Reorientation of columnar grains In principle, there are two different ways a columnar grain can adjust its orientation during solidification in order to accommodate a shift in the direction of the maximum temperature gradient in the weld pool, i.e.: (i) (ii) Through bowing Through renucleation.
240 (a)
[010]
dendrites
dendrites
[100]
+x (mm)
[010]
(b)
t4 ----
r--- 2.1
mm --.....{
Base plate
[010]
dendrites
[100]
dendrites
[010]
dendrites microstructure (Example 3.2;
Fig. 3.22. Schematic representation of the weld metal solidification (a) Top view of fusion zone, (b) Transverse section of fusion zone.
3.4.4.1 Bowing of crystals A continuous change in the crystal orientation due to bowing will result in curved columnar grains, as shown previously in Fig. 3.2(a). This type of grain morphology has been observed in for instance electron beam welded aluminium and iridium alloys+" Normally, the adjustment of the crystal orientation is promoted by multiple branching of dendrites present within the grains. Alternatively, the reorientation can be accommodated by the presence of surface defects at the solid/liquid interface, e.g. screw dislocations, twin boundaries, rotation boundaries, etc. The latter process presumes, however, a faceted growth morphology, and is therefore of minor interest in the present context.
Example (3.3)
Consider a curved columnar grain of iridium which grows from the fusion boundary towards the weld centre-line along a circle segment of length L, as shown schematically in Fig. 3.24. Based on the assumption that the bowing is accommodated solely by branching of [010] dendrites in the [100] direction, calculate the maximum local growth rate of the crystal during solidification.
241
~Ol ~
z
[010]
001 ]
Fig. 3.23. Reconstructed 3-D image of solidification microstructure in an electron beam welded Fe-15Ni15Cr single crystal. The letters (a), (b) and (c) refer to [100], [010] and [001] type of dendrites, respectively. After Rappaz et al.1?
8~./P
0/
~
-$
r:::-
,_._._.
d?
Fusion line - -
Solution
In principle, the solution to this problem is identical to that presented in Example (3.2). Referring to Fig. 3.24 the growth rate of [100] and the [010] dendrite stems is given by: R100
242 and
R
010
=v
cos cos(900
a
-
a)
It follows from Fig. 3.21 that growth will occur preferentially in the [010] direction as long as the interface normal angle a. is larger than 45, while the [100] direction is selected at smaller angles. This means that the local growth rate of the dendrites, in practice, never will exceed the welding speed v. 3.4.4.2 Renucleation of crystals In ingots and castings, three different mechanisms for nucleation of new grains ahead of the advancing interface are operative: 1,2 (i) (ii) (iii) Heterogeneous nucleation Dendrite fragmentation Grain detachment.
The former mechanism is of particular importance in welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. These particles can either be primary products of the weld metal deoxidation or stem from reactions between specific alloying elements which are deliberately introduced into the weld pool through the filler wire. The latter process is also known as inoculation. Nucleation potency of second phase particles In general, the effectiveness of individual particles to act as heterogeneous nucleation sites can be evaluated from a balance of interfacial energies, analogous to that described in Section 3.3.1 for epitaxial nucleation. It follows from the definition of the wetting angle ~ in Fig. 3.5 that the energy barrier to heterogeneous nucleation is a function of both the substrate/liquid interfacial energy 'Y SL, the substrate/embryo interfacial energy 'YES' and the embryo/liquid interfacial energy "IEL. Complete wetting is achieved when:
'Y SL
;;::: ES
'Y
+ 'Y EL
(3-11)
Under such conditions, the nucleus will readily grow from the liquid on the substrate.Unfortunately, data for interfacial energies are scarce and unreliable, which makes predictions based on equation (3-11) rather fortuitous. 18 In pure metals, experience has shown that the solid/liquid interfacial energies are roughly proportional to the melting point, as shown by the data in Fig. 3.25. On this basis, it can be expected that the higher melting point phases will reveal the highest 'Y SL values, and thus be nucleants for lower melting phases. A similar situation also exists in the case of non-metallic inclusions in liquid steel, where the high-melting point phases are seen to exhibit the highest solid/liquid interfacial energies (see Fig. 3.26). In contrast, very little information is available on the substrate/embryo interfacial energy "IES. For fully incoherent interfaces, 'YES would be expected to be of the order of 0.5 to 1 J m-25 However, this value will be greatly reduced if there is epitaxy between the inclusions and the nucleus, which results in a low lattice disregistry between the two phases. In general, assessment of the degree of atomic misfit between the nucleus n and the substrate s can be done on
243
t
~
0
ew
3200 '-
eTa eos
.Mo
eRe
E
CO>
2400~
Q5
:2
I I I
2---...
Fig. 3.25. Values of solid/liquid interfacial energy ~ SL of various metals as function of their melting points. Data from Mondolfo.!"
3000
Fig. 3.26. Values of interfacial energy ~SL for different types of non-metallic inclusions in liquid steel at 1600C as function of their melting points. Data compiled from miscellaneous sources.
244
METALLURGICAL
MODELLING
OF WELDING 19
~(hkl)s U(hkl)
n
_ -
L 1[ I(d
3 ;=1
[uvw]~
x 100%
(3-12)
where
(hkl)s
[uvw]s
a low-index plane of the substrate; a low-index direction in (hki), a low-index plane in the nucleated solid; a low-index direction in (hkl)n; the interatomic spacing along [uvw]n; = the interatomic spacing along [uvw]s; and = the angle between the [uvw]s and the [uvw]n.
(hkl)n
[uvw]n
d[uvw]n d[uvw]s
In practice, the undercooling ~T (which is a measure of the energy barrier to heterogeneous nucleation) increases monotonically with increasing values of the planar lattice disregistry, as shown by the data in Fig. 3.27. This means that the most potent catalyst particles are those which also provide a good epitaxial fit between the substrate and the embryo. Examples of such catalyst particles are TiAl3 in aluminium 18 and TiN in steel. 19 Nucleation of delta ferrite at titanium nitride will be considered below.
1000
CaS
SCaD ASi02
OAI2D OTiN
t
o o
o
(1)
-MnO
~TiC
CaS
eRE203 100
vwc
.5 10 (5
c;
~
=>
'0 C
0.1
0.01
12
16
20
24
28
Fig. 3.27. Relationship between planar lattice disregistry andundercooling steel. Data compiled from miscellaneous sources.
245
Example (3.4)
In low-alloy steel weld metals, titanium nitride can form in the melt due to interactions between dissolved titanium and nitrogen. Assume that the TiN particles are faceted and delimited by {lOa} planes. Calculate on the basis of equation (3-12) the minimum planar lattice disregistry between TiN and the nucleating delta-ferrite phase under the prevailing circumstances. Indicate also the plausible orientation relationship between the two phases. The lattice parameters of delta ferrite and TiN at 1520C may be taken equal to 0.293 and 0.431nm, respecti vel y.
Solution
Titanium nitride has the N aCI crystal structure, while delta ferrite is body-centred cubic, as shown in Fig. 3.28(a) and (b). It is evident from Fig. 3.29(a) that a straight cube-to-cube orientation relationship between TiN and 8-Fe will not result in a small lattice disregistry, However, the situation is largely improved if the two phases are rotated 45 with respect to each other (see Fig. 3.29(b, conforming to the following orientation relationship: (100)o-Fe II (100)TiN [110lo-Fe II [1OOlTiN The resulting crystallographic relationship at the interface is shown schematically in Fig. 3.29(c). Since the lattice arrangements are similar in this case, equation (3-12) reduces to:
B(lOO)
(100)
A comparison with the data in Fig. 3.27 shows that the calculated lattice disregistry conforms to an undercooling of about 1 to 2C. This value is sufficiently small to facilitate heterogeneous nucleation of new grains ahead of the advancing interface during solidification. Considering other inclusions with more complex crystal structures, the chances of obtaining a small planar lattice disregistry between the substrate and the delta ferrite nucleus are
(a)
(b)
Fig. 3.28. Crystal structures of phases considered in Example (3.4); (a) Titanium nitride, (b) Delta ferrite.
246
(100)S_Fe II
(100)TiN
[1DOJo_FeII [ 10D]TiN
I
TiN
I I
u-Fe
u-
TIN TIN
[11 O]~F
II [ 100] .
(c)
/---(b)
I I I I
TiN
(:'
Ti atoms
N atoms
o-Featoms
Fig. 3.29. Possible crystallographic relationships between titanium nitride and delta ferrite (Example (3.4; (a) Straight cube-to-cube orientation, (b) Twisted cube-to-cube orientation, (c) Details of lattice arrangement along coherent TiN/d-Pe interface.
rather poor (see Fig. 3.27). Nevertheless, such particles can act as favourable sites for heterogeneous nucleation if 'Y SL is sufficiently large compared with 'Y EL and 'YES' This is illustrated by the following example:
Example (3.5)
In low-alloy steel weld metals Al203 inclusions can form during the primary deoxidation stage as discussed in Section 2.12.4.2 (Chapter 2). Based on the classic theory of heterogeneous nucleation, evaluate the nucleation potency of such inclusions with respect delta ferrite.
"1-
247
Solution
It is readily seen from Fig. 3.27 that the planar lattice disregistry between delta ferrite and Al203 is very large, which indicates of a fully incoherent interface (i.e. 'YES = 0.75 J m"). Moreover,readings from Figs. 3.25 and 3.26 give the following average values for the delta ferrite/liquid and the inclusion/liquid interfacial energies:
'YEL
= 0.35 J m?
and
According to equation (3-11) complete wetting is achieved when 'Y SL ~ 'YES + 'Y EL" This requirement is clearly met under the prevailing circumstances. Similar calculations can also be performed for other types of non-metallic inclusions in steel weld metals. The results are presented graphically in Fig. 3.30. It is evident that the nucleation potency of the inclusions increases in the order Si02-MnO, A1203- Ti203-Si02MnO, A1203, reflecting a corresponding increase in the inclusion/liquid interfacial energy 'Y SL" The resulting change in the weld metal solidification microstructure is shown in Fig. 3.31, from which it is seen that both the average width and length of the columnar grains decrease with increasing Al203-contents in the inclusions. This observation is not surprising, considering the characteristic high solid/liquid interfacial energy between aluminium oxide and steel (see Fig. 3.26). The important effect of deoxidation practice on the weld metal solidification microstructure is well documented in the literature. 3,20-22 Rate of heterogeneous nucleation It can be inferred from the classic theory of heterogeneous nucleation that the nucleation rate
--~ (degrees)
1.0
180
120
90
60
t
~
0.8
E
0
"'.!:
0.6
AI203 0.4
'Q) (!J
..c:
<1
0.2 0
-1.6
-1.2
-0.8
-0.4
0.4
-----
0.8
1.2
1.6
(Y sC YEP)/Y EL
Fig. 3.30. Nucleation potency of different weld metal oxide inclusions with respect to delta ferrite.
248 (a)
(wt%) ----80
t
E
20
160
I95%
confidence
limit
.~ 120
C>
en c:
::l
..c
'0
i5
Q)
'
80
~ ~
C>
Q)
40
o '.
0
0.4
0.8
1.2
1.6
2.0
2.4
(b)
t
"(ij
o
500
20
40
60
80
100
~------------------~
95% confidence limit
Pure AI20a
E 400
en s:::
::t
C,
300
.co
'0
~ ~
g>
Q)
200
C>
~ 100
0.4
0.8
1.2
1.6
2.0
2.4
Fig. 3.31. Effect of deoxidation practice on the columnar grain structure in low-alloy steel weld metals; (a) Average width of columnar grains, (b) Average length of columnar grains. Data from Kluken et al.22
Nhet(i.e. number of nuclei which form per unit time and unit volume of the melt) is interrelated to the energy barrier f1GZet through the following equation."
Nhet -
I1N vexp
(I1GD
+I1GZet)] kT
(3-13)
where 11 is a frequency factor, N; is the density of nucleation sites per unit volume of the melt,
249
I1GD is the activation energy for diffusion of atoms across the interface, and k is the Boltzmann constant. Since I1GD is often negligible compared with ~GZet in liquids, equation (3-13) reduces to:
(3-14) Equation (3-14) shows that the nucleation rate depends both on N; and f1Ghet Hence, under full wetting conditions (f1Ghet~ 0), the number of nuclei which form per second and mm ' ahead of the advancing interface is directly proportional to the instantaneous concentration of catalyst particles in the melt. Examples of such particles are TiAl3 in aluminium and TiN/Al203 in steel. The important effect of controlled titanium additions and subsequent TiAl3 precipitation on the columnar grain structure in 1100 aluminium welds is illustrated in Fig. 3.32.
s:
Critical cell/dendrite alignment angle It follows from the minimum growth rate criterion and the definition of the local crystal growth rate in equation (3-10) that reorientation of the columnar grains will occur when the cellI dendrite alignment angle <f> reaches a critical value <f> *. The value of <f> * will, in turn, depend on the nucleation potency of the catalyst particles and can be estimated for different types of welds. If growth of the columnar grains is assumed to occur along a circle segment of length L (see Fig. 3.33), the critical cell/dendrite alignment angle is given by:
th*
't'
=(0-
I L
(3-15)
where w is the total grain rotation angle, and I is the average length of the columnar grains (in mm).
t
::t uj
500
[J
<> Y = 0%
L:J..
E 400
o Y= 50%
[J
Y=25%
Y = 750/0
.~ 300
0)
s=
15
.c:
Q)
:c
e
.~ 200
C)
-c 100
Fig. 3.32. Effect of titanium on the columnar grain structure in 1100 aluminium welds. The value Yis the fractional distance from fusion line to top surface of weld metal. Data from Yunjia et a1.23
250
Fig. 3.33. Characteristic growth pattern of columnar grains in bead-on-plate welds (schematic).
By introducing reasonable average values for wand I in the case of SA welding of lowalloy steel,22 we obtain:
<P * ~ 60 deg( ~ )
(3-16)
Calculated values for <P * in steel weld metals are presented in Fig. 3.34, using data from Kluken et al. 22 An expected, the critical cell/dendrite alignment angle in fully aluminium deoxidised steel welds is seen to be very small (of the order of 2), reflecting the fact that nucleation of delta ferrite occurs readily at Al203 inclusions. The value of <p * increases gradually with decreasing Al203 contents in the inclusions and reaches a maximum of about 4 for Si-Mn deoxidised steel weld metals. This situation can be attributed to less favourable nucleating opportunities for delta ferrite at silica-containing inclusions, which reduces the possibilities of obtaining a change in the crystal orientation during solidification through a nucleation and growth process. Dendrite fragmentation In principle, nucleation of new grains ahead of the advancing interface can also occur from random solid dendrite fragments contained in the weld pool. Although the source of these solid fragments has yet to be investigated, it is reasonable to assume that they are generated by some process of interface fragmentation due to thermal fluctuations in the melt or mechanical disturbances at the solidlliquid interface.' At present, it cannot be stated with certainty whether grain refinement by dendrite fragmentation is a significant process in fusion welding.i" Grain detachment Since the partially melted base metal grains at the fusion boundary are loosely held together by liquid films between them, there is also a possibility that some of these grains may detach themselves from the base metal and be trapped in the solidification front.I" Like dendrite fragments, such partially melted grains can act as seed crystals for the formation of new grains in the weld metal during solidification if they are able to survive sufficiently long in the melt.
251
t
o > .......
1.025----"1 1.020
~ ~ 1.015
o en ~
00
~
.-
0:-11.010
<N 0'"
F
C\I
en
1.0
aO
10
12
1l -
16
Fig. 3.34. Critical cell/dendrite alignment angle <f> * for reorientation of delta ferrite columnar grains during solidification of steel weld metals. Data from Kluken et al. 22
252
Fig. 3.36. Optical micrograph showing the characteristic cellular-dendritic substructure in a low-alloy steel weld. The metallographic section is normal to the columnar grain growth direction. After Kluken et al.22
Fig. 3.37. Schematic representation of the development of primary, secondary and tertiary dendrite arms along < 100> directions in cubic crystals.
253
A characteristic feature of cellular and cellular-dendritic growth is also that the boundary between two adjacent columnar grains will closely follow the contours of the original cell boundaries. An illustration of this point is contained in Fig. 3.38. Consequently, since all cell walls are preferential sites for segregation during solidification (see ion micrograph in Fig. 3.39), the presence of solute at the columnar grain boundaries can strongly alter the kinetics of subsequent solid state transformation reactions. The indications are that for instance phosphorus segregations at prior austenite grain boundaries will promote the formation of grain boundary ferrite in low-alloy steel weld metals during the 'Y to a transformation because of the associated increase in the Ae3-temperature.22,25 We will return to this question in Chapter 6.
(a)
(b)
Fig. 3.38. Optical micrographs showing alignment of austenite grain boundaries along original delta ferrite (cell) boundaries in a low-alloy steel weld; (a) Primary delta ferrite solidification microstructure, (b) Secondary austenite grain structure. After Kluken et al.22
254
Fig. 3.39.Ion (SIMS) micrograph showing evidence of phosphorus segregations at primary solidification (cell) boundaries in a low-alloy steel weld. After Kluken et al. 22
3.5.2 Stability of the solidification front The stability of the solid/liquid interface is critical in determining the microstructural characteristics of the weld metal. 3.5.2.1 Interface stability criterion The simplified treatment given here is based on the constitutional undercooling criterion which ignores the important effect of capillarity on the interface stability. Consider a simple binary system, the phase diagram of which is given in Fig. 3.40(a). During solidification, a solute-rich layer will form in front of the growing interface, as shown schematically in Fig. 3.40(b). At steady state, the mass of solute transferred from the liquid to the interface, R/CL)i' is equal to the mass of solute which accumulates in the solid, R/CS)i' minus the amout which diffuses back into the liquid, DL( dCz/dx)i. Hence, we may write:
(3-17) Provided that equilibrium exists at the solid/liquid interface, i.e. (CS)i = ko(CL)i' equation (3-17) can be rewritten as:
(deL) dx = __R. 1 (C ).(I-k
DL L
I 0
(3-18)
where DL is the diffusivity of the solute in the liquid, and ko is the equilibrium partition coefficient (ko < 1).
255
(a)
(CJ i
Concentration
(b)
c: o "Ci) o
o
;:;
o 0. E
=-:::::1
"0
::J
C" "5
Co
Distance (x)
(c)
Liquidus temperature (T L )
Distance (x) Fig. 3.40. Constitutional undercooling in alloy solidification; (a) Schematic representation of binary phase diagram, (b) Build-up of solute-enriched layer in front of solid/liquid interface, (c) Undercooled region ahead of solid/liquid interface"
256
With the aid of the phase diagram in Fig. 3.40(a) it is easy to verify that the equilibrium liquidus temperature increases with distance from the interface because of the lower solute content. Referring to Fig. 3.40(c) the latent heat of melting Mlm will diffuse away from the interface (thereby stabilising possible interface protuberances) if the actual temperature gradient in the liquid (dTA/dx\ is less than the equilibrium liquidus temperature gradient (dT idx)i' The latter gradient can be expressed as: ( dTL) dx
= (dTL)
i
dCL.
. (dCL )
l
(3-19)
i
dx
where (dTidCL)i denotes the slope of the liquidus curve in the phase diagram (designated By combining equations (3-18) and (3-19), we obtain the following criterion for the interface stability:
) G = dT - L L ( dx
=,
=-m
R. (C ).(l-k _l L DL L 1
)
0
(3-20)
Taking (CL)i = Co/ko and R, = RL this equation can alternatively be written as: GL RL
mLCo(l-ko) koDL
(3-21)
3.5.2.2 Factors affecting the interface stability It follows from equation (3-21) that the stability of the weld metal solidification front is controlled by the extent of constitutional undercooling ahead of the advancing interface, and is therefore influenced by factors such as the total amount of alloying or impurity elements present, Co' the local crystal growth rate, RL, and the thermal gradient in the weld pool, GL" The combined effects of alloy content, growth rate, and thermal gradient are presented schematically in Fig. 3.41. It can be seen from the figure that a planar solidification front is favoured by a high GiRL ratio, in agreement with predictions based on equation (3-21). At lower GL/RL ratios, the morphology of the interface changes to cellular, cellular-dendritic or dendritic, depending on the degree of constitutional undercooling ahead of the advancing interface. Normally, the GriRL ratio close to the fusion line is large enough to facilitate cellular solidification. The temperature gradient in the weld pool decreases, however, with distance from the fusion boundary and hence, a cellular-dendritic type of substructure is often observed in the central areas of the columnar grain region.V This substructure may change to dendritic close to the weld centre-line if the degree of constitutional undercooling in front of the advancing interface is sufficiently large. The various types of growth products that may develop during normal solidification of fusion welds are shown schematically in Fig. 3.42.
Example (3.6)
In aluminium welding, binary AI-Si and AI-Mg alloys are frequently used as filler metals. Consider stringer bead deposition with an AI-4.5wt%Si filler wire at a constant welding speed of 1.5mm S-I. Based on equation (3-21) calculate the critical (minimum) temperature gradient
257
G L fA
L ------
Fig. 3.41. Schematic representation of the combined effect of crystal growth rate RL and melt thermal gradient G L on the weld metal solidification microstructure.
Fusion zone
v
Small RL,
Large G.
Cellular-dendritic
~
HAZgrains
Fig. 3.42. Schematic diagram illustrating structure across the fusion zone.
micro-
in the weld pool which gives a planar solidification front. Relevant physical data for the AI-Si system are given below."
mL = -7.5C per wt% DL
S-1
Solution
When mL, ko and DL are known the critical temperature gradient (GL)cr. can readily be calculated from equation (3-21). Taking RL ::::: v, we obtain:
mLCo(l-ko)R
L-
mm
-1
258
METALLURGICAL
MODELLING
OF WELDING
The above calculations show that the temperature gradient in the weld pool must be extremely large in order to promote a planar solidification front. The calculated value for (GL)cr. corresponds to a cooling rate of about 92 400C s:' (C.R. = GLRL). Besides low-heat input electron beam and laser welding, such high cooling rates are rarely observed in fusion welding (see Fig. 3.3). On this basis it is not surprising to find that the weld metal solidification microstructure is normally of the cellular or the dendritic type. For a rough evaluation of the weld metal solidification microstructure the diagram in Fig. 3.43 can be used. This diagram summarises the various microstructures which can be obtained (using a typical alloy' with a melting range of 50C) when the imposed temperature gradient G L and crystal growth rate RL are varied. Moving from the lower left to the upper right along the lines at 45 leads to a refinement of the structure without changing the morphology (constant GLIRL ratio). Crossing these lines by passing from the upper left to the lower right leads to changes in the morphol~gy from planar to cellular or dendritic growth, while the scale of the microstructure remains essentially the same. The gray bands define the regions over which one structure changes into another. An example of the application of this diagram is given below.
Example (3.7)
Consider GTA butt welding of a 2mm thin sheet of aluminium (AI-Mg alloy) under the following conditions: I
II OA, U
15V, v
= 4mm
S-I,
1")
= 0.6,
To
= 20C
Based on the diagram in Fig. 3.43, estimate the weld metal solidification microstructure in different positions from the fusion boundary. In these calculations we shall assume that the actual thermal gradient in the weld pool is equal to the average thermal gradient within the temperature interval from 680 to 550C, as evaluated from the Rosenthal thin plate solution (equation (1-83)). Relevant thermal data for the AI-Mg alloy are given in Table 1.1 (Chapter
I). Solution
The contours of the fusion boundary and the 680 and the 550C isotherms can be calculated according to the procedure shown in Example (1.10). If we neglect the latent heat of melting, the corresponding eoln3 ratios at these temperatures become: T
Tm
650C:
8m8
n3
2nad(Hm -Ho) q0
88 "s
(650 - 20)
259
()
0
E E
t
~
10
101
~----~------~
10.2
RL,mm/s
10
~
Fig. 3.43. Variation of weld metal solidification microstructure with GLand RL" The diagram is based on the ideas of Kurz and Fisher.?
80
n3
= 1.19[(550-20)]
(650 - 20)
1.00
Substitution of the above values into equation (1-83) gives the isothermal contours shown in Fig. 3.44. It follows from Fig. 3.44 that the thermal gradient decreases from about 25e mm' at the fusion boundary to approximately 100e mm ! close to the weld centre-line. This occurs parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm S-I. According to Fig. 3.43 the calculated values of GL and RL conform to a cellular-dendritic solidification microstructure within the central regions of the weld and an equiaxed-dendritic microstructure close to the weld centre-line. Both types of substructures are commonly observed in aluminium weldments.Fr"
Weld pool (680C) Fusion boundary (650C) HAZ (550C)
-30
-20
-10
+10
+20
Fig. 3.44. Predicted shape of fusion boundary and neighbouring isotherms during thin plate aluminium welding (Example (3.6)).
260
3.5.3 Dendrite morphology Within the fully dendritic region of the weld, the dendrite morphology remains largely unchanged over a wide range of cooling rates. Nevertheless, it will become finer as the heat is extracted at greater rates. 3.5.3.1 Dendrite tip radius A combination of stability analysis with diffusion considerations and the specification of a dendrite tip shape and a radius selection criterion can lead to an unique solution for the morphology and solute distribution around a growing dendrite.29,30 Based on such analyses it is possible to calculate the dendrite tip radius rd as function of the growth rate R L and the thermal gradient G L in the weld pool for a given alloy system.!? Referring to Fig. 3.45, the dendrite growth is seen to occur between two stability limits. At low speeds, the rd vs RL curve reaches a threshold where the tip radius increases rapidly and finall y approaches an infinite value corresponding to a planar solidification front. At high speeds, an absolute stability .limit is also found. This corresponds to the limiting case where the transport of solute as;1"OSS the interface becomes too sluggish to keep pace with the advancing solidification front. However, within the central areas of the curve, the dendrite tip radius is seen to decrease steadily with increasing values of RL, typically as l/(RL) 112. Note that the thermalgradient in the weld pool GL will have a strong influence on the low growth rate stability limit. In contrast, the absolute stability limit is essentially independent of GL, as indicated in Fig. 3.45.
100
60
40
20
10
=c
ur ::J
E :i.
6 4
2
\.,rGL=
""
10 KIm
........
i=
0-
0.6
0.4
0.2 0.1 ----~ 0.1 ~~~ 0.2 0.4 __~ __ ~~~-L 2 4 6
Growth rate, mm/s
__ ~ __ ~~~~
10
20
40 60 100
Fig. 3.45. Predicted relationship between dendrite tip radius r d' crystal growth rate RL and melt thermal gradient GL for a Fe-15Cr-15Ni alloy. Data from David and Vitek.4
SOLIDIFICATION
BEHAVIOUR
OF FUSION
WELDS
261
3.5.3.2 Primary dendrite arm spacing The primary dendrite arm spacing (defined in Fig. 3.46) is an important characteristic of the solidification microstructure and has a marked effect on the weld metal mechanical properties. If it is assumed that the dendrite envelope, representing the mean cross-section of the trunk, can be described by an ellipse (see Fig. 3.47), the radius of curvature of the ellipse is given by:2
rd=-
b2
g
(3-22)
In general, the half width b of the dendrite stem is proportional to the primary arm spacing A.1 In cases where the arrangement of the dendrite trunks can be represented by a simple close-packed hexagonal array, the value of b is exactly equal to Al / -fj.2 Similarly, the total length of the dendrite stem g can be calculated by considering the difference between the tip temperature Ttip and the root temperature Troot' as shown in Fig. 3.48. Taking the thermal gradient in the mushy zone equal to G t we obtain:
(3-23)
This gives the following relationship between the primary dendrite arm spacing A.1, the tip radius "a and the thermal gradient GL:
Al
* = ( 311~Lrd
12 /
(3-24)
'--y--J
Fig. 3.47. Geometrical relationship between the radius of curvature r d and the width to length ratio of an elliptical dendrite.
262
~g
I I I
Heat flow
III(
Fig. 3.48. Definition of the dendrite tip Ttip and root Troot temperatures.
If we also take into account that the dendrite tip radius r d is inversely proportional to the square root of the crystal growth rate RL within the central range of RL (see Fig. 3.45), equation (3-24) reduces to: (3-25)
where c 1 is a kinetic constant which is characteristic of the alloy system under consideration. It is evident from the above analysis that the primary dendrite ann spacing cannot readily be characterised by one single parameter (e.g. the cooling rate), since its dependence on G L and RL have different exponents.
Example (3.8)
Based on equation (3-25), show that the following relationships exist between the primary dendrite ann spacing }q, the net arc power qo' and the weld travel speed v during thick plate and thin plate welding, respectively: Thick plate welding: (3-26) Thin plate welding: (3-27) where c2 and c3 are proportionality constants.
263
Solution Under 3-D heat flow conditions, the pseudo-steady state temperature distribution is given by the Rosenthal thick plate solution (equation (1-45)). For points located on the weld centre-line behind the heat source y = z = 0, and R* = -x. Hence, equation (1-45) reduces to: T-T ()
=-~
21tAx
The thermal gradient G L in the mushy zone close to the weld centre-line can be obtained by differentiating of the above equation with respect to the x-coordinate: .
(3-28)
By inserting the appropriate expressions for GL and RL into equation (3-25) (noting that RL = v at the weld centre-line), we arrive at an expression for A.l which is identical with the one presented in equation (3-26). Similarly, under 2-D heat flow conditions, the pseudo-steady state temperature distribution is given by the Rosenthal thin plate solution (equation 1-81). For points located on the weld centre-line behind the heat source y = 0 and r = -x = vt. If Ixl is sufficiently large, it is a fair approximation to set Ko(u)::::; exp(-u)-v'It I Zu . Hence, equation (1-81) reduces to: T-T
o
=-.;;:....()--
q I vd
pc~41tat
(3-29) From this we see that the primary dendrite arm spacing A.l during thin plate welding is interrelated to qo and v through a relationship of the type shown in equation (3-27). In Fig. 3.49 the validity of equation (3-26) has been checked against the experimental data of Jordan and Coleman.F who measured the primary dendrite arm spacing in different GMA AI-Mg-Mn welds. It is evident from this plot that their data-points can approximately be represented by straight lines passing through the origin, as required by the theory. Moreover, a closer inspection of the figure shows that the dendrite arm spacing close to the centre- line varies systematically from the bottom to the top of the weld. This observation is not surprising, considering the fact that the rate of solidification increases progressively from the toe to the surface of the plate.
Example (3.9)
Consider GTA butt welding of a 2mm thin aluminium sheet (AI-Mg alloy) under conditions similar to those employed in Example (3.7). Based on equation (3-25) calculate the relative change in the primary dendrite arm spacing A I from the fusion boundary to the weld centreline during solidification.
264
14
12
./
10
9-/
./ /
8
j)./
o
./
all /
6
./ ./. .../ /
AS'
0
Parallel to plate
surface
Just above toe ofweJd
.//
./ /
/'
10
20
30
40
50
60
70
80
Fig. 3.49. Experimental verification of equation (3-26). Data from Jordan and Coleman.F
Solution
As shown in Fig. 3.44, the thermal gradient, Gv decreases from about 25C mm ! at the fusion boundary to approximately 10C mnr ' close to the weld centre-line. This occurs parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s". Taking the primary dendrite arm spacing at the fusion boundary equal to 'h1*, the ratio 'h1l'h 1* in different positions of the weld becomes: Central region of columnar zone:
A;
(15)112(2.1 )114
Weld centre-line:
In contrast to that predicted above, the smallest dendrite arm spacing is normally observed at the weld centre-Iine.F' This has to do with the fact that the constant c 1 in equation (3-25), in practice, decreases with increasing distance from the fusion boundary due to solute segregation, which gradually reduces the coarsening rate of the dendrites. 3.5.3.3 Secondary dendrite arm spacing It follows from the above analysis that the primary dendrite arm spacing, once it has been
SOLIDIFICATION
BEHAVIOUR
OF FUSION
WELDS
265
established, will remain constant during subsequent cooling of the weld. This is not true of the secondary arms (see definition in Fig. 3.46), which undergo a continuous ripening process. Hence, their size and morphology will change with time as the thicker branches grow larger at the expense of the smaller ones.? The driving force for the ripening process is the disparity in chemical potential (interfacial energy) between branches with different curvature. Since this process is analogous to the Ostwald ripening of precipitates (see equation (2-76)), the spacing of the branches, x.2' will be a simple cube root function of the local solidification time to: 1,2
A2 = 5.5(M
where M is a mobility term defined as.? M
z
to)1/3
(3-30)
(3-31)
The parameter
n in equation
where 0' is the solid/liquid interfacial energy, and ~Sf is the entropy of fusion. In practice, the value of M can easily vary by an order of magnitude. Nevertheless, its effect on the secondary dendrite arm spacing is rather weak, since A2 is proportional to the cube root of M. Hence, a plot of the secondary dendrite arm spacing vs the local solidification time in a logarithmic diagram will normally reveal a straight-line relationship between x.2 and to with a slope close to 0.33, as shown in Fig. 3.50.
t
~
10
8
6 4
c<
C\I
1 ~--~------~------~--~--~~------~ 20 40 10
60
100
200
Fig. 3.50. Relation between secondary dendrite arm spacing A2 and local solidification time to in type
AISI 310 stainless steel welds. Data from Kou and Lee.31
266
Example (3.10)
Based on equation (3-30), derive a relationship between the secondary dendrite arm spacing A2, the net arc power qo' and the weld travel speed v during thick plate welding. Calculate then the secondary dendrite arm spacing in the centre of a thick GTA AI-Si weld deposited under the following conditions: 1= 140A, U
13V, v
To = 20C
Relevant physical data for the AI-Si system are given belowr' Co mL ko Ceut.
DL Q
A i1T*
r;
= 4.5 wt% = -7.5C per wt% = 0.13 = 12.6wt% = 5.5 X 10-3 mm? s-1 = 2 XI0-4mmoC = 0.167Wmm-1C-1 = 53C = 630C
Solution
The local solidification time to is the time for the dendrite array to pass an arbitrary point in the weld and is therefore a measure of the solidification rate. Referring to Fig. 3.48, the local solidification time is defined as: (3-33)
By inserting the appropriate expressions for GL (equation (3-28 and RL in position y we obtain the following relationship between to' q 0 and v during thick plate welding:
i1T*
m
= z = 0,
t -
o - 21CA(T
_ T)2
0
(q
0
/ v)
(3-34)
(3-35)
-1
SOLIDIFICATION
BEHAVIOUR
OF FUSION
WELDS
267
and
A
2
= 5.5[(2.1
Jl
~4Jlm Although reliable experimental data are not available for a direct comparison, the calculated value for A.2 is considered reasonably correct. In Al-Cu castings a secondary dendrite arm spacing of about 4Jlm corresponds to a cooling rate of the order of 103 C S-I, as shown in Fig. 3.51. In the present example, the cooling rate at the solid/liquid interface is close to:
0
C.R.= G R
L L
= 2nA(Tm -To)2
(q 0
/
v)
Cs-1
= 2145Cs-1
The operating parameters given in Example (3.10) are also applicable to single pass butt welding of thin aluminium plates. Based on equation (3-30), estimate the secondary dendrite arm spacing in the centre of a 2mm thick AI-Si butt weld deposited under such conditions.
Solution
In thin plate welding the local solidification time to is obtained by combining equations (3-29) and (3-33). Noting that RL = v at the weld centre-line, we get:
(3-36)
AI-4.5
wt%
Cu
Cooling rate, C/s ----Fig. 3.51.Relation between secondary dendrite arm spacing A,2 and cooling rate C.R. in Al-Cu castings. Data compiled by Munitz.F
268
(3-37)
A
2
It is evident from the above calculation that the secondary dendrite arm spacing is significantly coarser in thin plate welding compared with thick plate welding. This observation is not surprising, considering the pertinent difference in the heat transfer mode and thus the cooling rate between these two types of weldments. In the former case, we have: C.R.= G R = 21tAPCV (Tm - To)3 L L (qo / d)2
2
A comparison with the data in Fig. 3.51 shows that a cooling rate of about 78C compatible with a secondary dendrite arm spacing of 13 urn.
S-1
is
269
Fig. 3.52.Example of grain refinement in an AI-Mg-Si plasma arc weld due to TiAl3-precipitation fusion line, CL: centre-line). Courtesy ofM.!. Onsoien, SINTEF, Trondheim, Norway.
(FL:
i.,
(3-38)
During thick plate welding (3-D heat flow) the temperature gradient at the weld centre-line is given by equation (3-28). Taking RL = v, a combination of equations (3-28) and (3-38) gives:
(3-39)
where c4 is a kinetic constant which is characteristic of the alloy system under consideration. Similarly, for welding of medium thick plates (mixed heat flow), we get: qo
(v)n '
c5
0~n~1
(3-40)
where c5 is a new kinetic constant, and n is an exponent which varies between zero and unity depending on the mode of heat flow (i.e. thin, medium thick or thick plate welding).
270 (a)
(b)
Co
Heat flow
Fig. 3.53. Schematic diagrams showing different dendrite growth morphologies in castings; (a) Columnar dendritic growth, (b) Equiaxed dendritic growth. The diagrams are based on the ideas of Kurz and Fisher.?
Equation (3-40) predicts that the columnar to equiaxed transition in fusion welds occurs at critical combinations of the net arc power qo and the welding speed v. Thus, a decrease in qo must always be compensated by a corresponding increase in v in order to maintain equiaxed dendritic growth at the weld centre-line. This is also in agreement with general experience (see Fig. 3.54).
Example (3.12)
Based on the experimental data in Fig. 3.54, calculate the critical GLIRL ratio which provides equiaxed dendritic growth during thick plate welding of AI-Mg alloys (5083 series). Relevant thermal data for the AI-Mg system are given in Table 1.1 (Chapter 1).
271
t
~ ~ c. ~
ctS
1.6
1.2
CD
0.8
0.4
10
20
30
40
50
60
Fig. 3.54. Effect of net arc power qo and welding speed v on the columnar to equiaxed transition in
Solution
The critical GLIRL ratio is obtained by rearranging equation (3-39): 21t A (Tm-To)2 t.; =---~-~qov
I
cr.
= 21txO.149(650-20)2 400 X 8
"Cs mm
=116 "Cs mm
By multiplying fer. with the welding speed we see that the calculated value corresponds to a critical temperature gradient of about 928C mm -1. This means that the thermal conditions existing in welding will favour growth of equiaxed dendrites close to the weld centre-line. A requirement is, of course, that the melt contains a sufficient number of seed crystals to facilitate heterogeneous nucleation of new grains ahead of the advancing solid/liquid interface.
272
3.6.2 Nucleation mechanisms Several nucleation mechanisms have been proposed to explain the columnar to equiaxed transition during weld pool solidification, includingr'? (i) (ii) (iii) Heterogeneous nucleation Dendrite fragmentation Grain detachment.
The former mechanism is particularly relevant to welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. As already mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal deoxidation or stem from reactions between specific alloying elements which are deliberately introduced into the weld pool through the filler wire. The important effect of deoxidation practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA welds is shown in Figs. 3.55 and 3.56.
Under such conditions it is fairly simple to derive an expression for the solute concentration in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance gives:
CL d: = Cs d:
+ dCL(L*
- z - dz)
(3-41)
Since dz.
273
;g
e-
60
A
D
'a s '"0
CJ)
c:
0>
14 mm/s
3mm/s Bmm/s
0.29 wtO/o Ti
40
'5
~ ctS
c .2
0" CD
20
'LL
(a) 00 0.01 0.02 0.03 0.04 0.05
t5 ccs
t
.~
~
800
A
:::t
cD
'(i)
N
600
0.29 wtO/o Ti
c:
C) Q)
400
C>
> (b)
Cl)
200
0.01
0,02
0.03
0.04
0.05
Fig. 3.55. Effect of deoxidation practice (aluminium additions) on the columnar to equiaxed transition in ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the welds vs aluminium content, (b) Surface equiaxed grain size vs aluminium content. Data from Villafuerte
l-k
0)
J
o
d; - JdCL (L* - Z) C
Co L
cL
(3-43)
274 (a)
ca t."'C
en c
O')
40
'5 c
~ co
g
o
20~--------~~
U
u.
0.1
0.2
0.3
0.4
0.5
(b)
t
cD
'0
N
wt%
E :::1. 600
c: 'e Ol
CD
400
~ ~
0)
200
0.2
0.3
0.4
0.5
Titanium content,
wt/o
Fig. 3.56. Effect of deoxidation practice (titanium additions) on the equiaxed transition in ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the weld vs titanium content, (b) Surface equiaxed grain size vs titanium content. Data from Villafuerte et al.21
from which (3-44) and (3-45) wherefs is the fraction solidified (equal to z/L "),
275
~.
t
~
co
0
E (])
0
del
Distance,
Fig. 3.57. Solute redistribution during non-equilibrium
Equation (3-44) is valid up to CL = Ceut. where the remaining melt solidifies in the form of intercellular or interdendritic eutectics. The eutectic fractionfeut. is, in turn, given as:
f eut. = [ zsu: C
o
)lI k
C
o-1)
(3-46)
Figure 3.58 shows how the Scheil equation can be used for an evaluation of the microsegregation pattern in binary alloy systems by considering a small volume element of length L * which solidifies perpendicular to the cell/dendrite growth direction. Since the Scheil equation does not allow for solid state diffusion during solidification, a slightly refined version of this equation also exists in the literature: 1
(3-47)
(3-48)
276
T .L
C"
k----------
9 --------~
Fig. 3.SS.Idealised model for micro segregation in ingots and castings (schematic).
Equation (3-47) has been used by Brody and Flemings-" to evaluate the effect of solid diffusion on the amount of eutectic in different cast structures. The extent of this diffusion depends on the dimensionless product ex* ko and it becomes significant only for values of ex * ko greater than about 0.1. I Note that in cellular or cellular-dendritic growth the primary dendrite arm spacing Al provides a measure of the diffusion length. Taking L * = A1/2, equation (3-48) can be rewritten as: (3-49) In contrast, during equiaxed dendritic growth the secondary dendrite arm spacing A2 is a more appropriate dimension for the solidification microstructure, since the back diffusion process here occurs mainly between secondary arms and not between primary trunks.i In such cases we may write: (3-50)
Example (3.13)
Consider GTA welding of an AI-2wt%Cu tions: Welding speed Cooling rate Primary dendrite arm spacing
277
Estimate on the basis of the Scheil equation the degree of microsegregation occurring during weld metal solidification. Relevant physical data for the Al-Cu system are given below: ko Ceut. ~T* Ds
Solution
S-I
The local solidification time can be calculated from equation (3-33). Taking RL = v, we get:
Since the solidification conditions in this case facilitate the formation of a cellular-dendritic type of substructure close to the weld centre-line (see Fig. 3.43), the characteristic diffusion length L * is determined by the primary dendrite arm spacing hI. The product 0. * ko is then given as: a*k = 4(3Xl0- )(6.6Xl0- )0.17=5.5XI0-4 o (12 X 10-3)2 Because the numerical value of 0. * ko is very small, the contribution from diffusion in the solid state can be neglected. Hence, the extent of microsegregation occurring during solidification can be evaluated from equation (3-45). Taking Co = 2wt% and ko = 0.17, we obtain:
7 2
The results are presented graphically in Fig. 3.59. As expected, the copper concentration is seen to increase monotonically from the core to the periphery (surface) of the dendrite stem. When the eutectic composition is reached, the remaining fraction solidifies as AI(5.6wt%Cu) + CuAI2 According to equation (3-46) the eutectic fraction is equal to:
-12 )
. ~3%
These results should be compared with the experimental data of Brooks and Baskes'" replotted in Fig. 3.60. It is evident that the measured copper concentration profile for the AI2wt%Cu GTA weld is similar to that inferred from the Scheil equation, although the observed dendrite core concentration lies significantly above the predicted one. Consequently, the Scheil equation gives a reasonable description of the segregation pattern during weld metal solidification, in spite of the simplifying assumptions inherent in the model.
278
z.u rn
5
~ ~
4 5
t
0
CSL
z
L*= A1/2
~ ... ....
c:
Q)
0 0
c: :::J
Fig. 3.59. Predicted Cu concentration profile based on the Scheil equation (Example (3.13)).
c:
o
E 5
Q)
o c:
::1
10
Position,
30
Jl m
Fig. 3.60. Electron microprobe analysis of Cu across primary solidification 2wt% Cu GTA weld. Data from Brooks and Baskes.P
3.7.2 Macrosegregation Macrosegregation in the form of solute banding is also frequently observed after weld metal solidification.Y This type of segregation arises from a periodic enrichment and/or depletion of solute elements, caused by a non-steady advancement of the solidification front (see Fig.
SOLIDIFICATION
BEHAVIOUR
OF FUSION
WELDS
279
Fig. 3.61. Example of solute banding (dark stripes) in a phosphorus-containing land and Davis.l"
3.61). Although the origin of the phenomenon is not yet fully understood, it is reasonable to assume that the pertinent fluctuations in the solidification rate occurs as a result of frequent variations in the heat flux during welding. Direct experimental evidence for such a correlation can be obtained from the data of Garland and Davis.l" Moreover, there is a pronounced tendency for alloying and impurity elements to segregate to the weld centre-line where the columnar grains growing from each side impinge. This, in turn, may produce hot tearing as a result of the formation of low-melting eutectics between the dendrite arms. In general, the risk of hot tearing decreases with increasing width to depth ratio of the weld because of a more favourable crystal growth mode, as illustrated in Fig. 3.62. 3.7.3 Gas porosity Troublesome impurities in fusion welds are those which precipitate, alone or in combination with other elements, to form various gaseous reaction products." These reaction products may be simple diatomic gases such as H2 and N2 or more complex gaseous compounds like CO, or H20. An illustration of gas porosity due to hydrogen evolution is contained in Fig. 3.63. Following the discussion in Chapter 2, absorption of gases occurs readily in the hot part of the weld pool during welding because of interactions with the surrounding arc atmosphere. During the subsequent stage of cooling a supersaturation rapidly increases due to the associated decrease in the element solubility with decreasing temperatures (see Figs. 2.7 and 2.8). The system will respond to this supersaturation by rejection of dissolved elements from the liquid, either through a gas/metal reaction (desorption) or by precipitation of new phases. The latter incident may result in porosity if the gas bubbles become trapped in the weld metal solidification front. 3.7.3.1 Nucleation of gas bubbles In liquid metals a bubble will be stable if the gas pressure inside the bubble is sufficiently high to balance the external forces. These external forces are the liquid/vapour interfacial energy
280
.J:
Q)
Fig. 3.62. Effect of weld width to depth ratio on the tendency to centre-line cracking; (a) Correct width to depth ratio, (b) Incorrect width to depth ratio.
Fig. 3.63. Example of gas porosity in a GTA 7106 aluminium weld. After D'annessa.'?
Fig. 3.64. Growth of a gas bubble due to diffusion in the liquid phase (schematic).
281
0'*, the metallostatic pressure head Pm' and the ambient pressure Pa. Thus, for a stable bubble,
we may write:
Pg = Pa
+ Pm +-
20'*
rg
(3-51)
where P g is the total gas pressure inside the bubble, and rg is the radius of the gas bubble. For shallow welds, the contribution from the metallostatic pressure head Pm can be ignored. Hence, equation (3-51) reduces to: P gar ~ P
+g
20'*
(3-52)
Since 0'* is typically of the order of 1 J m-2 (9.87 atm urn) for most gas-metal systems, we may write: P gar ~P
+-g
19.74
(3-53)
It is evident from equation (3-53) that the interfacial energy term is negligible at large values of 'g. However, if the radius of curvature becomes sufficiently small, extremely large pressures are required to maintain a stable bubble. Thus, there is a bubble nucleation problem, which is formally similar to that of nucleation of a solid from a liquid (discussed in Section 3.3.1). In fact, it can be shown on the basis of classic nucleation theory that the driving force normally associated with rejection of dissolved gases in liquid metals is by far too small to allow for homogeneous nucleation of gas bubbles in the weld pool during cooling. This, in turn, implies that solid particles (e.g. inclusions) entrained in the liquid metal will be the most probable sites for gas bubble formation in fusion welds. 3.7.3.2 Growth and detachment of gas bubbles As shown in Fig. 3.64, growth of gas bubbles in liquids is a diffusion-controlled the time dependence of the mean bubble radius is given by:38
process where
Q* = CL
Ce
, Q*
> 10
(3-55)
Pg where C L is the molar concentration of solute in the supersaturated liquid, C e is the equilibrium molar concentration of the solute at the gaslliquid interface, and Pg is the gas density (in the same units as CL and Ce). Equations (3-54) and (3-55) may be used to estimate the growth rate of a bubble while it is still attached to the solid/liquid interface. The bubble becomes detached when the buoyancy force, which is pushing it upwards, exceeds the surface tension force, which tends to keep it attached to the solid surface. The bubble radius at which detachment occurs is given by the socalled Fritz equationr"
282
r (crit.)
g
= 7.5 x
10-4~[__ _]1/2
2_0_* gc(PZ-Pg)
(3-56)
where gc is the gravity constant, and ~ is the wetting angle (in degrees). Based on equations (3-54) and (3-56) it is possible to evaluate the conditions for growth and detachment of gas bubbles during weld metal solidification. This is a subject of considerable importance in welding, since the pore formation will inevitably affect the mechanical integrity of the weldment.
Example (3.14)
Consider GTA butt welding of a 3mm thin AI-Mg sheet under the following conditions: 1= 150A, U
13V, v
= 5mm
S-I,
'Tl
= 0.6,
To
= 20C
Suppose that gas bubbles form at the solidlliquid interface during solidification due to rejection of dissolved hydrogen from a supersaturated liquid. Based on equations (3-54) and (3-56) estimate the maximum theoretical radius of the gas bubbles and the critical radius at which the bubbles detach themselves from the solidlliquid interface during welding. Relevant physical data for the AI-Mg system are given below: Tm 650C 200C = 0.149 W mrrr ' C-l = 0.0027 J mm-3C-1
st=
A pc
CL
C, DL gc ~ 0*
PI
= 2ml H2 per 100g fused metal (at 25C). = 0.7 ml H2 per 100g fused metal (at 25C). DB = 0.32 mm2 s-1 = 9.81 m s-2 = 100 = I J m-2
= PAl
= 2350 kg m-3
Solution
Since particles located at the solid/liquid interface are the most probable sites for hydrogen gas evolution, the local solidification time to provides a conservative estimate of the growth time t in equation (3-54). From equation (3-36), we have: 200(150 x 13 x 0.6 / 3)2 s = 1.9s 21t x 0.149 x 0.0027(5)2(650 - 20)3 In order to calculate the growth constant from equation (3-55), it is necessary to convert the concentration driving force to molar units:
SOLIDIFICATION
BEHAVIOUR
OF FUSION
WELDS
283
C = 2x2x2350 mol Hcm=' =3.84x10-4moIHcm-3 L 24465 x 1000 C = 2xO.7x2350 mol Hcm-3 = 1.34 x 10-4 mol Hcm=' e 24465 x 1000 The molar density of the gas Pg is obtained from the ideal gas law:
p =L =
g
RT
82.06 x 923
The value of the growth constant is thus: U* = CL -Ce = (3.84-1.34)x10-4 Pg 0.13x10-4 =19.23
The maximum theoretical radius of the gas bubbles can now be evaluated from equation (3-54) by inserting the appropriate values for U*, DH, and to:
rg(max)
x 1.9 mm
Similarly, the critical radius at which the hydrogen bubbles become detached may be estimated from equation (3-56). Since PAl PH2' we obtain:
::= 7.5
10-2
By inserting this value into equation (3-54) it is also possible to estimate the average bubble detachment frequency under the prevailing circumstances:
i, =_1_=(
2Q*
rg(erit.)
t(erit.)
J2
=
DH
=(2
0.7
19.23J20.32S-1
966s-1
Since rg (erit.) rg (max), the maximum pore radius will probably be closer to O.7mm than 30mm in a real welding situation. This is also in agreement with practical experience (see Fig. 3.65).
284
t~ 70 50 99
90
70
50
0.01
0.1
10
3.7.3.3 Separation of gas bubbles It follows that the gas bubbles will start to migrate towards the surface of the weld immediately after they become detached from the solid/liquid interface. Small, spherical bubbles (characterised by a bubble Reynold number less than 2) will rise at a terminal velocity determined by Stokes law:38 "
= 1:j..l gc(pt-Pg)
d2
(3-57)
where dg is the diameter of the gas bubbles, and f..l is the viscosity of the liquid. Larger gas bubbles (characterised by a bubble Reynold number between 2 and 400) will also rise in a rectilinear manner, but their terminal velocity may be as much as 50% greater than that predicted from Stokes law.38 Depending on their flotation rate, such ascending gas bubbles will either escape to the weld surface or be trapped in the weld metal solidification front in the form of macroscopic gas porosity (see Fig. 3.63).
Example (3.15)
Based on Stokes law (equation (3-57 calculate the rising velocity of a 0.2mm large hydrogen bubble ascending in liquid aluminium. Relevant physical data for liquid aluminium are given below:
285
Since PI
d2
(2 x 10-4)2 (18)9.4xl0
-4
(9.81)(2350)xl03mms-1
= 54.5mms-1
It is evident from the above calculations that the flotation rate of such gas bubbles is quite high and of the same order of magnitude as the weld pool fluid flow velocity (discussed in Section 2.11.2). Hence, the buoyancy force would be expected to playa significant role in the separation process of detached gas bubbles in the weld pool. On this basis it is not surprising to find that a change in the welding position (e.g. from flat to overhead) results in a dramatic increase in the volume of porosity during GTAW of aluminium alloys (see Fig. 3.66). Although a great deal has been reported on the causes and effects of porosity in weld metals (see Ref. 37 for an excellent discussion), little is known about the mechanism of pore formation relative to solidification mechanics, nucleation, growth and transport of gas bubbles in the weld pool. Consequently, a more fundamental approach to the porosity problem in fusion welding (along the lines indicated above) is necessary in order to obtain a verified, quantitative understanding of the phenomenon.
GTAW
1.0
8 ,.. ::::.
E
C)
ur e 0
Q)
Co
0.6
0
E :::s 15
>
Welding orientation, degrees Fig. 3.66. Porosity in GTA welds deposited on 2mm sheets of aluminium to overhead. Data compiled by Devletian and Woods.37
286
3.7.4 Removal of
microsegregations
during cooling
As shown in Section 3.7.1, the characteristic growth pattern of cellular and dendritic -solidification, in combination with the rapid cooling rates normally associated with fusion welding, lead to extensive segregation of alloying and impurity elements to the intercellular or interdendritic spaces. Segregation produced by this means is remarkably persistent, and can in certain cases only be eliminated by prolonged high-temperature heat treatment. A simplified analysis of homogenisation of micro segregations in fusion welds is given below.
3.7.4.1
Diffusion model It is a reasonable approximation to regard microsegregations in cast structures as periodic, where the concentration at any point and time along an arbitrary line, C(x,t), is given by:41 C(x, t) = Cavg. + (Cmax. - Cavg.) sln(nx 1 lavg. )exp(
. -1t 2
Dst 1 lavg.)
. 2
(3-58)
where Cavg. and Cmax. are as indicated in Fig. 3.67, lavg. is the average distance between adjacent maxima and minima, and Ds is the diffusivity of the solute in the solid. Equation (3-58) states that the 'concentration remains constant and equal to Cavg.at positions x = 0, x = lavg., x = 2lavg. etc., while the peak of the sine wave is attained at distances x = lavg/2, x = 5lavg/2 etc. during the decay. If only the peak concentration is considered, the sine term becomes equal to unity and equation (3-58) reduces to: (3-59) from which the homogenisation time during isothermal heat treatment
thorn.
can be obtained:
thorn. -
<; ]
(3-60)
Since the diffusion length lavg., in practice, is equal to the half dendrite arm spacing (AI 12 or A.2/2), equation (3-60) predicts that the homogenisation time is proportional to the square of Al or A2. The latter parameters are, in tum, determined by the thermal conditions existing within the mushy zone during weld metal solidification, and are therefore sensitive to variations in welding variables such as the net arc power q 0 and the travel speed v. to continuous cooling During cooling of the weld in the solid state, some equalisation of microsegregations will occur through diffusion. The extent of this diffusion can be reported in terms of an equivalent isothermal homogenisation time at a chosen reference temperature Tr. If Qs denotes the activation energy for diffusion of the solute in the solid, the equivalent isothermal homogenisation time t; at T; is given by: t;
3.7.4.2 Application
= exp(Qs
IRTr)
f exp(-Qs
tl
ti
I RT)dt
(3-61)
287
iavg.
2avg.
3avg.
The integral on the right-hand side of equation (3-61) represents the kinetic strength of the weld thermal cycle with respect to homogenisation (solid state diffusion), and can be determined by means of numerical methods when the cooling programme is known. The extent of solute diffusion may then be evaluated from equation (3-60) by inserting representative values for t, and lavg.: (3-62)
Example (3.16)
ln low-alloy steel, both carbon, phosphorus, and manganese are known to segregate to the interdendritic spaces during solidification. Consider SA welding on a thick plate of steel under the following conditions: I
450A, U
30V, v
2.6mm
S-I,
To
= 20C
Based on equations (3-61) and (3-62) calculate the extent of homogenisation occurring within the solid weld metal during cooling in the austenite regime (i.e. from 1520 to 700C). Relevant data for the diffusivity of carbon, phosphorus, and manganese in austenite are given below: Dl; (rnm-
s')
S-I)
S-I)
23.43 exp (- 1477151RT) 51 exp (- 2301201RT) 140 exp (- 2860001RT) 0.025 W mm' C-l = 15 urn 8.314 J K-I mol "
Dfo
(mm?
D1n
A
lavg.
(rnm?
AI/2
288
Solution
During thick plate welding, the cooling programme can be calculated from equation (1-45). For points located on the weld centre-line behind the heat source y = z = 0, and R* = -x = vt. Hence, equation (1-45) reduces to: T-T
o
=_0_
Iv
2nAt
T=20+
From this it is seen that the actual transformation temperatures (i.e. 1520 and 700C) are reached after 20.9 and 46.2s, respectively. If 1350C is used as a reference temperature, the kinetic strength of the cooling cycle with respect to solute diffusion can be expressed as:
t2=46.2
11350 =
exp(Qs /1623R)
tl
=20.9
Numerical integration of this equation over the weld cooling cycle gives: t1350(C)
= 7.2s,
t1350(P) = 8.0s,
t1350(Mn)
= 9.3s
At 1350C (1623 K) the diffusivities of C, P and Mn in austenite are: D~ (1350C) D~ (1350C) (1350C) 4.1 X 10-4 mm? s-1 2.0 X 10-6 mm? s-1 8.7 X 10-8 mm? S-1
D1n
= exp[-
1t
10-57
= exp[
= 0.50
= exp[Mn
2 1t
(8. 7 X (15 X 10
1~:~9.3]
)
= 0.97
289
It is evident from the above calculations that interdendritic segregations of carbon are readily removed during cooling of the weld from the solidification temperature due to a high diffusivity of C in austenite. In contrast, segregations of phosphorus and manganese are much more persistent, since the diffusion cannot keep pace with the falling temperature. In the latter case the initial concentration gradients will largely be maintained down to temperatures where the austenite to ferrite transformation occurs (see ion micrograph of phosphorus segregations in Fig. 3.39). This, in turn, can promote the formation of different types of transformation products (ranging from ferrite to martensite depending on the nature of the segregants) along the primary solidification boundaries due to local variations in the steel harden ability. An example of martensite banding in an AISI 4340 SA steel weld is shown in Fig. 3.68. It should be noted that the conditions for equalisation of interdendritic segregations (e.g. phosphorus) are not dramatically altered by a change in the heat input as long as the t; /l;vg. ratio in equation (3-62) remains fairly constant. For instance, if the net arc power qo in the above example is reduced by a factor of four (conforming to welding with covered electrodes), the corresponding change in lavg. and t 1350 (P) will be:
lavg.
= Al
/2
.
z
15 {4l1ffi
= 7.511ffi
tz;
(7.5)
This value is reasonably close to that obtained (0.036 s J.lm-2), as shown in the above example.
during high-heat
Fig. 3.68. Banding of martensite (M) along primary solidification boundaries in an AISI 4340 SA steel weld. After Burck.F
290
291
Liquid (L)
Tper.
Fig. 3.70. Peritectic reaction by which the secondary 'Y p -phase grows along the surface of the primary &p -phase (schematic).
Fig. 3.71. Peritectic transformation involving long-range diffusion of solute atoms through the secondary 'Y p -phase (schematic).
292
Delta ferrite
Austenite
Fig. 3.72. Primary delta ferrite solidification with subsequent growth of austenite along the boundaries of the primary 8Fe-phase (schematic).
3.8.2.2 Primary precipitation of austenite Under certain welding conditions the weld metal can solidify directly as austenite without any primary precipitation of delta ferrite (see Fig. 3.73). This solidification mode has been observed during welding with covered electrodes on high-carbon steels.l" Because of competitive growth occurring during the initial stage of the solidification process, the columnar austenite grains will have one of their < 100> axes aligned in the direction of the maximum temperature gradient in the weld pool. Consequently, the weld metal will develop a solidification texture which is similar to that observed during primary precipitation of delta ferrite. 51 3.8.2.3 Primary precipitation of both delta ferrite ant! austenite In certain low-alloy steel weld metals the austenite grain boundaries will systematically cross the original delta ferrite solidification boundaries, as indicated in Fig. 3.74 because of a shift in the mechanism of the peritectic transformation.3,22,24,54-56 This point is more clearly illustrated in Fig. 3.75, which shows ion (SIMS) micrographs of phosphorus and boron segregations
<100>y-Fe
Columnar grain
293
Fig. 3.74. Optical micrograph showing austenite grain boundaries crossing primary delta ferrite solidification boundaries in a low-alloy SA steel weld (indicated by arrows). After Kluken et al.22
at prior solidification (cell) and austenite grain boundaries", respectively in a SA steel weld metal deposited under the shield of a basic flux. It is obvious from Fig."3.75 that there is no matching between the two types of boundaries in this particular case. It follows from the analysis of Kluken et al.22 that the observed shift in the mechanism of the peritectic transformation can probably be attributed to heterogeneous nucleation of austenite at inclusions (e.g. AI203), which is energetically more favourable than nucleation at 8p/8pe grain boundaries. Under such conditions, the austenite is not bound by an orientation relationship with the delta ferrite and is thus free to grow across the original delta ferrite columnar grain boundaries, as shown schematically in Fig. 3.76. The austenite grains will therefore adopt a morphology which is different from the columnar one.24,55,56
References
1. 2. 3. 4. 5. 6. 7. 8. 9. M.C. Flemings: Solidification Processing, 1974, New York, McGraw-Hill Book Company. W. Kurz and D.J. Fisher: Fundamentals of Solidification, 3rd Edn, 1989, Aedermannsdorf (Switzerland), Trans. Tech. Publications. G.J. Davis and J.G. Garland: Int. Metall. Rev., 1975, 20, 83-106. S.A. David and J.M. Vitek: Int. Mater. Rev., 1989,34,213-245. D.A. Porter and K.E. Easterling: Phase Transformations in Metals andAlloys, 1981, Wokingham (England), Van Nostrand Reinhold Co. Ltd. K.E. Easterling: Introduction to the Physical Metallurgy of Welding , 198'3, London, Butterworths & Co (Publisher) Ltd. 0. Grong, T.A. Siewert and G.R. Edwards: Weld. l., 1986,65, 279s-288s. G.M. Oreper, T.W. Eagar and J. Szekely: Weld. l., 1983,62, 307s-312s. Y.H. Wang and S. Kou: Proc. Int. Conf. on Advances in Welding Science, Gatlinburg, TN, May, 1986,65-69, Publ. ASM International.
and AIO-, both types of ions will be sampled in the SIMS analysis. traces of aluminium oxide
the white spots observed within the interior of the grains are probably
294
(a)
(b)
Fig. 3. 75. Ion (SIMS) micrographs showing austenite grain boundaries crossing primary delta ferrite solidification boundaries in a low-alloy SA steel weld; (a) Phosphorus segregations at primary solidification (cell) boundaries, (b) Boron segregations at prior austenite grain boundaries (same area as in Fig. 3.75(a)). After Kluken et al. 22
295
Fig. 3.76. Primary delta ferrite solidification with subsequent nucleation of austenite at inclusions (schematic). 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. K.C. Mills and B.J. Keene: Int. Mater. Rev., 1990,35,185-216. T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Me ta ll. Trans., 1991, 22B, 243-257. C.R. Heiple and J.R. Roper: Weld. J., 1982,61, 97s-102s. C.R. Heiple and P. Burgardt: Weld. J., 1985,64, 159s-162s. T. Habrekke, H.O. Knagenhjem and J.O. Berge: Proc. 9th. Int. Conf. on Offshore Mech. and Arctic Eng in. , Houston, TX, Feb., 1990,511-515, Publ. ASME. H. Biloni: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and P. Haasen), 1983, Amsterdam, North-Holland Physics Publ., 478-579. T. Senda, F. Matsuda and M. Kato: Techn. Reports of the Osaka University (Japan), 1970,20, 527-558. M. Rappaz, S.A. David, J .M. Vitek and L.A. Boatner: Metall. Trans. A, 1989, 20A, 1125-1138. L.F. Mondolfo: Mat. Sci. and Techno!., 1989,5,118-122. B.L. Bramfitt: Me ta ll. Trans., 1970, 1, 1987-1995. G.N. Heintze and R. McPherson: Weld. J., 1986,65, 71s-81s. J.C. Villafuerte, E. Pardo and H.W. Kerr: Metall. Trans. A, 1990, 21A, 2009-2019. A.O. Kluken, 0. Grong and G. Rervik: Me tall. Trans. A, 1990, 21A, 2047-2058. H. Yunjia, R.H. Frost, D.L. Olson and G.R. Edwards: Weld. J., 1989,68, 280s-289s. C.E. Cross, 0. Grong, S. Liu and J.F. Capes: Applied Metallography, (Ed. G.F. Vander Voort), 1986, New York, Van Nostrand Reinhold Inc., 197-210. A.O. Kluken and 0. Grong: Proc. Int. Conf. on Recent Trends in Welding Sci. and Technol. (TWR 89), Gatlinburg, TN, May, 1989,781-786, Publ. ASM International. S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons. M.F. Jordan and M.C. Coleman: Brit. Weld. J., 1968,15,553-558. C.E. Cross and G.R. Edwards: Treatise on Mat. Sci. Techno!., 1989,31, 171-187. W. Kurz, B. Giovanola and R. Trivedi: Acta Metall., 1986,34, 823-830.
296
30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56.
R. Trivedi and W. Kurz: Acta Me tall., 1986,34,1663-1670. S. Kou and Y. Lee: Me tall. Trans., 1982, 13A, 1141-1152. A. Munitz: Me tall. Trans., 1985, 16B, 149-161. F. Matsuda, K. Nakata, K. Tsukamoto and K. Arai: Trans. JWRI, 1983,12,81-87. H.D. Brody and M.C. Flemings: Trans. AIME, 1966,236,615-623. J.A. Brooks and M.1. Baskes: Proc. Int. Con! on Trends in Welding Research (TWR 86), Gatlinburg, TN, May, 1986, 93-99, Publ. ASM International. J.G. Garland and GJ. Davis: Brit. Weld. J., 1970,17,171-175. J.H. Devletian and W.E. Woods: Weld. Res. Council Bull., 1983,290, 1-18. J. Szekely and NJ. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John Wiley & Sons, Inc. A.T. D'annessa: Weld. J., 1967,46, 491s-499s. Y. Tomii, A. Sakaguchi and M. Mizuno: Proc. 4th Int. Con! on Aluminium Weldments, Tokyo (Japan), 1988,46-60, Publ. Japan Light Metal Welding and Construction Association. J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons, Inc. P. Burck: M.Sc. thesis, 1984, Colorado School of Mines, Golden, Colorado (USA). M. Hillert: Solidification and Casting of Metals, 1979, London, The Metals Society, 81-87. H. W. Kerr, J. Cisse and O.F. Bolling: Acta Me tall. , 1974,22, 677-686. J. Cisse, G.F. Bolling and H.W. Kerr: 1. Cryst. Growth, 1972,13/14,777-781. J. Cisse, H.W. Kerr and O.F. Bolling: Metall. Trans., 1974,5,633-641. H. Fredriksson: Met. Sci., 1976, 10, 77-86. H. Fredriksson and J. Stjerndahl: Metall. Trans., 1977, 8A, 1107-1115. H. Fredriksson and J. Stjerndahl: Met. Sci., 1982, 10,575-585. N.S. Pottore, C.1. Garcia andAJ. DeArdo: Metall. Trans., 1991, 22A, 1871-1880. A.O. Kluken, 0. Grong and J. Hjelen: Metall. Trans., 1991, 22A, 657-663. N. Suutala, T. Takalo and T. Moisio: Me tall. Trans., 1979, lOA, 1183-1190. S.A. David: Weld. J., 1981,60, 63s-71s. A.A.B. Sugden and H.K.D.H. Bhadeshia: Metall. Trans., 1987, 19A, 669-674. J.G. Garland and P.R. Kirkwood: Proc. Int. Symp. on Welding of Line Pipe Steels, S1. Louis (USA), 1977, 176-227, Publ. Welding Research Council (New York). R.C. Cochrane: Weld. in the World, 1983,21,16-29.
thermal diffusivity
(mm? s ')
(~%AI)weld
difference between total and acid soluble Al in weld metal (wt%) half width of dendrite stem (Jim) body-centred cubic structure various kinetic constants and temperature-dependent parameters
lattice parameter (nm) b equilibrium ~/OL solvus temperature (C) bcc area of embryo-liquid face (nr') area of embryo-substrate terface (m'') inter-
in-
297
eutectic fraction fraction solidified shape factor face-centred cubic structure length of dendrite stem (urn) gravity constant (m
S-2)
<;
Co
average solute concentration in solid phase (wt%) initial alloy concentration (wt%) equilibrium concentration of solute at gaslliquid interphase (wt% or mol crrr ') eutectic concentration (wt%) concentration of alloying element in liquid phase (wt% or mol crrr ') maximum solute concentration in solid phase (wt%) concentration of alloying element in solid phase (wt%) solute concentration at position x and time t (wt%) cooling rate (OC S-1 or K plate thickness (mm) the interatomic spacing along [zzvw], the interatomic spacing along [uvw]s diffusion coefficient in liquid (rnm- s-l) diffusion coefficient in solid (mm? S-I) diffusion coefficient of arbitrary element X (mrrr' S-I) electron beam welding
S-I)
t.:
is
f(3) fcc g ge GL GMAW GTAW ~GD
Ce
Ceut. CL
temperature gradient in weld pool (C mm ' or K rrr ') gas metal arc welding gas tungsten arc welding activation energy for diffusion of atoms across the interface (J) energy barrier to heterogeneous nucleation (J) energy barrier to homo geneous nucleation (J) free energy change associated with the embryo formation (J m=') heat content per unit volume at the melting point (J mm -3) latent heat of melting (J rrr ') a low-index plane in the nucleated solid a low-index substrate amperage (A) Boltzmann constant (1.381 X 10-23 J K-l) equilibrium partition coefficient modified Bessel function of second kind and zero order plane of the
<:
Cs C(x,t)
~GZet ~GZom
~Gv
C.R. d
d[uvw]n
d[uvw]s
Hm-Ho
DL
~m (hkl)n
Ds
Dx
(hkl)s
I EBW f1 fb k frequency factor (s+) detachment frequency of gas bubbles (S-I) critical GLIRL ratio which facilitates equiaxed dendritic growthr'C s mm -2 ) ko
fer.
Ko
298
length (mm)
lavg.
of columnar
grain
rs
average distance between adjacent maxima and minima concentrations (urn) total length of circle segment (mm) total length of volume element (mm or urn) slope of the liquidus curve (C perwt%) mobility term (mm? exponent dimensionless rameter operating paS-I)
* r, R R*
Rhkl
universal gas constant (8.314 J K-l mol ') three-dimensional vector (mm) radius
L*
mL
dendrite growth rate in hkl direction (mm S-I) steady-state growth rate of solid/liquid interface (mm S-I) local crystal growth rate (mm
S-I)
M
n
nj
Ri
RL
Nhet
RN
nominal crystal growth rate (mm S-I) submerged arc welding secondary spectrometry ion mass
SAW SIMS
Nv
density of nucleation sites per unit volume of melt (nuclei per mrrr') analytical weld metal oxygen content (wt%) ambient pressure (atm) total gas pressure in bubbles (atm) metallostatic (atm) pressure head
[%OJanal.
SMAW f1Sf
shielded metal arc welding entropy of fusion per unit volume (J m-3 K-l) time (s)
Pa Pg
t1, t2 thorn. to t,
Pm
qo
net arc power (W) activation energy for diffusion of solute in solid (kJ mol=l) two-dimensional tor (mm) radius vee-
local solidification time (s) equi valent isothermal mogenisation time (s) temperature (C or K) ambient temperature K) ho-
Qs
T r To
(C or
rd rg
TA
299
primary peritectic phase efficiency factor interface normal angle (degrees) dimensionless rameter molar density diffusion pa-
TL
liquidus temperature according to phase diagram (C or K) melting point (C or K) peritectic temperature (C or K) reference temperature K)
8p
'll ex
Tm
r;
ex *
t;
Troot
C C or
Pg
of gas (mol
crrr ')
density of liquid (kg m-3 or g cm=') volume heat capacity (J mm>' C-I) cell/dendrite alignment angle (degrees) critical cell/dendrite ment angle (degrees) Gibbs-Thomson (mm 'c ormK) dimensionless stant dimensionless align-
PI
r:
f1T f1T
pc
<P temperature difference between tip and root of dendrite stem (C or K) a low-index (hkl)n a low-index (hkl)s voltage (V) rising velocity of gas bubbles (mm S-I) volume of solid embryo (rrr') welding speed (mm x-axis/welding (mm) y-axis/transverse (mm) z-axis/through rection (mm) dimensionless ness delta ferrite lattice disregistry between a nucleus n and a substrate s direction in
Q
4>*
[uvw]n
coefficient
[zzvw],
direction
in
Q*
growth con-
U ug VE
v x
e em
~
temperature
dimensionless temperature at the melting point wetting angle (degrees) viscosity of liquid metal (kg m-1 S-I) angle between [uvw]s [uvw], (degrees) and
S-I)
direction
direction
"I
Fe
austenite secondary peritectic phase total grain rotation angle (degrees) solidlliquid interfacial energy (J m-2) liquid/vapour ergy (J m-2) interfacial en-
z
8
thickness di-
'Yp
plate thickco
8Fe
B(hkl)s (hkl)n
a*
300
'YEL
embryolliquid interfacial energy (J m-2) embryo/substrate energy (J m-2) substrate/liquid energy (J m-2) interfacial
hI
'YES
primary dendrite arm spacing close to fusion boundary (urn) secondary dendrite arm spacing (urn)
'YSL
interfacial
h2
(W
4.1 Introduction
Precipitate stability is an important aspect of welding metallurgy. Normally, modern structural steels and aluminium alloys derive their balanced package of high strength, ductility and toughness via optimised thermomechanical processing to produce a fine-grained, precipitation strengthened matrix. This delicate balance of microalloy precipitation and microstructure, however, is significantly disturbed by the heat of welding processes, which, in tum, affects the mechanical integrity of the weldment. When a commercial alloy is subjected to welding or heat treatment several competitive processes are operative which may contribute to a change in the volume fraction and size distribution of the base metal precipitates. The two most important are: 1 (i) (ii) Particle coarsening (Ostwald ripening) Particle dissolution (reversion)
Referring to Fig. 4.1, particle coarsening occurs typically at temperatures well below the equilibrium solvus Teq. of the precipitates, while particle dissolution is the dominating mechanism at higher temperatures. On the other hand, there exists no clear line of demarcation between these two processes, which means that particle coarsening can take place simultaneously with reversion in certain regions of the weld where the peak temperature of the thermal cycle falls within the 'gray zone' in Fig. 4.1. Nevertheless, it is important to regard them as separate processes, since the reaction kinetics are so different (coarsening is driven by the surface energy alone, whereas dissolution, which involves a change in the total volume fraction, is driven by the free energy change of transformation).
302
Fig.4.1. Schematic diagram showing the characteristic temperature ranges where specific physical reactions occur during reheating of grain size-controlled and dispersion-hardened materials. (4-1)
= Mio
- TtlS
= -RTln
(4-2)
where iVlO and tlS are the standard enthalpy and entropy of reaction, respectively. The other symbols have their usual meaning (see Appendix 4.1). When pure AnBm is used as a standard state, the activity of theprecipitate (aAnBm) is equal to unity. In addition, for dilute solutions it is a fair approximation to set aAz[%A] and aBz[%B], where the matrix concentrations of elements A and B are either in wt% or at% * . Hence, the solubility product can be written as:
log[%A]n[%B]m
=---=
ss:
R'
llHo R'T
c* --
D* T
(4-3)
*For the solute, the standard state is usually a hypothetical equal to unity as long as Henry's law is obeyed.
1% solution.
303
Table 4.1 gives a summary of equilibrium solubility products for a wide range of precipitates in low-alloy steels and aluminium alloys. In addition to the compounds listed in Table 4.1, different types of mixed precipitates may form within systems which contain more than two alloying elements.v" However, since the presence of such rnultiphase particles largely increases the complexity of the analysis, only pure binary intermetallics will be considered below. 4.2.2 Equilibrium dissolution temperature Based on equation (4-3) it is possible to calculate the equilibrium dissolution temperature Td of the precipitates. By rearranging this equation, we get:
D* T =-------d
C*
-log[%A]:[%B];
(4-4)
where [%A]o and [%B]o refer to the analytical content of elements A and B in the base metal, respecti vel y. Equation (4-4) shows that the equilibrium dissolution temperature increases with increasing concentrations of solute in the matrix. This is in agreement with the Le Chatelier's principle.
Table 4.1 Equilibrium solubility products for different types of precipitates in low-alloy steels and aluminium alloys. Data compiled from miscellaneous sources.
Materiall phase Type of Precipi tate TiN TiC NbN Low-alloy steel (austenite)'] NbC VN VC AIN Mo2C AI-Mg-Si:j: AI-Cu-Mg:j: Mg2Si CuMg MgZn AI-Zn-Mg:j: Zn2Mg C* log [%A]n [%B]m
= ~so/R'
0.32 5.33 4.04 2.26 3.02 6.72 1.79 5.0 5.85 6.64 5.33 7.72
D* =tJ!/R'
8000 10475 10230 6770 7840 9500 7184 7375 5010 4005 2985 4255
t
:j:
in wt% in at. %
304
Example (4.1)
Consider a low-alloy steel with the following chemical composition: 0.03wt%Nb, 0.04wt%AI, 0.006wt%Ti, 0.008wt%N Calculate on the basis of the reported solubility products in Table 4.1 the equilibrium dissolution temperature of each of the following three nitride precipitates, i.e. NbN, AIN, and TiN.
Solution
The equilibrium dissolution temperature of the precipitates can be computed from equation (44) by inserting the correct values for C* and D* from Table 4.1:
T (NbN)
d
T (AIN) =
d
T (TiN)
d
It is evident from these calculations that precipitates of the NbN and the AIN type will dissolve readily at temperatures above 1050 to 1100C, while TiN is thermodynamically stable up to about 1450C. In practice, however, a certain degree of superheating is always required to overcome the inherent kinetic barrier against dissolution, particularly if the heating rate is high. Consequently, in a real welding situation the actual dissolution temperature of the precipitates may be considerably higher than that inferred from simple thermodynamic calculations based on the solubility product (to be discussed later). 4.2.3 Stable and metastable solvus boundaries Due to the lack of adequate phase diagrams for the complex alloy systems involved, thermodynamic calculations based on the solubility product represent in many cases the only practical means of estimating the solid solubility of alloying elements in commercial low-alloy steels and aluminium alloys. 4.2.3.1 Equilibrium precipitates In the case of large, incoherent precipitates (where the Gibbs-Thomson effect can be neglected), the concentration of element A in equilibrium with pure An B m at different temperatures can be inferred directly from equation (4-3). If we replace R" by R (i.e. switch from common to natural logarithms), this equation yields:
[%A] = exp(Mo
[%B]mln
/ nR) ex (-
MfD
nRT
(4-5)
305
Equation (4-5) describes the solvus surface within the solvent-rich corner of the phase diagram. However, when a pure binary compound dissolves the concentration of elements A and B in solid solution is fixed by the stoichiometry of the reaction. The following relationship exists between [%B] and [%A]:
[wt%B]=[wt%B]
or
m MB
n MA
([wt%A]
0
-[wt%A1)
(4-6)
[at.%B]= [at.%BL -:
([at.%AL -[at.%A])
where MA and M B are the atomic weight of elements A and B, respectively. Figure 4.2 shows a graphical representation of equations (4-5) and (4-6), and the corresponding change in the matrix concentrations during dissolution of pure AnB m for a given set of starting conditions. Alternatively, we can express T as function of the product [%A]n [%B]m by utilising equations (4-3) and (4-6). The combination of these equations provides a mathematical description of the 'solvus boundary' of an equilibrium precipitate in a multi-component alloy system. It is evident from the graphical representation in Fig. 4.3 that the solid solubility will always in-
displacements
(equilibrium condi-
306
::s ~
Q)
c.. iQ)
Concentration of element A
Fig. 4.3. Factors affecting the solid solubility of a binary intermetallic compound in a multi-component alloy system (schematic).
crease with increasing temperature when f)Jl0 is positive. This type of behaviour is characteristic of intermetallics in metals and alloys, since the dissolution process in such systems is endothermic'? As a result, increased additions of a second alloying element B will also reduce the solubility of the first alloying element A by shifting the 'solvus boundary' towards higher temperatures when an intermetallic compound between A and B is formed. With the aid of Fig. 4.3 it is easy to verify that the equilibrium volume fraction of the precipitates I. at a fixed temperature is given by: eq.
feq.-fmax.
- (1- [%A]J
[%Alo
(4-7)
where fmax. is the maximum possible volume fraction precipitated at absolute zero. Equation (4-7) provides a basis for estimating the equilibrium volume fraction of binary intermetallics in complex alloy systems at different temperatures in cases where the concentration of element B is sufficiently high to tie-up all A in the form of precipitates. Similarly, if A is present in an overstoichiometric amount with respect to B, we may write:
i.; - f max.
(1
[%B] [%BL
(4-8)
PRECIPITATE
STABILITY
IN WELDS
307
Example (4.2)
In AI-Mg-Si alloys the equilibrium Mg2Si phase may form during prolonged high temperature annealing. Consider a pure ternary alloy which contains 0.75 wt% (0.83 at.%) Mg and 1.0 wt% (0.96 at. %) Si. Estimate on the basis of the solubility product the equilibrium volume fraction of Mg2Si at 400C. Make also a sketch of the Mg2Si solvus in a vertical section through the ternary AI-Mg-Si phase diagram. Relevant physical data for the AI-Mg-Si systemare given below:
MSi
MMg
MAL
28.1 g mol! 24.3 g mol! 27.0 g mol-l 76.7 g mol 2.7 g cnr' 2.0 g crrr '
!
MMg2Si
PAL
PMg2Si Solution
The maximum possible volume fraction of Mg2Si precipitated at absolute zero (fmax) can be estimated from a simple mass balance by considering the stoichiometry of the reaction:
f
max.
(%)""MMg2Si--.E.dL[wt%Mg]
nMMg PMg2Si
0
= 76.7 2.70.75%
2x24.32.0
=1.6% Moreover, the solubility product [at. % Mg]2 [at. % Si] at 400C (673K) can be obtained from equation (4-3) by utilising data from Table 4.1: log [at.% Mg]2 [at.% Si] =5.85 - 5010/673 from which [at.% Mg]2 [at.% Si]
= -1.59
= 2.55
X 10-2
If we also take into account the stoichiometry of the reaction, the solubility product can be expressed solely in terms of the Mg-concentration. Substituting [at.% Si]
[at.% Mg])
into the above equation gives [at.% Mg] ~ 0.20. The equilibrium volume fraction of Mg2Si at 400Cis thus: feq. =L; 1-[ [at.%Mg] ] at. %Mg 0
0.83
%=1.2%
Similarly, the equilibrium Mg2Si solvus can be calculated from the solubility product by substituting [at.% Si] = 0.96 - 1/2 (0.83 - [at.% Mg])
308
into equation (4-3). By inserting data from Table 4.1 and rearranging this equation, we get: T
eq.
CC)
- 273
It is seen from the graphical representation of the above equation in Fig. 4.4 that the Mg2Si compound is thermodynamically stable up to about 300C. At higher temperatures the phase will start to dissolve until the process is completed at 560C. It is obvious from these calculations that the microstructure of averaged AI-Mg-Si alloys should be very persistent to the heat of welding processes. In practice, only a narrow solutionised zone forms adjacent to the fusion boundary. However, within this zone significant strength recovery may occur after welding due to reprecipitation of hardening phases from the supersaturated solid solution. Consequently, in such weldments the ultimate HAZ strength level is usually higher than that of the base metal, as illustrated in Fig. 4.5. 4.2.3.2 Metastable precipitates In practice, the solid solubility is also affected by the size of the particles.
600
500
o o
~ ::s ~Q)
t-
g
'iii o
c:
c. ~ 300
o Co E
o
~ .Q
Cd
(ij
'E
z
o
c:
0
0
I
0.2 0.1
I
0.4 0.2
I
0.6
0.8
I
I
1.0
0.5
I
at% Mg
0.3
0.4
II
ats Mg Si
2
309
......
_<,
,.,y
\ \
\,V
\
spherical precipitate is acted on by an external pressure of say 1 atm, the same precipitate is also subjected to an extra pressure M due to the curvature of the particle/matrix interface, just as a soap bubble exerts an extra pressure on its content (see Fig. 4.6(a)). The pressure M is given as:" &=2y
(4-9)
where 'Y is the particle/matrix interfacial energy, and r is the radius of the precipitate. Because of this extra pressure, the Gibbs energy of a small precipitate will be higher than that of a large one, which, in turn, increases its solubility (see Fig. 4.6(b )). The important influence of particle curvature on the solid solubility has been extensively investigated and reported in the literature.l-" Usually, the phenomenon is referred to as the capillary or the Gibbs-Thompson effect. In the following we shall assume that the thermodynamic and crystallographic properties of the metastable precipitates are similar to those of the equilibrium phase and that the reduced thermal stability is only associated with capillary effects. For single phase precipitates in binary alloy systems, it is fairly simple to show that the concentration of solute across a curved interface, [%A]r' is interrelated to the equilibrium concentration of solute across a planar interface, [%A], through the following equation:"
[%A]
::::[%A]
exp( 2y V RTr
m )
= [%A]
exp(~) RT
(4-10)
where Vm is the molar volume of the precipitate (in m3 mol:"), and Q is the contribution of the interface curvature to the reaction enthalpy (equal to 2')' Vm/r).
310 (a)
METALLURGICAL
MODELLING
OF WELDING
Matrix
(b)
~ ~
t
Large precipitate
c:
CD
Q)
Ul .c
.c
Concentration
Fig. 4.6. Effect of interfacial energy on the solubility of small particles; (a) Schematic representation of spherical particles embedded in a metal matrix, (b) Integral molar Gibbs energy of matrix and precipitates at a constant temperature.
Assuming that this relationship also holds in the case of binary intermetallics, a combination of equations (4-5) and (4-10) gives:
[%A] r
-nQ)]
(4-11)
where
[wt%B]
rOn
= [wt%B]
- m MB ([wt%A]
MA
-[wt%A]
(4-12)
311
or [at. %B]r
= [at. %B]o
Alternatively, we can express T as a function of the product [%A]rn [%B]rm. This gives the following expression for the solvus temperature of metastable precipitates T' eq. :
T'
eq.
c: -log[%A]~
(4-13)
It is evident from the graphical representation of equation (4-13) in Fig. 4.7 that the solid solubility at a given temperature is significantly increased at small particle radii. Taking as an example v = 0.5 J m-2, Vm = 10-5 m3 mol:", R = 8.314 J K-I mol:", T= 500 K, we obtain from equation (4-10):
[%A]
= [%A] exp(~) RT
~ [%A](1
+~)
RT
or
[%A]r [%A] ~1+ 2xO.5xlO-5 109 =1+2.4 8.314 x 500(r) r
where r is the particle radius in nm. From this it is seen that quite large solubility differences can arise for particles in the range from r = 1 - SOnm.
t
.-"
O~O
)-0" ..--Q
- - - - - - -
312
METALLURGICAL
MODELLING
OF WELDING
Example (4.3)
In AI-Mg-Si alloys metastable (hardening) f3"(Mg2Si)-precipitates may form during artificial ageing in the temperature range from 160-200C. Consider a T6 heat treated ternary alloy which contains 0.75 wt% (0.83 at.%) Mg and 1.0 wt% (0.96 at.%) Si. Based on equation (413) make a sketch of the metastable f3"(Mg2Si) solvus in a vertical section through the ternary AI-Mg-Si phase diagram. In these calculations we shall assume that the thermodynamic properties of the metastable f3"(Mg2Si) phase are similar to those of the equilibrium rJ (Mg2Si) phase, i.e. the reduced thermal stability is only related to the Gibbs-Thomson effect. Relevant physical data for the AI-Mg-Si system are given below:
MMg2Si PMg2Si
'Y
r Solution
=--2-=--10-
MMgSi
PMg2 S. I
m mol- =3 84x10.
m mol-
n = 2yVm
r
The metastable f3"(Mg2Si) solvus can now be calculated from the solubility product by substituting
[at. %Si]r = 0.96 -1/2(0.83 - [at. % Mg]r)
into equation (4-13). By inserting data from Table 4.1 and rearranging this equation, we get:
T'
eq.
CC)=
- [at. %Mg]r]
It is evident from the graphical representation of the above equation in Fig. 4.8 that the particle curvature has a dramatic effect on the solid solubility. A comparison with Fig. 4.4 shows that the dissolution temperature drops from about 560C in the case of the equilibrium Mg2Si phase to approximately 225C for the metastable f3"(Mg.Sij-phase. On this basis it is not surprising to find that artificially aged (T6 heat treated) AI-Mg-Si alloys suffer from severe softening in the HAZ after welding, as shown schematically in Fig. 4.9. Moreover, it is
313
,
0
250
o 200
ca 'Q)
cD '::J
c. E Q)
150 100 50 0
....
0 0
0.2 0.1
0.4 0.2
0.6 0.3
0.8 0.4
1.0 0.5
at% Mg at% Mg Si
2
'~--""""'~After
__
artificial ageing
I I
..... ~--o#----After
natural ageing
I
I------~
~----Immediatelylafter
I I I
welding
~----------HAZ---------~
Fig. 4.9. Response of artificially aged AI-Mg-Si matic).
314
evident that the characteristic low dissolution temperature of the precipitates also gives rise to the formation of a heat affected zone which is significantly wider than that observed during welding of averaged AI-Mg-Si alloys." This shows that the response of age-hardenable aluminium alloys to welding and thermal processing depends strongly on the initial base metal temper condition. With the aid of equation (4-11) it is also possible to calculate an average (apparent) metastable solvus boundary enthalpy for hardening ~"(Mg2Si)-precipitates in AI-Mg-Si alloys. A closer evaluation of the exponent gives:
Q'
app.
~ 29kJrnol-1 This value is in close agreement with the reported solvus boundary enthalpy for ~"(Mg2Si)precipitates in 6082-T6 aluminium alloys.9,10
em
r3 - r3
(4-15)
315
where c1 is a kinetic constant, and Qs is the activation energy for the coarsening process (for binary intermetallics Qs may be taken equal to the activation energy for diffusion of the less mobile constituent atom of the precipitates in the matrix). 4.3.2.1 Kinetic strength of thermal cycle It follows that the extent of particle coarsening occurring during a weld thermal cycle can be calculated by integration of equation (4-15) between the limits t = tl and t = t2:
r3 - r~
= C1
f~
12 tl
(4-16)
The integral on the right-hand side of equation (4-16) represents the kinetic strength of the thermal cycle with respect to particle coarsening, and can be determined by means of numerical methods when the weld thermal (T -t) programme is known. The resulting radius of the precipitates may t~en be evaluated from equation (4-16) by inserting representative values for the constants Yo and c1 (e.g. obtained from quantitative particle measurements). 4.3.2.2 Model limitations A salient assumption in the classic Lifshitz- Wagner theory is that the particles coarsen at almost constant volume fraction, i.e. no solute is lost to the surrounding matrix during the coarsening process. Consequently, equation (4-16) should only be applied in cases where the peak temperature of the thermal cycle is well below the equilibrium solvus of the precipitates.
Example (4.4)
Consider stringer bead deposition (GMAW) on a thick plate of a Ti-rnicroalloyed steel under the following conditions: 1= 370A, U = 30V,
v
To = 20DC
Assume that the base metal contains a fine dispersion of TiN precipitates in the as-received condition. Calculate on the basis of equation (4-16) and the Rosenthal thick plate solution (equation (1-45) the extent of particle coarsening occurring within the fully transformed heat affected zone during welding. Relevant physical data for titanium-microalloyed steels are given below:
A
a
Yo
r:
:
c1 Qs
Solution
1520C 0.025 W mrrr ' C-I 5 mm? s' 6nm '6.67 X 1013 nm'' K s ' 240 kJ mol-l (activation energy for diffusion of Ti in austenite)
In the present example the problem is to calculate the size of the TiN precipitates in different
316
METALLURGICAL
MODELLING
OF WELDING
positions from the fusion boundary. This requires detailed information about the weld thermal programme, as shown in Fig. 4.10(a). By substituting the appropriate values for qo' A.,a and v into the Rosenthal thick plate solution, the governing heat flow equation becomes:
56532 T=20+-*-exp[-O.3(R
R
+x)]
where R* refers to the three-dimensional radius vector in the moving coordinate system (designated R in equation (1-45)), while x is the welding direction (equal to vt at pseudo-steady state). Since titanium nitride is thermodynamically stable up to the melting point of the steel, equation (4-16) can be used to calculate the extent of particle coarsening occurring within the transformed parts of the HAZ. In the present example, we may write:
1013 ~exp(
J
12
-288671 T)dt
where the times tl and t2 are defined in Fig. 4.10(a). The kinetic strength of the weld thermal cycle with respect to particle coarsening can now be evaluated from these two equations by utilising the numerical integration procedure shown in Fig. 4.10(b). The results from such computations are presented graphically in Fig. 4.11. It is evident from this figure that significant coarsening of the precipitates occurs within the HAZ during welding, particularly in regions close to the fusion boundary where the peak temperature of the thermal cycle is high. A comparison with the experimental data of Ion et al.14 (reproduced in Fig. 4.12) shows that the theory gives a fairly good prediction of particle size as a function of the peak temperature, provided that the kinetic constant c 1 in equation (416) can be estimated with a reasonable degree of accuracy. In practice, however, the numerical value of c 1 will vary significantly with the chemical composition and thermal history of the base metal (see Fig. 4.13). This means that empirical calibration of equation (4-16) to experimental data is always required to avoid systematic deviations between theory and experiments.
317
(a)
~00C
I
I
me
(b)
Fig. 4.10. Kinetic strength of weld thermal cycle with respect to particle coarsening (Example (4.4)); (a) HAZ temperature-time programme (schematic), (b) Numerical integration procedure (schematic).
318
30
I I
t
c:
''-
,
1
25
N
1
I I I I
E 20
::r:
-0
Q)
(j)
E 0
s::
:S
I
I
Q)
1-0
jg
lQ5
I
:~
I 1 I
ro = 6 nm 1---....-4-.:::0:::._:=;;;._._._._._._._._._._._._._._._.,._.
I I 1 I I I
1000
1200
1400
1600
50
r-------I I
I I
Ata/s= 155 (E
=-------1
l-
I I I I I I I I I I I
I
--------------------10
15
10
I I
20 .
25
Particle radius, nm
Fig. 4.12. Measured size distribution of TiN before (broken lines) and after (full lines) weld thermal simulation. Operational conditions as in Example (4.4). Data from Ion et al.14
ing precipitates. Nevertheless, it will be shown below that at least some of them are sufficiently accurate to capture the essential physics of the problem and to give valuable quantitative information on the extent of particle dissolution occurring during the weld thermal cycle.
319
1
E
c:
100 80 60 40
W ::l :0
Q)
13
<U a.
:e
20 Steel A o Steel B I:J. Steel C
Annealing time, S ~
10
Fig. 4.13. Effects of annealing time and steel chemical composition on the mean particle size of TiN. Data from Matsuda and Okumura.P
4.4.1.1 The invariant size approximation The model described here is due to Whelan.17 Consider a spherical particle embedded in an infinite matrix, as shown schematically in Fig. 4.14. The corresponding matrix concentration profile is plotted in the lower part of the figure. In this case the concentration of the constituent element A is higher close to the particle/matrix interface than in the bulk. Hence, there is a tendency for the element to diffuse away from the particle and into the surrounding matrix (i.e. the particle dissolves). Based on the assumption that the particle/matrix interface is stationary (i.e. the diffusion field has no memory of the past position of the interface), Whelan17 arrived at the following expression for the dissolution rate of a spherical precipitate at a constant temperature:
(4-17)
where r is the particle radius, a is the dimensionless supersaturation (defined in Fig. 4.14), and D m is the element bulk diffusivity. The term l/r on the right-hand side of equation (4-17) stems from the steady-state part of the diffusion field, while the (ItVt) term arises from the transient pact, Because of the complex form of equation (4-17) it cannot be integrated analytically and hence, numerical methods must be applied. However, if the transient part of equation (4-17) is neglected (conforming to the solution after long times), it is possible to obtain a simple expression for the particle radius as a function of time:
(4-18)
where r0 is the initial particle radius. Equation (4-18) is identical with the so-called invariant-field solution developed independ-
320
c: ,2
Q)
e 1::
(l=
c. I
C - C.
m
I
<.J c:
Distance
Fig. 4.14. Schematic representation of the concentration profile around a dissolving spherical particle in an infinite matrix.
ently by Aaron et al.16 and is valid after a certain period of time, provided that there is no impingement of diffusion fields from neighbouring precipitates. As shown in Fig. 4.15, this simplified solution gives a reasonable description of the dissolution kinetics of small spherical precipitates in steel during reheating above the Acl-temperature. Following the treatment of Agren," the time required for complete dissolution of a spherical precipitate td can be obtained from equation (4-18) by setting r = 0:
(r )2 t = _0 d 2aDm
(4-19)
f as a function
2aDm t
( J
}3/2
(4-20)
where 10 is the initial particle volume fraction. The former equation shows that the dissolution time cle size roo
Example (4.5)
td
The following example illustrates the direct application of equations (4-18) and (4-19). Consider a niobium-microalloyed steel which contains a fine dispersion of NbC precipitates. Provided that impingement of diffusion fields from neighbouring particles can be neglected, calculate the total time required for complete dissolution of a 100 nm large NbC precipitate at
PRECIPITATE
STABILITY
IN WELDS
321
ti.4
~ en ::J :c ~
Q)
E 1.2 1.0
13 O.B
S a.
Time,s ----
Fig. 4.15. Dissolution kinetics of spherical cementite particles in austenite at 850C. Data from Agren.18
1350C. Data for the steel chemical composition and the diffusivity of Nb in austenite at 1350C are given below: Steel chemical composition: 0.12 wt% C, 0.05 wt% Nb Diffusivity of Nb in austenite at 1350C:
In the present example it is reasonable to assume that the dissolution rate of the precipitate is controlled by diffusion of Nb in austenite. For a single NbC precipitate embedded in a Nbdepleted matrix, the dimensionless supersaturation becomes:
ex
Nb
Ci(Nb) :::;:--p
C (Nb)
322
from the solubility product by utilising data from Table 4.1. If we assume that the carbon concentration at the interface is constant and equal to the nominal value of 0.12 wt% (i.e. the stoichiometry of the dissolution reaction is neglected), equation (4-5) reduces to:
C.(Nb) ~ exp(43.26/8.314) [0.12]
I
ex ( p
129578) 8.314x1623
~ O.IOwt%
92.9+12.0
0.10 ~--=1.1 88.6
92.9
100wt%=88.6wt%
This gives:
3 xIO-'
Nb
The dissolution time td can now be. calculated from equation (4-19) by inserting the appropriate values for ro' Ci..Nb' and DNb:
(100 X 10-6)2 2(1.13 x 10-3)(5.39 x 10-7)
S
= 8.2s
A comparison with Fig. 4.16 shows that the predicted value is off by a factor of about 4 compared with that obtained from more sophisticated numerical calculations. This degree of accuracy is acceptable and justifies the use of equation (4-18) for prediction of the dissolution rate of spherical precipitates under different thermal conditions provided that the model is calibrated against experimental data points. 4.4.1.2 Application to continuous heating and cooling Application of the model to continuous heating and cooling requires numerical integration of equation (4-18) over the weld thermal cycle:
t..,
r2 = (r
)2 -2Jo
o m dt
(4-21)
Under such conditions the volume fraction of the precipitates is given by:
(4-22)
Equations (4-21) and (4-22) provide a basis for predicting the extent of particle dissolution occurring within the HAZ during welding in the absence of impingement of diffusion fields from neighbouring precipitates.
323
10
1350C
(/)
1
(/) (/)
;:;
cD E
1 c .Q 10. "S (5
Q 10.2
10
.. ~
Fig. 4.16. The dissolution time of NbC in austenite at 1350C as function of initial particle diameter 2ro for different Nb and C levels (numerical solution). Data from Suzuki et al. 6 Example (4.6)
Consider stringer bead deposition (SAW) on a thick plate of a Nb-microalloyed wt% C - 0.03 wt% Nb) under the following conditions: I
steel (0.10
= 450A,
To
= 20C
Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received condition. Calculate on the basis of equation (4-22) and the Rosenthal thick plate solution (equation (1-45 the extent of particle dissolution occurring within the fully transformed HAZ during welding. Relevant physical data for Nb-microalloyed steels are given below: 1520C 0.025 W mrrr ' C-I 5 mrrr' S-l 910C 60nm
324 DNb(mm2
MNb
S-I)
343000)
Me
Solution
In the present example the problem is to calculate the variation in the fifo ratio across the fully transformed HAZ. By substituting the appropriate values for qo' A, Q, and v into the Rosenthal thick plate solution, the governing heat flow equation becomes:
81646 T( C) = 20+-*-exp[-0.5(R R
o
+x)]
Since it is reasonable to assume that the dissolution rate of the precipitates is controlled by diffusion of Nb in austenite, the dimensionless supersaturation reduces to: a
Nb
z---
C/Nb) Cp (Nb)
As shown in example (4.5), the equilibrium concentration of niobium at the particle/matrix interface (in wt%) can be estimated from the solubility product by utilising data from Table 4.1. If we assume that the carbon concentration at the interface is constant and equal to the nominal value of 0.10 wt%, equation (4-5) becomes:
C. (Nb)
I
= 181gexp( _12!~78)
This gives:
a
Nb z
= 20.5exp(-
129578) RT
DNb
Ci
Nb
and
325
Here the lower and upper integration limits refer to the total time spent in the thermal cycle from AC3 to Tp and down again to Ac3 The extent of particle dissolution occurring within the HAZ during welding can now be calculated in an iterative manner by numerical integration of the above equation over the weld thermal cycle. The results from such computations are presented graphically in Fig. 4.17. It is evident from these data that NbC starts to dissolve when the peak temperature of the thermal cycle Tp exceeds the equilibrium dissolution temperature Td of the precipitate. The process is completed when Tp approaches 1330C, conforming to a temperature interval of 190C. This shows that considerable superheating is required in order to overcome the inherent kinetic barrier against particle dissolution under the prevailing circumstances. 4.4.2 Numerical solution In the previous treatment, no consideration is given to impingement of diffusion fields from neighbouring precipitates or the position of the particle/matrix interface during the dissolution process. In certain cases, however, such phenomena will have a marked effect on the dissolution kinetics. 18-22 A good example is AI-Mg-Si alloys where the hardening J3"(Mg2Si)-phase forms a very fine distribution of needle-shaped precipitates along < 100> directions in the aluminium matrix. These precipitates are closely spaced and will therefore interact strongly with each other during dissolution (coupled reversion).
1.0 0.8
__ -----~L
o o
o
II
i i i j o o
o
----
No dissolution-
-i. -'I
i i
0;
C't)
C)
i i - - -j-
I
Complete olssolunon - - - - -------
..
800
900
1000
1100
1200
1300
1400
150
Peak temperature, C ~ Fig. 4.17. Dissolution of NbC during steel welding (Example (4.6.
326
4.4.2.1 Two-dimensional diffusion model For rod or needle-shaped precipitates in a finite, depleted matrix, the rate of dissolution can be calculated by numerical methods from a simplified two-dimensional diffusion model. Assuming that the precipitates are mainly aligned in one crystallographic direction, it is reasonable to approximate their distribution by that of a face-centered cubic (close-packed) space lattice, as shown in Fig. 4.18(a). If planes are placed midway between the nearest-neighbour particles, they enclose each particle in a separate cell. Since symmetry demands that the net flux of solute through the cell boundaries is zero, the dissolution zone is approximately defined by an inscribed cylinder whose volume is equivalent to that of the hexagonal cell. The modelling principles outlined in Fig. 4.18(a) and (b) have previously been used by a number of other investigators to describe particle dissolution during isothermal heat treatment.18-22 Consequently, readers who are unfamiliar with the concept should consult the original papers for further details. It follows from Fig. 4.18(b) that the rate of reversion can be reported as: (4-23)
(b)
~2r~
c:
~
I I
: ,...... -- .... :
I I
t--C
E '"E
Q)
(J
o t:>t--~
Distance Fig. 4.18. Numerical model for dissolution of rod-shaped particles in a finite, depleted matrix; (a) Dissolution cell geometry, (b) Particle/matrix concentration profile (moving boundary).
c:
327
For a specific alloy, the ratio between r 0 and L (the mean interparticle spacing) can be calculated from a simple mass balance, assuming that all solute is tied-up in precipitates. Taking this ratio equal to 0.06 for rod-shaped precipitates in diluted alloys," the kinetics of particle dissolution during isothermal heat treatment have been examined for a wide range of operational conditions. These results are presented in a general form in Fig. 4.19 by the use of the following groups of dimensionless parameters:
Dimension ess time
I.
't = --2
Dmt
(ro)
(4-24)
(4-25)
The data in Fig. 4.19 suggest that the reaction kinetics during the initial stage of the process are approximately described by the relation: (4-26) where c2 is a kinetic constant, and n1 is a time exponent (assumed constant and equal to 0.5 under the prevailing circumstances). The rate of particle dissolution will gradually decline with increasing values of r as a result of impingement of diffusion fields from neighbouring precipitates which reduces a. In practice, this is seen as a continuous decrease in the slope ofthe!lfo-'t curves in Fig. 4.19 (nl < 0.5). In such cases equation (4-26) will only be valid within small increments of r.
t
C)
-0.2
:::.. -0.4
o
-0.6
.. 0.8 ""'-----------'------
~2
o
log
....-..--..1..------------1
2
't
Fig. 4.19. Dissolution kinetics of rod-shaped particles in a finite, depleted matrix. Data from Myhr and Grong.?
328
4.4.2.2 Generic model Myhr and Grong? have shown how this model can be applied to specific alloy systems. From equation (4-26) we have:
(4-27) where n1 ~ 0.5. This equation can further be simplified if we assume that C. C 1
and Cp C.:
(4-28)
For isothermal heat treatment at a chosen reference temperature (Trl) , the rate of particle dissolution is determined by the retention time t rl. Let t;l denote the maximum hold time required for complete dissolution of the precipitates. It follows that equation (4-28) can be written in a general form by normalising t rl with respect to t;l. The parameter t;l is obtained by settingf/io = 0:
tr* =
]
c3 D m (T) r]
]-1[ c. (T)
1
]-lIn
r]
(4-29)
where c3 is a new kinetic constant. If heat treatment is carried out at a different temperature (T ~ T rl) t 1* is simply given by:
(4-30)
By inserting the approximate expressions for C, and D m into equation (4-30) (see previous examples), and rearranging equation (4-28), we obtain:
t * =t * exp [(
I
r]
--+Q~PP.
nR
1
o, J( --1 1 Jl R T T
~
(4-31)
and Qs is the activation energy for diffusion of the less mobile constitutive atom of the precipitates.
329
Equations (4-31) and (4-32) exploit some good modelling techniques. For example, the use of a dimensionless time eliminates an unknown kinetic constant which premultiplies t and t~ in the derivation of equation (4-32). Moreover, by raising the dimensionless time to a power n1 means that the premultiplying constant, here unity, is independent of the value of n1, and is itself also dimensionless. Finally, the form of equation (4-31) eliminates further unknown kinetic constants, and may readily be calibrated using an experimental time t;1 at a reference temperature. Figure 4.20 shows the variation inflfo with time (on log axes), from a range of isothermal experiments carried out on 6082- T6 aluminium alloys, using hardness (or electrical conductivity) measurements to evaluateflfo' The curve (equation (4-32)) extrapolates back to a slope of 0.5 (the exponent n1) for the case of the early stages of dissolution before impingement of the diffusion fields. The exponent n1 is seen to fall to lower values when the proportion dissolved is higher, in agreement with the theoretical curves in Fig. 4.19. 4.4.2.3 Application to continuous heating and cooling Myhr and Grong" have also shown how this model can be applied to situations where the temperature varies with time (as in welding). In order to obtain a general description of particle dissolution under non-isothermal conditions, it is convenient to introduce the related concepts of an isokinetic reaction and the kinetic strength of a thermal cycle.P A reaction is said to be isokinetic if the increments of transformation in infinitesimal isothermal time steps are additive. Christiarr'" defines this mathematically by stating that a reaction is isokinetic if the evolution equation for some state variable X may be written in the form:
Cl
AT T
-0.4
= 375C
*T=275C T = 250C
.Q
-1.2
-1.6
-8
-6
-2
o
in 6082-T6 aluminium
Fig. 4.20. 'Master-curve' for dissolution of hardening (j"(Mg2Si)-precipitates alloys. Data from Myhr and Grong.?
330
dt
H(T)
(4-33)
where G(X) and H(T) are arbitrary functions of X and T, respectively. For a given thermal history Ttt), this essentially means that the differential equation contains separable variables of X and T. The same criterion may also be applied to the models derived above. In the case of coupled reversion, we may write: (4-34) and (4-35) Since dfldl, and t; are unique functions ofJand T, respectively, the additivity condition is satisfied. Consequently, when the temperature varies with time, we replace the term tit; in equation (4-34) by dt I t; and integrate over the thermal cycle, giving:
I =
1
f
t o
dt = (1t*
1
f /f
)11 nl
0
(4-36)
This integral is called the kinetic strength of the thermal cycle with respect to reversion. The resulting volume fraction of the precipitates following a heating cycle is then found by evaluating the integral II numerically (e.g. by utilising input data from Table 4.2) and replacing tit; with II in equation (4-32), yielding a value for/lfo from the master curve of Fig. 4.20. Case Study (4.1) By utilising equation (4-36) and the general heat flow model for welding on medium thick plates (i.e. equation (1-104, it is possible to calculate the variation in the fifo ratio (i.e. the solute distribution) across the HAZ of single pass 6082- T6 aluminium weldments for a wide range of operational conditions (see Table 4.3). The results from such computations are presented graphically in Fig. 4.21 -4.23. When stringer bead deposition is carried out on a plate of medium thickness, the solute distribution in the transverse y direction is expected to vary with distance from the plate surface due to a continuous change in the heat flow conditions. A closer inspection of Figs. 4.21 and 4.22 shows that this is correctly accounted for in the present model. In contrast, a full penetration butt weld will always reveal a similar solute distribution in the transverse direction of the weld, as shown in Fig. 4.23. This situation arises from the lack of a temperature gradient in the through-thickness z direction of the plate.
Table 4.2 Basic input data in dissolution model for hardening (3" (Mg2Si)- precipitates in 6082- T6 aluminium alloys. Data from Myhr and Grong.?
Parameter
:
30kJ mer:'
Qs
130 kJ mol '
tr}
(375C) 600 s
Value
(a)
1.0
-.
f
o
0 ~
t
s
0
400
....
')'-
-----~m~~ru~~@Q-----...
IJ ., .,
300
200 100
e
Q)
~ ::I
-a a,
CD
.m
0.
10
20 30 ~ ~Ym,mm
40
50
(b)
1.0
600
f
_0
0.8 0.6
.
<.~.
500
400
f
o
0
:;:::
.2 0.2 300
Q)
PI
a.
____
C..2l]P~t~<i(s~o~tlgn
.... --o.
____
E .m
.
Y
200
100
7a CD
a..
10
20
30 Ymtmm
----
40
Fig. 4.21. Dissolution of (3"(Mg2Si)-precipitates during aluminium welding (Weld 1); (a) Upper plate surface, (b) Lower plate surface. Operational conditions as in Table 4.3. Table 4.3 Operational conditions used in aluminium welding experiments (Case study 4.1).
Weld No.* 1 2 3 Material Plate thickness (mm) 15 15 13 Net arc power (kW) 9.1 5.7 14.0 Welding speed (mm S-I) 4.2 9.1 5.8
* Calibration of heat flow model is done by including an empirical correction for heat consumed in melting of the parent material (thermal data for the AA6082 alloy are given in Table 1.1, Chapter 1).
, .0
~------~--~--~------~------~600
500
400
o
0
!
~ ~
~
:;::
.~
____
:J ... 300
..0
~Qo~e!ele~i~~u.!iO!!.
...
s
200 ~ Q)
c,
100 0
10 llYm,mm
15
20
(b)
1.0
600 500
:::::
..2
______
.)'
,0.
to
!
0
400
~ 300 c.
Q)
...
~ :.1
E
s
200 ~ Q)
n,
10 ym,mm
15
Fig. 4.22. Dissolution of J3"(Mg2Si)-precipitates during aluminium welding (Weld 2); (a) Upper plate surface, (b) Centre of plate. Operational conditions as in Table 4.3.
4.4.2.4 Process diagrams for single pass 60B2-T6 butt welds For single pass butt welding of plates, the heat flow model (equation 1-104 can largely be simplified if the net arc power is kept sufficiently high compared with the plate thickness (i.e, qjd>O.5 kW mm-'), Under such conditions the mode of heat flow becomes essentially onedimensional, and the temperature distribution is determined by the ratio qjvd, kJ mm -2 (see Sections 1.10.3.3 and 1.10.4.1 in Chapter 1).
333
I
S
o
400
o ~ ~
(1)
.~
,0
300 ~
----~~~~~~~~~-----.......
E 2
200
100
td CD
o,
~----------~----~------~----~----~o o ~ ~ 00
.6.Ym,mm
Fig. 4.23. Dissolution of ~"(Mg2Si)-precipitates ditions as in Table 4.3. during aluminium welding (Weld 3). Operational con-
Figure 4.24 shows plots of the variation in the fifo ratio across the HAZ of 6082- T6 aluminium weldments for different values of qfvd. It follows that a narrow width of the dissolution zone requires the use of a low energy per mm/ of the weld. In practice, this can be achieved by the use of an efficient welding process (e.g. electron beam or laser welding) which facilitates deposition off a full penetration butt weld without employing a groove preparation (i.e. eliminates the need for filler metals).
0.8
0.6
Scale: ~Omml Distance from fusion line ~ Fig. 4.24. Process diagram showing the solute distribution within the HAZ of single-pass 6082- T6 aluminium butt welds for different values of q 0 Ivd.
334
References
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. R.D. Doherty: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and P. Haasen), 1983, Amsterdam, North-Holland Physics Publ., 934-1030. K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths & Co., Ltd. H. Adrian and F.B. Pickering: Mater. Sci. Techno!., 1991,7,176-182. B. Loberg, A. Nordgren, J. Strid and K.E. Easterling: Me ta ll. Trans., 1984, 15A, 33-41. J. Strid and K.E. Easterling: Acta Metall., 1985,33,2057-2074. S. Suzuki, G.C. Weatherly and D.C. Houghton: Acta. Metall., 1987,35,341-352. J.L. Petty-Galis and R.D. Goolsby: J. Mater. Sci., 1989,24, 1439-1446. D.A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham (England), Van Nostrand Reinhold Co. Ltd. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1991,39,2693-2702; ibid., 2703-2708. H.R. Shercliff and M.F. Ashby: Acta Metall. Mater., 1990,38, 1789-1802; ibid., 1803-1812. J.M. Lifshitz and V.V. Slyozov: J. Phys. Chern. Solids, 1961,19,35-50. C. Wagner: Z. Electrochem., 1961,65, 581-591. L.C. Brown: Acta Metall. Mater., 1992,40, 1293-1303. J.C. Ion, K.E. Easterling and M.E Ashby: Acta Metall., 1984,32, 1949-1962. S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18,198-205. H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970,41,4404-4410. M.J. Whelan: Metal Sci. J., 1969,3, 95-97. J. Agren: Scand. J. Metall., 1990, 19,2-8. R.A. Tanzilli and R.W. Heckel: Trans. Met. Soc. AIME, 1968,242,2313-2321. H.B. Aaron and R. Kotler: Me ta ll. Trans., 1971,2,393-408. R. Asthana and S.K. Pabi: Mat. Sci. Eng., 1990, A128, 253-258. U.H. Tundal and N. Ryum: Metall. Trans., 1992, 23A, 433-444; ibid., 445-449. H.R. Shercliff, 0. Grong, O.R. Myhr and M.F. Ashby: Proc. 3rd Int. Conf. on Aluminium Alloys - Their Physical and Mechanical Properties, Trondheim, Norway, June 1992, Vol. III, pp. 357-369, The University of Trondheim, The Norwegian Inst. of Technol. J.W. Christian: Phase Transformations in Metals and Alloys, 1975, Oxford, Pergamon Press.
24.
S-I)
[%A]r
335
standard free energy of reaction (J mol-lor kJ mol ') gas metal arc welding arbitrary function of T standard enthalpy of reaction (J mol-lor kJ mor ') amperage (A) kinetic strength of thermal cycle with respect to reversion equilibrium constant mean interparticle spacing in dis-solution model (m, urn or nm) integer atomic weight of element A (g mol:") atomic weight of binary precipitate (g mol ")
B [%B]
arbitrary alloying element equilibrium concentration of element B in matrix (wt% or at.%) analytical content of element B in alloy (wt% or at. %) equilibrium concentration of element B across a curved interface particle/matrix (wt% or at.%) various kinetic constants and temperature-dependent parameters normalised (dimensionless) entropy of reaction concentration of solute at interface particle/matrix (wt% or at. %) nominal alloy composition (wt% or at.%) concentration of solute in matrix (mol nr ', wt% or at.%) concentration of solute inside the precipitates (wt% or at.%) plate thickness (mm) normalised enthalpy of reaction (K) element diffusivity (mm- s' or m2 s') particle volume fraction initial particle volume fraction equilibrium particle volume fraction maximum particle volume fraction at absolute zero arbitrary function of X
scr
GMAW H(n flHO
[%B]o
[%B]r
I II
cIc3
K L
C*
Ci
m
MA
Co Cm
MAnBm
MB Cp
atomic weight of element B (g mol ') integer time exponent in dissolution model pressure caused by curvature effects (J rrr-') net arc power (W) apparent (metastable) solvus boundary enthalpy (kJ mol:') activation energy for diffusion (kJ mol ') particle radius (m, urn or nm) initial particle radius (m, urn ornm)
n ni M
d D*
Dm
qo
;
Qs
t, t.;
fmax.
G(X)
r ro
336
R
uni versal gas constant (8.314J K-I mor ') universal gas constant multiplied by InlO (19.14J K-I
T6
R'
u
v x
S-I)
mol ")
welding direction (mm) state variable transverse direction (mm) width of HAZ referred to fusion boundary (mm)
R*
three-dimensional radius vector in Rosenthal equation (mm) standard entropy of reaction (J K mor ')
x
y
SAW
submerged arc welding time (s) time necessary for complete dissolution of precipitate(s)
through-thickness (mm)
direction
(m3 mol:")
tl ' t2
integration limits (s) time necessary for complete reversion at T( s) retention time at T rl (s) time necessary for complete reversion at T, 1(s)
molar volume of precipitate per mole of the diffusate (m3 mol ") density of precipitate An Bm
(g
cm" or kg m=')
conductivity (W
thermal
mm-Ioe-I)
dimensionless supersaturation in dissolution model density (g cnr' or kg nr ') arc efficiency factor particle/matrix energy (J m-2) interfacial
melting point (Oe or K) peak temperature of thermal cycle r-c or K) chosen reference temperature in dissolution model (Oe or
K)
dimensionless time in dissolution model contribution of particle curvature to reaction enthalpy (J mol-lor kJ mol+) equilibrium phase in AI-MgSi alloys hardening precipitate in AIMg-Si alloys
Teq.
T' eq.
5.1 Introduction
Grain growth is an important aspect of welding metallurgy. Normal grain growth in metals and alloys is a thermally activated process driven by the reduction in the grain boundary energy. Physically, it occurs by growth of the larger grains at the expense of the smaller ones which tend to shrink. Under isothermal heat treatment conditions, normal grain growth is well described by the following empirical equation: 1
15l1n _15l1n o
(5-1)
where 15 is the mean grain size (diameter), 15 0 is the initial grain size, n is the time exponent, t is the isothermal annealing time, Qapp. is the apparent activation energy for grain growth, and Cl is a kinetic constant. The other symbols have their usual meaning. For most metals and alloys the time exponent n in equation (5-1) varies typically in the range from 0.1 to 0.4, as shown in Fig. 5.1. Only in the case of ultrapure metals annealed at very high temperatures the time exponent may approach a constant value of 0.5. This corresponds to the limiting case where the grain boundary migration rate is directly proportional to the driving pressure 'YID ('Ydenotes the grain boundary interfacial energy per unit area, while D is the grain size). It is well recognised that alloying and impurity elements both in the dissolved state and in the form of inclusions or second phase particles will retard grain growth.v" Consequently, a comprehensive theoretical treatment of grain growth in welds must include a consideration of such effects. The present analysis will therefore start with a closer examination of factors affecting the grain boundary mobility in metals and alloys under conditions applicable to welding.
A critical aspect of modelling grain growth is the quantitative description of grain structures, which is essential in making a comparison between theoretical predictions and experimental
338
Z.R.Sn
Zr -IJ-IJ-O-II0.1
rrJ
Fig. 5.1. Temperature-dependence of the time exponent in isothermal grain growth (Z.R.: zone refined, H.P.: high purity). Data compiled by Hu and Rath.l
measurements. Different parameters are used to describe the size and the shape of individual grains. In three dimensions, individual grain volumes cannot be determined directly from measurements made in single cross sections through the structure. Therefore, certain geometric assumptions must be employed to obtain these quantities. Since most grain size measurements seek to correlate the interaction of grain boundaries with specific properties (e.g. the transformation behaviour), an estimate of the grain boundary area per unit volume S; is often required. This parameter can be calculated without assumptions concerning grain shape and size distribution from measurements of the mean linear grain intercept D *.5 (5-2) If the mean grain diameter D is required from Sv, this may be obtained by assuming a spherical grain shape. Noting that each boundary is shared by two adjacent grains, we obtain:
2S = 41t(D 12)
v
i1t(DI2)3
3
(5-3)
from which (5-4) It follows that the mean linear grain intercept D* is always smaller than the actual grain size A common observation in metals and alloys is that the size distributions of grain aggregates
D.
GRAIN
GROWTH
IN WELDS
339
at different annealing times become equivalent when the measured grain size parameter, D, is normalised (scaled) by the time-dependent average of this metric, D (see Fig. 5.2). This means that grain structures can be completely characterised, in a statistical sense, by simple probability functions of the standard deviation of the distribution together with the time dependence of the average size scale D. 5.2.2 Driving pressure for grain growth The thermodynamic driving pressure for motion of a spherically curved element of grain boundary is given by: 7
PG _ 2y
-
(5-5)
where 'Yis the grain boundary interfacial energy, and p * is the radius of curvature. It is conventional practice to replace p * in equation (5-5) with some measure of the average grain size, such as the mean linear grain intercept D*, or with the diameter of the equivalent spherical volume of some geometrically modelled average grain size. Experimental measurements performed on high purity aluminium indicate that p" ~ 3.23 D*J This observation is consistent with the model of Hellman and Hillert,8 which predicts that the curvature of the most critical element of grain boundary that must be stabilised is p * = 3 D. Under such conditions, the driving pressure becomes:
p
G
='!:..l 3D
(5-6)
12
Zone refined iron
10
:' c
:J
1
~
1-
25 min
8 Rayleigh distribution
ca ... :c ~
u,
0.5
1.0
2.5
3.0
Fig. 5.2. Comparison of measured grain size data in iron with the Rayleigh and log-normal distribution functions F. The similarity of the size distributions at different annealing times illustrates the self-similar scaling behaviour of normal grain growth. After Pande."
340
METALLURGICAL
MODELLING
OF WELDING
In practice, the numerical constant in equation (5-6) can vary by, at least, a factor of three, depending on the assumptions of the models. Consequently, in the general case the average driving pressure is given by: (5-7) where
C2
is a constant which is characteristic of the system under consideration. 5.2.3 Drag from impurity elements in solid solution
In the dissolved state impurity and alloying elements will retard grain growth through elastic attraction of the atoms towards the open structure of the grain boundary. The boundary must then either drag the impurity atoms along (so that its speed is limited by the diffusion rate of these atoms) or break away if the impurity concentration is sufficiently small or the driving pressure or temperature is high enough. An analysis of such effects can be done on the basis of the classic impurity drag theories of Chan/ and Lucke and Stuwe," which deal with the following two extreme cases: (i) A low velocity limit, where the rate of grain boundary migration is controlled by diffusion of impurity atoms perpendicular to the boundary. (ii) A high velocity limit, where the grain boundary migration process is mainly governed by the diffusion of solvent atoms across the boundary (i.e. controlled by the rate of boundary self diffusion). The low velocity limit is associated with either a low driving pressure P G or a high impurity level Co, and is characterised by a linear type of relationship between the grain boundary migration rate Va and P G:2
V
a
p
E
'Peo
(5-8)
Here e denotes the intrinsic drag coefficient, while \}I is a parameter depending on the diffusivity and the interaction energy between the grain boundary and the impurity atoms. For the other extreme (i.e. the high velocity limit), the grain boundary migration rate Vb is described by the relationship." (5-9) where \}I /~2 denotes another complex function of the impurity diffusivity and the interaction energy between the grain boundary and the impurity atoms. In the case of spherical grains, the classic grain growth equation predicts that the grain boundary migration rate V is a power function of the driving pressure ("I / D). This is readily seen by differentiating equation (5-1) and inserting "I (the grain boundary interfacial energy) into the resulting expression:
341
D)(1ln-l)
v =1. dD
2 dt
=!!.c
2
3
exp(-
Qapp.
RT
J(1/
(5-10)
where c3 is a new kinetic constant. Note that at very high driving pressures or low impurity levels, equation (5-9) approaches the limiting case where the migration rate Vb becomes directly proportional to PG, corresponding to a time exponent n = 0.5 in the grain growth equation. At this point the grain boundary will break away from its surrounding impurity atmosphere and migrate at a rate close to the rate of boundary self diffusion. In most cases, however, the relationship between V and P G derived from equation (5-10) will be different from the theoretical one due to the empirical nature of the grain growth equation. For high driving pressures and intermediate impurity concentrations, the classic impurity drag theories predict a discontinuous transition from the high to the low velocity limit. It has, however, been argued by Vandermeer? that the observed transition may be considered as continuous. Thus, in a log V vs log P G plot it would appear as a steep curve connecting the lines for the high and low velocity extremes together, as illustrated schematically in Fig. 5.3. 5.2.4 Drag from a random particle distribution The retardation of grain growth by second phase particles was first theoretically investigated by Zener.lO There seems to be general agreement that the maximum pinnig force Fp exerted by a single particle of radius r on a grain boundary is given by:4,lo Fp
'IT,,!
(5-11)
1
Transition region Low velocity limit
Fig. 5.3. Schematic variation of grain boundary migration rate V with driving pressure P G according to the classic impurity drag theories of Cahn? and LUcke and Stuwe.' The diagram is based on the ideas of Vandermeer."
342
If the number of interacting particles per unit area of the grain boundary is taken equal to na, the resulting retardation pressure becomes:
Pz
'Tr"Y
rna
(5-12)
Assuming that only one half of the particles which touches the grain boundary will interact with a maximum force, na is related to Nv (the number of particles per unit volume) through the following equation.v'" (5-13) Given that the particles are spherical and of uniform size, N; can be expressed as: (5-14) where f is the particle volume fraction. A combination of equations (5-12), (5-13), and (5-14) leads to the well-known expression for the so-called Zener drag (or Zener retardation pressurej.v'?
p
=lyl
4
(5-15)
In practice, the numerical constant in equation (5-:-15)can vary by, at least, a factor of five, depending on the assumptions of the models.'! Consequently, in the general case the Zener retardation pressure is given by:
P =c
Z 4
yl r
(5-16)
where
C4
is a constant which is characteristic of the system under consideration. 5.2.5 Combined effect of impurities and particles
As already stated in the introduction of this chapter, the time exponent n is a measure of the resistance to grain boundary motion in the presence of impurity and alloying elements in solid solution. Based on equation (5-1) Hu and Rathl,12 have shown that the grain boundary migration rate V is related to the effective driving pressure M G and the time exponent n through the following equation:
v = M[M> G JOIn-I)
(5-17)
where M is the grain boundary mobility. In alloys containing grain boundary pinning precipitates, the effective driving pressure M G is defined as the numerical difference between P G and Pz. By inserting the correct expressions for P G and Pz into equation (5-17), we obtain:
343
M(c
2
y)(1/n-l) [
_ C4
C 2
L
r
(5-18)
It follows from equation (5-18) that the grain boundary migration rate Vbecomes inversely proportional to the average grain size D when n = 0.5 and f = O. This corresponds to the limiting case where the grain boundary will break away from its surrounding impurity atmosphere and migrate at a rate which is controlled by the diffusion of solvent atoms across the boundary. In most cases, however, the observed relationship between V and M G will be different from the theoretical one due to drag from second phase particles if> 0) orimpurity elements in solid solution (n < 0.5).
~ = 2M (c y)(1/n-l) dt 0 2
exp( -Q
app.
1 c I RT) -=:=
[ .u: I"
D
C 2
(5-19)
(C21')(1ln-l)
and
dD
dt
= M exp(-Q
0
app.
(5-20)
From this it is seen that the parameters M; * and k are true physical constants which are related to the grain boundary mobility and the pinning efficiency of the precipitates, respectively. 5.3.1 Limiting grain size Equation (5-20) shows that the grain structure is stabilised when (dD/dt) (limiting) grain size is given by: D1im,
= kf
The parameter k (which in the following is referred to as the Zener coefficient) is defined as the ratio between the numerical constants in equations (5-7) and (5-16), respectively. In the original Zener's model k = 4/3, while other investigators have arrived at different results.8,1l-14 As shown in Fig. 5.4, the limiting grain size may vary by over one order of magnitude, depending upon the assumptions of the models. This makes it difficult to apply equations (5-20) and (5-21) for quantitative grain size analyses without further background information on the Zener coefficient.
344
t
E
200
.5 10 100
100
200
r/f, Jlm ---.
300
400
by
Fig. 5.4. Relation between limiting grain size DUm., particle radius r, and volume fractionjpredicted different models. Example (5. J)
Consider multipass GMA welding on a thick steel plate under the following conditions: I = 350A, U
30V, v
5 mm
S-I,
11
0.8, To
20C
Based on the models of Zener, 10 Hellman and Hillert,8 and Gladman+' estimate the limiting austenite grain size D lim. in the transformed parts of the weld HAZ when the oxygen and sulphur contents of the as-deposited weld metal are 0.04 and 0.01 wt%, respectively.
Solution
As shown in Chapter 2 of this textbook the volume fraction of oxide and sulphide inclusions can be calculated from equation (2-75):
f ~10-2[5.0(0.04) + 5.4(0.01-0.003)]
= 2.38X 10-3
Similarly, the average radius of the grain boundary pinning inclusions can be obtained from equation (2-79):
r(
Jl
m) 0.35 ( E)1/3
11
1/3
1000x
5)
='0.21 m
Jl
This gives the following values for the limiting austenite grain size:
Zener:
15.
11m.
=..c=4 3f
0.21 3 2. 38 X 10-3
Jl
m=118
. Jl
345
Dlim, =97=
8 r
0.21
Gladman: and
DZ1'm
= 0.52~ = 0.52
DI1'm
r 0.21 "m = 9 "m ( 1ower 1" imit) , = 0.1-f = 0.1 2.38xlO- 3 r-" r-"
As expected, the limiting austenite grain size is seen to vary by more than one order of magnitude, depending on the assumptions of the models. In practice, the Zener coefficient in low-alloy steel weld metals falls within the range from 0.32 to 0.93, as shown in Fig. 5.5. The average value of k is close to 0.52, which is the same as that inferred from the Gladman model (upper limit). When it comes to intermetallic compounds such as titanium nitride, the Zener coefficient varies typically between 0.75 and 0.25 during grain growth in the austenite regime.I6,I7 This suggests that k:::: 0.50 is a reasonable estimate of the grain boundary pinning efficiency of oxides and nitrides in steel. 5.3.2 Grain boundary mobility Direct application of equation (5-20) requires also reliable information on the time exponent n and the grain boundary mobility M. When n = 0.5 andf= 0, the classic impurity drag theories predict that the activation energy Qapp, should be close to the value for boundary self diffusion in the matrix material.i-' This borderline case is approximately attained in steel welding, as shown in Fig. 5.6(a) and (b), since the driving pressure for austenite grain growth immediately following the dissolution of the pinning precipitates is usually so large that the grain boundary migration rate approaches the higher velocity limit defined in equation (5-9).18 On this basis it is not surprising to find that Qapp, falls within the range reported for lattice self diffusion (284 kJ mol:') and boundary self diffusion (170 kJ mol ') in pure v-iron'? during welding.l'' In most cases, however, the activation energy will be different from the theoretical one due to complex interactions between impurity atoms and grain boundaries (characterised by a time exponent n < 0.5). Under such conditions, the value of Qapp, has no physical meaning. 1 5.3.3 Grain growth mechanisms Equation (5-20) provides a basis for evaluating the grain growth inhibiting effect of impurity elements and second phase particles under different thermal conditions. This also includes situations where the grain boundary pinning precipitates either coarsen or dissolve during the heat treatment process. 5.3.3.1 Generic grain growth model Equation (5-20) can readily be integrated to give the average grain size time. In the general case we may write:
as a function of
D
Do
_ [1/ D -1/ dD D.
lim.
](lIn-1)
= Mo*J
t 0
= II
(5-22)
N .CiS
E ~ cD 80
c:
100
T~ Ol
'2
60
40 20 0 r = O.32J..Lm f = 2.1 x1 0
. 3
S ~Q)
Cf)
:::3
1000
1400
(b)
150
50
50
100
r/f,
150
Jlffi --.
200
250
300
Fig. 5.5. Evaluation of the Zener coefficient in steel weld metals containing stable oxide and sulphide inclusions; (a) Determination of DUm. from isothermal grain growth data (holding time: 30 min), (b) Variation in Dlim. with the inclusion r/fratio. Data from Skaland and Grong.15
where DUm. is the limiting grain size (defined in equation (5-21. The integral I] on the right-hand side of equation (5-22) represents the kinetic strength of the thermal cycle with respect to grain growth and can be determined by numerical methods when the temperature-time programme is known. In practice, however, it is not necessary to solve this integral to evaluate, the grain growth mechanisms. Consequently, the left-hand side of equation
347
(a)
0.4
t
10 .......
x 0 ...J
C)
Steel A 0.3
10.2
0.1
~ Tp: 1350C
o
Tp: 1200C
1.2
0.4
0.8 ------
(b)
o Steel A
Steel B
Fig. 5.6. Evaluation of the time exponent n and the activation energy Qapp. for austenite grain growth in steel under thermal conditions applicable to welding; (a) Time exponent n, (b) Activation energy Qapp.' Data from Akselsen et al. 18
(5-22) can be solved explicitly for different values of D lim.' n, and II' The results may then be presented in the form of novel diagrams which show the competition between the various processes that lead to grain growth during heat treatment of metals and alloys. A more thorough documentation of the predictive power of the model and its applicability to welding is given in Section 5.4. 5.3.3.2 Grain growth in the absence of pinning precipitates In the absence of grain boundary pinning precipitates, we have: f = 0, D lim Under such conditions, equation (5-22) reduces to:
00,
= O.
348
_[l/D](lln-l) Do
15
_ dD
-I
I
(5-23)
15l/n
= Iiv +
o
(~)I n
(5-24)
Referring to Fig. 5.7, the average grain size D becomes a simple cube root function of 11 when n = 0.5 and Do = O. In other situations (n < 0.5), the grains will coarsen at a slower rate due to drag from alloying and impurity elements in solid solution. This is seen as a general reduction in the slope of the D-/lcurves in Fig. 5.7. The important austenite grain growth inhibiting effect of phosphorus and free nitrogen in steel following particle dissolution is shown in Fig. 5.8.
5.3.3.3 Grain growth in the presence of stable precipitates
If grain growth occurs in the presence of stable precipitates, the limiting grain size (Dlim) in equation (5-22) becomes constant and independent of the thermal cycle. In the specific case when n = 0.5 the integral on the left-hand side of equation (5-22) has the following analytical solution: -DDUm.
--
- (DUm.)
In(l- D / DUm.)
= 11 -
DoDUm. - (DUm.)
/ ~im.)
--
(5-25)
* In(I-15o
t
E
::l
200
u:)
100
10000
20000 ~ , [J.tm]1/n
30000 _
40000
Fig. 5.7. Predicted variation in average grain size D with II and n for f growth).
= 0 and
Do
= 0 ('free'
grain
349
(a)
200 Steel A
t
E
10
;-
150
~ 100
6
------
Number of cycles
Steel 8
t
E
-c -c
10
=:-
100
6 ----
Number of cycles
Fig. 5.8. Illustration of the austenite grain growth inhibiting effect of phosphorus and free nitrogen in low-alloy steel during reheating above the AC3 temperature (multi-cycle weld thermal simulation); (a) Steel A (50ppm P, 20ppm N), (b) Steel B (180ppm P, 80ppm N). Data from Akselsen et al.18
from which the average grain size D is readily obtained. In other cases, numerical methods must be employed to evaluate D. It is evident from the graphical representation of equation (5-25) in Fig. 5.9 that the grain growth inhibiting effect of the precipitates is very small during the initial stage of the process when D D lim.' Under such conditions the grains will coarsen at a rate which is comparable with that observed for free grain growth (n = 0.5,! = 0). The grain coarsening process becomes gradually retarded as the average grain size increases because of the associated reduction in the effective driving pressure MG until it comes to a complete stop when MG = 0 (i.e. D = DUm.)
350 600
1
10"'
400
~ 300 200
______________________ ~~~~~~O!_~ _
---
..........
- _ .. ---'_
....
- ...... --_
-------
100
50000
100000
~ ,Jlm 2
150000
-200000
with II and
DUm.
Do
= 0 (stable
300 250
f
E
:::1.
10"'
200
150
100 50 00
50000
11 '
150000
200000
Fig. 5.10. Predicted variation in average grain size D with II and n for DUm. = 250Jlm and Do = 0 (stable precipitates) ..Dotted curves correspond to grain growth in the absence of pinning precipitates.
351
If grain growth at the same time occurs under the action of a constant drag from impurity elements in solid solution, the situation becomes more complex. As shown in Fig. 5.10, a decrease in the time exponent from say 0.5 to 0.2 gives rise to a marked reduction in the slope of the V-II curves, similar to that observed in Fig. 5.7 for particle-free systems if = 0). However, the predicted grain coarsening rate is lower than that evaluated from equation (5-24) due to the extra drag exerted by the grain boundary pinning precipitates. This leads ultimately to a stabilisation of the microstructure when V = V lim.. 5.3.3.4 Grain growth in the presence of growing precipitates Very little information is available in the literature on the matrix grain growth behaviour of metals and alloys in the presence of growing second phase particles. So far, virtually all modelling work has been carried out on two phase (X-~ titanium alloys. 14 Unfortunately, none of these models can be extended to more complex alloy systems such as steels or aluminium alloys. When grain growth occurs in the presence of growing second phase particles, Dlim. will no longer be constant due to the associated increase in the particle rIJ ratio with time. As shown in Chapter 4 of this textbook, the Lifshitz-Wagner theory2o,21 provides a basis for modelling particle growth during welding and heat treatment of metals and alloys in cases where the peak temperature of the thermal cycle is kept well below the equilibrium solvus of the precipitates. Under such conditions, the particles will coarsen at almost constant volume fraction if = fo), in accordance with equation (4-16):
t r3 - r3 o
= c5 f!exPC-Q / RT)dt = I 2 T s
(5-26)
where Qs is the activation energy for the coarsening process, C5 is a kinetic constant, and 12 is the kinetic strength of the thermal cycle with respect to particle coarsening. The other symbols have their usual meaning. If the base metal contains particles of an initial radius r 0 and volume fractionj'., the limiting o grain size at 12= 0 (V lim.) can be defined as:
DC!
lim.
= k!.iL t;
3
-0
(5-27)
from which
3 r: = (fo
I k) (DUm.)
(5-28)
= (fo
I k) (DUm)
3 -
(5-29)
By combining equations (5-26), (5-28), and (5-29), we arrive at the following relationship between (V lim) and 12: (5-30)
352
It is seen from equation (5-30) that the limiting grain size in the presence of growing particles depends on the product (k/fo)3 12 In practice, the grain boundary pinning effect of the precipitates is determined by the relative rates of particle coarsening and grain growth in the material, i.e. whether the grain boundary mobility is sufficiently high to keep pace with the increase in DUm. during heat treatment. Generally, the pinning conditions are defined by the (k/fo)3 12//1 ratio, which after substitution and rearranging yields:
t
csJ
(k / 10)312 / II = (k / 10)3
-....;;...t-------
t J
In cases where the parameters Cs ,Qs' M; *, and Qapp. are known, the average grain size D can readily be evaluated from equations (5-22), (5-30), and (5-31) by utilising an appropriate integration procedure. However, since Qs normally differs from Qapp., the (k/fo)3 12//1 ratio will depend on the thermal path during continuous heating and cooling. Consequently, solution of these coupled equations generally requires stepwise integration in temperature-time space via a fourth heat flow equation. This problem will be dealt with in Section 5.4. The situation becomes much simpler if heat treatment is carried out isothermally. Under such conditions the product (k/fo)3 12 will only differ from II by a proportionality constant m, which is characteristic of the system under consideration. Accordingly, equation (5-30) can be rewritten as: DUm. = [(DUm.)
-0
+ mIl]
113
(5-32)
From this we see that the two coupled equations (5-22) and (5-32) can be .solved explicitly for different values of D~m., n, m, and II. Hence, it is possible to present the results in the form of novel 'mechanism maps' which show the competition between particle coarsening and grain growth during isothermal heat treatment for a wide range of operational conditions. Examples of such diagrams are given in Figs. 5.11 and 5.12. It is evident from these figures that the grain coarsening behaviour during isothermal heat treatment is very sensitive to variations in the proportionality constant m. For large values of m, the matrix grains will coarsen at a rate which is comp-arable with that observed in Fig. 5.7 for particle-free systems if = 0). This corresponds to a situation where the grain boundary pinning precipitates will completely outgrow the matrix grains. It is interesting to note that particle outgrowing is more likely to occur if the time exponent n is small, as shown in Fig. 5.12, because of the associated reduction in the grain boundary mobility in the presence of impurity elements in solid solution. In other systems, where the proportionality constant m is closer to unity, the reduced coarsening rate of the precipitates gives rise to a higher Zener retardation pressure and ultimately to a stagnation in the matrix grain growth. In the limiting case, when m = 0, the grain growth behaviour becomes idential to that observed in Figs. 5.9 and 5.10 for stable precipitates.
353
250
:i.
10
150 100
50
10000
20000
111J.lm2
30000
~
40000
50000
DUm.
= 50Jlm, n
= 0.5,
and
50
E ::1. lei' 20
30
10
10000
20000
~1f.1m1/n
30000
~
40000
50000
15
DUm.
354
Consider a titanium-micro alloyed steel with the following chemical composition: Ti(total): 0.016 wt%, Ti(soluble): 0.009 wt%, N: 0.006 wt% Assume that the base metal contains an uniform dispersion of TiN precipitates in the asreceived condition, conforming to a limiting austenite grain size D~m. of 50 urn. Provided that boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), estimate on the basis of Fig. 5.11 the average austenite grain size Dr in the material after 25 s of isothermal annealing at 1300C. Relevant physical data for titanium-microalloyed steels are given below:
Qs
<s
Qapp. Mo *
k
PTiN Psteel
240 kJ mor ' (activation energy for diffusion ofTi in austenite) 6.67 X 104 f..lm3 K s-I 224 kJ mol " 2.14 X 1010 Jlm2 S-I 0.5 5.2 g crrr ' 7.8 g crrr '
Solution
The initial volume fraction of TiN in the material can be estimated from simple stoichiometric calculations by considering the difference between total and soluble titanium. Taking the atomic weight ofTi and N equal to 47.9 and 14.0 g mol:', respectively, we obtain:
f
o
~1.4xl0-4
From this we see that the initial radius of the TiN precipitates in the base metal is close to: fo D-o 1.4XI0-4 0.5 50 0 014
r o =k
li tm.
t-"
um==U.
11m t-"
Since heat treatment is carried out under isothermal conditions, the parameters m and II can be obtained directly from equations (5-31) and (5-22) without performing a numerical integration:
= (0.511.4><
=27 J.Lm
10-4)3 (6.67 X 10 11573)exp( -240000/8.314 x 1573) ill 2.14 x 1010 exp( -224000/8.314 x 1573) Jl
355
RT)t
= 2.14
x 10
10
exp( -224000/8.314
The average austenite grain size can now be read from Fig. 5.11 by linear interpolation between the curves for m = 10 and 100J.lm. This gives:
Although experimental data are not available for a direct comparison, the predicted grain size is of the expected order of magnitude. From this it is obvious that considerable austenite grain growth may occur in titanium-micro alloyed steels because of particle coarsening, in spite of the fact that TiN, from a thermodynamic standpoint, is stable up to the melting point of the steel. The process can, to some extent, be counteracted by the use of a finer dispersion of TiN precipitates in the material. For example, if the initial particle radius is reduced by a factor of five (conforming to a change in I5gm. from 50 to 10 um), the austenite grain size of the annealed material decreases from 75 to 65 urn, as shown in Fig. 5.13. Nevertheless, since particle coarsening is a physical phenomenon occurring during high temperature heat treatment of metals and alloys, austenite grain growth cannot be avoided. This explains why, for instance, conventional titanium-microalloyed steels are not suitable for high heat input welding due to their tendency to form brittle zones of Widmanstatten ferrite and upper bainite in the coarse grained HAZ region adjacent to the fusion boundary.F
= 0.5
t
E
200 150
10
::t.
100
50
10000
20000
~ ,flm2
30000
----
40000
50000
DUm. =
356
5.3.3.5 Grain growth in the presence of dissolving precipitates Little information is available in the literature on the matrix grain growth behaviour of metals and alloys in the presence of dissolving precipitates. As shown in Chapter 4, the model of Whelan-' provides a basis for calculating the dissolution rate of single precipitates embedded in an infinite matrix. If the transient part of the diffusion field is neglected, the variation in the particle radius r with time t at a constant temperature is given by equation (4-18)~ (5-33) where u is the dimensionless supersaturation (defined in Fig. 4.14), and Dm is the element diffusivity. Application of the model to continuous heating and cooling requires numerical integration of equation (5-33) over the weld thermal cycle:
t r2
= r2 o
2f aD m dt
()
= r2 ()
I3
(5-34)
where 13 is the kinetic strength of the thermal cycle with respect to particle dissolution. From this relation the following expression for the particle volume fraction can be derived (see equation (4-22), Chapter 4):
(5-35) where i; is the initial particle volume fraction. By substituting D lim. = k(rlj). and DUm. = kir; /fo) into equations (5-34) and (5-35), it is possible to obtain a simple mathematical relation which describes the variation in the limiting grain size with 13 during particle dissolution. After some manipulation, we obtain:
.(5-36)
It is seen from equation (5-36) that the limiting grain size increases from DUm. at 13 = 0 to infinite when 13 = (fo Ik)2 ( DUm. )2. Since the magnitude of the Zener drag, in practice, depends on the relative rates of grain growth and particle dissolution in the material, the pinning conditions are defined by the (k/fo) 2 13 I II ratio:
I
t
exp(
-Qapp.
I RT)dt
(5-37)
357
Equation (5-37) shows that the (k/fo) 2 13 / I} ratio is contingent upon the thermal path during continuous heating and cooling. Consequently, application of the model to welding generally requires numerical integration of the coupled equations (5-22), (5-36), and (5-37) over the weld thermal cycle. However, the integration procedure is largely simplified if heat treatment is carried out isothermally. In such cases the product (k/fo) 2 13 will only differ from II by a proportionality constant m", which is characteristic of the system under consideration. By substituting m*II into equation (5-36), we obtain:
(5-38) From this we see that the two coupled equations (5-22) and (5-38) can be solved explicitly for different values of D~m., n, m *, and II' Hence, it is possible to present the results in the form of novel 'mechanism maps' which show the competition between particle dissolution and grain growth during isothermal heat treatment for a wide range of operational conditions. Examples of such diagrams are given in Figs. 5.14 and 5.15. As expected, the stability of the second phase particles is sensitive to variations in the proportionality constant m *. Normally, the precipitates will exert a drag on the grain boundaries as long as they are present in the metal matrix. However, when the dissolution process is completed, the matrix grains are free to grow without any interference from precipitates. This
250
t
E
200 150
10
::1.
100
DUm.
358
40
30
20
10
Do
DUm.
= 0 (dissolving precipitates).
means that the grains, after prolonged high temperature annealing, will coarsen at a rate which is comparable with that observed in Fig. 5.7 for particle-free systems. In the limiting case, when m" = 0, the grain growth behaviour becomes identical to that shown in Figs. 5.9 and 5.10 for stable precipitates.
Example (5.3)
Consider a niobium-micro alloyed steel with the following composition: Nb(total): 0.025 wt%, Nb(soluble): 0.010 wt%, C: 0.10 wt%
Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received condition, conforming to a limiting austenite grain size DUm. of 50 urn, Provided that the boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), estimate on the basis of Fig. 5.14 the average austenite grain size D'Yin the material after 25 s of isothermal annealing at 1300C. Relevant physical data for niobium-microalloyed steels are given below:
DNb (um? s ') : Qapp.
5.9 X 1010 exp (-343 OOOIRT) 224 kJ mor ' 2.14 X 1010 J.lm2 S-l 0.5 6.5 g crn' 7.8 g cm'
k
PNbC Psteel
359
Solution
The initial volume fraction of NbC in the material can be estimated from simple stoichiometric calculations by considering the difference between total and soluble niobium. Taking the atomic weight ofNb and C equal to 92.9 and 12.0 g mor ', respectively, we obtain:
f
o
(0.025-0.010)(92.9+12.0)x7.8 92.9x6.5xl00
= 2x10-4
From this we see that the radius of the NbC precipitates in the base metal is close to:
As shown in Example 4.6 (Chapter 4), the dimensionless supersaturation of niobium adjacent to the particle/matrix interface during dissolution can be written as: a
Nb
CiNb
= 1.02X 10-3 By substituting this value into the expression for the proportionality constant m *, we obtain:
m * = (k /
f
o
)2
2a
Nb
Nb /
RT)
(0.5/2
10-4)2 2(1.02 x 10-3 )(5. 9 x 1010 exp(-343000/ 8.314 x 1573 2.14 x 1010 exp(-224000/8.314 x 1573)
~4
The average austenite grain size can now be read from Fig. 5.14 by interpolation between the curves for m * = 1 and f = 0 (free grain growth). This gives:
Since the calculated value of Dr is reasonably close to that observed for a particle-free system, it means that the presence of a fine dispersion of NbC in the base metal has no significant effect .on the resulting austenite grain size '. under the prevailing circumstances. Other
360
types of niobium microalloyed steels may reveal a different grain coarsening behaviour, depending on the chemical composition, size distribution, and initial volume fraction of the base metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibition is always succeeded by grain coarsening as long as the precipitates are thermally unstable.
2rcA
IV(I)
exp] -(r * )2
I 4at]
(5-39)
T - To =
112
exp[ - y2 I 4at]
(5-40)
'A
ity, pc is the volume heat capacity, d is the plate thickness, and r * is the two-dimensional radius vector in the y-z plane (equal to ~ y2 + Z2 ). Although these heat flow equations are not general applicable, they provide a reasonable estimate of the temperature-time pattern within the HAZ of single pass steel welds.F Relevant thermal data for different types of steels are given in Table 1.1 (Chapter 1).
361
5.4.1.2 Grain growth model The use of equation (5-22) for prediction of the HAZ grain structure requires quantitative information about the different kinetic constants entering the model. In the following we shall assume that data for the time exponent n, the activation energies Qapp. and Qs' the temperature dependence of the dimensionless supersaturation a and the element diffusivity Dm as well as the initial and limiting grain sizes Do and 15f;m. are available in the literature. From knowledge of these parameters it is possible to calibrate the model against data from real or simulated welds by adjusting the remaining kinetic constants so thata good agreement is obtained between theory and experiments. 5.4.1.3 Calibration procedure The calibration procedure involves the following basic steps: Evaluation of experimental data Suppos~hat the mean grain size at two different locations in the HAZ is known (designated D; and D2, respectively). At each of these locations the peak temperature Tp and the cooling time within a specific temperature range (e.g. from 800 to 500ae), ilts/s' are given. The thermal cycles for the points can then be computed from either equation (5-39) or (5-40). Using these temperature profiles and an appropriate value for the activation energy Q app., the 11/M 0 * ratio is calculated from the integral: II / M; ::: exp( -Qapp.
t1
f
t2
RT)dt
(5-41)
where the limits t 1 and t2 refer to the total time spent in the thermal cycle from the chosen reference temperature T; to the peak temperature Tp and down again to T': Tuning of coarsening model When the parameters n, Qapp., Qs' Do, and DO lim. are known, it is possible to evaluate the remaining (unknown) kinetic constants from equations (5-22) and (5-30) by an iterative procedure. The following parameter is defined for this purpose: (5-42) The next step is to calculate the integral
12 I Cs =
f -yex
t,
~1
p( -Qs I RT)dt
(5-43)
for the same pairs of values of Tp and I1t8/5 as above. The differential grain growth equation can now be solved by selecting an appropriate starting value for M 0 * and evaluating the corresponding Q-value which conforms to a mean grain size of VI and V2, respectively. The computations are repeated by adjusting M; * until a contour in M; *-Q space is built up for each grain size. The accepted values of M; * and Q are then found by considering the intersection point between the two curves, as shown schematically in Fig. 5.16(a).
362 (a)
M* o
(b)
M~
Fig. 5.16. Method for calibrating unknown kinetic constants (a) M*o-Q (coarsening model), (b) M*o-Q* (dissolution model).
to experimental
Tuning of dissolution model In this case the unknown kinetic constants are combined in a single calibration parameter Q*, defined as: (5-44) where a" and supersaturation
DJ:z include all constants entering the expressions for the dimensionless
(Y
363
J
12 II
eXp[-(LVf*
+ Qd)1 RT]dt
(5-45)
where str is the standard enthalpy of the dissolution reaction per mole of the diffusate (defined by the solubility product in equation (4-5), Chapter 4), and Qd is the activation energy for diffusion of the less mobile constituent atom of the precipitates in the matrix. By calculating the integral in equation (5-45) for the same pairs of values of Tp and tltS/5 as above and selecting an appropriate starting value for M 0 * in equation (5-41), it is possible to build up a contour in Mo*-Q* space for each grain size that satisfies the differential grain growth equation. The accepted values of Mo* and Q* are then found by considering the intersection point between the curves representing D; and D2 in Fig. 5.16(b}. 5.4.1.4 Axes and features of diagrams The microstructural information calculated using the mathematical models described in the previous sections may be plotted on various kinds of welding diagrams. The graphical representation chosen here has been adopted from Ashby et al.24,25 The process diagrams have axes of weld input energy and peak temperature, and display contours of constant grain size along with information about the thermal stability of the grain boundary pinning precipitates. The axes can, in turn, be converted into an equivalent cooling time, tltS/5' and isothermal zone width tlr* or tlYm through equations (5-39) and (5-40): Thick plate welding (2-D heat flow)
(5-46)
dr * _
>
[ (n/2)epc ]
point).
q0 / v
1I2[
~Tp-ro
~Tm-To
(5-47)
where Tp s r;
tit
8/5
-~------
4APC
(5-48)
(5-49)
364
Welding diagrams of this kind are very useful, since they summarise the effect of the important process variables in a systematic and illustrative manner at the same time as they provide a good overall indication of the grain growth behaviour of materials during welding. In addition, the diagrams can be used for quantitative predictions of the austenite grain size across the HAZ of steel welds for a wide range of operational conditions. This will be illustrated below in various numerical examples. 5.4.2 Case studies The following section describes grain growth diagrams for different types of steels. The parameters used to construct the maps are either tabulated or included in the text. Some of these are taken from the literature, while others are arrived at by fitting the theory of the previous sections to data from real or simulated welds according to the procedure shown in Fig. 5.16. To obtain a consistent basis, all grain sizes reported here conform to the mean linear intercept.28 Conversion to three-dimensional grain sizes may then be done through equation (5-4) or by the use of other appropriate conversion factors (e.g. 1.776 as recommended by Ashby et al. 24,25). 5.4.2.1 Titanium-microalloyed steels Titanium-microalloyed steels are widely used in welded structures. From a thermodynamic standpoint, additions of small amounts of titanium to steel would be expected to impede austenite grain coarsening during welding by virtue of its ability to form stable nitrides even at high temperatures (see discussion in Chapter 4). However, certain restrictions must be adhered during casting and subsequent thermomechanical processing of the steel so that the number density of TiN particles is sufficiently high to retard grain growth.P This currently limits the use of titanium for austenite grain size control to continuously cast and controlled rolled steels. Typical HAZ grain growth diagrams for Ti-microalloyed steels can be constructed on the basis the experimental data reported by Ion et al. 25 Tables 5.1 and 5.2 contain information about the steel chemical composition and the parameters used in the computations, while Fig. 5.17 gives an overall indication of the accuracy of the predictions after calibration of the model against two experimental data points.
Table 5.1 Chemical composition of Ti-microalloyed
C 0.12 Si 0.23 Mn 1.53 P 0.005
Table 5.2 Data used to construct welding maps for Ti-microalloyed ous sources).
n
Q/
jjo
lim.
Do
(urn) 9
(urn) 20
t Activation
365
E :i. 60 cD N 50 w
80 70
Ti-microalloyed steel
.~ 40 C)
"0
~ ::J
30
(JJ
co 20 Q)
10
10
20
30
40
50
60
70
80
Predicted grain size, Jl m ----Fig. 5.17. Comparison between measured and predicted HAZ austenite grain sizes after calibration of model to data reported by Ion et al. 25 for Ti-microalloyed steel (simulated thick plate welds).
The response of the base material to welding under 2-D and 1-D heat flow conditions is shown in Fig. 5.18(a) and (b), respectively. As expected, the presence of TiN particles is seen to retard austenite grain growth within the heat affected zone during welding. However, since particle coarsening is a physical process occurring at temperatures well below the equilibrium solvus of the precipitates, the problem cannot be eliminated. This means that a coarse grained region will always form adjacent to the fusion boundary, even at very low heat inputs, as indicated by the nomograms in Fig. 5.18(a) and (b).
Example (5.4)
= 6mm
S-1,
11 = 0.95, To = 20C
Evaluate on the basis of the nomograms in Fig. 5.18(a) the variation in the austenite grain size across the fully transformed HAZ after welding. Estimate also the total width of the HAZ (referred to the fusion boundary) under the prevailing circumstances.
Solution
First we calculate the net heat input per unit length of the weld: Iv= llx1xU
q0
= 0.95x500X30kJmm-1
1000 x v
1000 x 6
=2.38 kJ mrrr ' Readings from Fig. 5.18(a) give the HAZ austenite grain size profile shown in Fig. 5.19.
366 (a)
Thick plate welding (2-D heat flow) -Prior austenite grain size (J.tm) Relative size of pinning precipitates (r/ro)
20 10 30 40
Tm
50
f
~ ~ ~
,
8
\ \
-E -,
~
6
40
-r
\ \ \
en
J5
30
<J
Q)-
'5 C. 4
~
\
.S .c Q)
Z
<ti Q)
20
.s
0
Ol
(5
o
2 10
900
1000
1100
1200
1300
1400
1500
(b)
Thin plate welding (1-0 heat flow) Prior austenite grain size (JLm) - - - Relative size of pinn.ing preclpitates (r/ro)
10J!m 20gm 30J!m 0.20
T m
t
C\I
E .E
.c
S a. .s 1U Q)
(j)
E:
<
:0
0
0.10
0.05
o~---~---~------~---~------~---~---~----~ 900
1000
1100
~ __~
_
1400
~ __~~~
1500
Fig. 5.18. HAZ grain growth diagrams for titanium-microalloyed steel; (a) Thick plate welding (2-D heat flow), (b) Thin plate welding (l-D heat flow). No preheating (To = 200e).
367
SAW (Ti-microalloyed
steel)
('t)
I-E
+-
1500
Peak temperature,
C ------
Fig. 5.19. Predicted variation in austenite grain size across the fully transformed HAZ of a Ti-microalloyed steel weld (Example 5.4).
The total width of the fully transformed HAZ can now be estimated from equation (5-47) by using data from Table 1.1 (Chapter 1). Taking the AC3 -temperature equal to 910cC, we obtain:
tir
A
m -
0.95 x 500 x 30
1I2[
1 '\1"910- 20
1 '\1"1520- 20
] mm
=2.6mm Based on the quoted value of.Q in Table 5.2, it is also possible to estimate the initial volume fraction of TiN in the base metal. Taking Cs::::: 6.67 X 104 Jlm3 K S-1 and k::::: 0.5 for titaniummicro alloyed steels, we obtain from equation (5-42):
f =k
o
113 C5 . ( Q)
= 0.5 6.67 x 10
9.9xlo14
(...
4 )1/3
= 2 X 10-4
A comparison with the experimental data of Ringer et al. 17 shows that a volume fraction of 2 X 10-4 is reasonably close to that measured by microscopic assessment methods.
5.4.2.2 Niobium-microalloyed steels This class of steel is used for a variety of applications, including ship building, pressure vessels, oil platforms and bridges. The steels contain small amounts of microalloying elements such as niobium, vanadium, and aluminium which during thermomechanical processing combine with carbon and nitrogen to form fine dispersions of grain boundary pinning precipitates. They are readily weldable, but suffer from severe HAZ grain coarsening due to dissolution of the carbo-nitrides at elevated temperatures. 18,22
368
Typical HAZ grain growth diagrams for Nb-microalloyed steels (with chemical composition as in Table 5.3) can be constructed on the basis of the experimental data of Ion et al. 25 The parameters used to compute the diagrams are listed in Table 5.4. Figure 5.20 shows a correlation between predicted and observed grain sizes after calibration of the model against two experimental data points. It is evident from the nomograms in Fig. 5.21 (a) and (b) that considerable austenite grain growth occurs during welding of niobium-microalloyed steels because of dissolution of the grain boundary pinning precipitates. On the average, the HAZ austenite grain size adjacent to the fusion boundary is four to six times larger than that observed for titanium-microalloyed steels. This gives rise to a high HAZ hardenability, which facilitates formation of low temperature transformation products such as martensite and bainite during welding .22
Example (5.5)
In Table 5.4 the quoted value for the effective grain boundary pinning constant n* is 5.0 X 1018 urn? S-I. Based on equations (5-27) and (5-44), estimate the initial volume fractionj', and radius roof NbC in the parent material under the prevailing circumstances.
Table 5.3 Chemical composition of Nb-microalloyed
C 0.12 Si 0.16
Mn
0.91
P 0.002
250
E ::l cU
'(h
N
200
@
150
* o
c '(~ 13
C) "C
~ ::s
U)
100
<D
50
Fig. 5.20. Comparison between measured and predicted HAZ austenite grain sizes after calibration of model to data reported by Ion et al,25 for Nb-microalloyed steel (real and simulated thick plate welds),
369
(a)
Thick Qlate welding (2-D heat flow) -Prior austemte grain size (urn) - - - Relative volume fraction 01 pinning precipitates (fIfo)
t 10'-~~~~~Im~Trr-"~~'\~~"~~150
8
10gm
SO).!m 100 m
200
300
40 1
!!?
30
... 00
<l
:
4
20 .~ (5
cD
o
2
10
1000
oL-~~L-~~~Q~~~~~~~~~~~~o
900
1100
1200
1300
C ~
1400
1500
Peak temperature,
(b)
T
f 0.20
<:o
~
:0;
'5
Q)
0.10
.s
s:
.:l.
Qi
0.05
1000 1100
1200
1300
C
Peak temperature,
__~~_
1400
1500
Fig. 5.21. HAZ grain growth diagrams for niobium-microalloyed steel; (a) Thick plate welding (2-D heat flow), (b) Thin plate welding' (l-D heat flow). No preheating (To 20C).
370
Table 5.4 Data used to construct welding maps for Nb- microalloyed steel (compiled from miscellaneous sources).
n
o.;
sn=
(kJ mor ') 130
Qi
DO
(J.Lm) 11
lim.
D0 (J.Lm) 6
(J.Lm2S-I) 37 X 109
M*
n*
(J.Lm2S-I) 5.0 X 1018
tEstimated from the solubility product of NbC in austenite. tActivation energy for diffusion of Nb in austenite.
below:
20.5 (estimated from the solubility product of NbC in austenite, Example 4.6) D~ 5.9 X 1010 J.lm2 S-1 (diffusion constant for Nb in austenite) k : 0.5 (Zener coefficient) -0 DUm.: 11 urn (from Table 5.4)
Solution
The initial volume fraction of NbC can be estimated from equation (5-44). After rearranging this equation, we obtain:
10
J1/2
=3.5xl0-4
The corresponding radius of the precipitates can now be obtained from equation (5-27):
r,
10.5)IIJ.lmzO.008J.lm
(i.e.i Snm)
Although experimental data are not available for a direct comparison, the calculated values for I; and ro are reasonable and of the expected order of magnitude (see Example 5.3). 5.4.2.3 C-Mn steel weld metals In C-Mn steel weld metals the volume fraction of non-metallic inclusions is considerably higher than for normal cast steel products because of the limited time available for growth and separation of the deoxidation products (see discussion in Chapter 2). For the same reason, the weld metal inclusions are also significantly smaller in dimension and more finely dispersed. Since oxides and sulphides will neither coarsen nor dissolve during the weld thermal cycle, they will ultimately lead to a stabilisation of the austenite grain structure within the reheated regions of multipass steel weld metals. This is illustrated schematically in Fig. 5.22. Grain growth diagrams for C-Mn steel weld metals (with chemical compostion as in Table 5.5) can be constructed on the basis of the theory described in the previous sections by selecting reasonable average values for the kinetic constants in equation (5-22). The combination of parameters listed in Table 5.6 gives a fair agreement between predictions and measurements, as shown in Fig. 5.23. The response of the material to reheating above the AC3-temperature under 2-D and I-D heat flow conditions is illustrated in Fig. 5.24(a) and (b). As expected, the presence of finely
371
Base plate
Q)
:--I
1(1)
Fully transformed
HAZ~
'00
C '(5
- ..-----I
C,
'c
l-E
Q) ....., en :J
Peak temperature Fig. 5.22. Schematic diagram illustrating the effect of non-metallic inclusions on the weld metal grain coarsening behaviour, Table 5.5 Chemical composition of C-Mn steel weld metal used by Kluken et al. 29 (in wt%),
c
0.09
0
0.034
Si 0.48
Mn 1.86
P 0.01
S 0.01
N 0.005
Nb 0.004
V
0.02
Al 0.018
Ti 0.005
Table 5.6 Data used to construct welding maps for C-Mn steel weld metal (compiled from miscellaneous sources),
n
: (kJ
224
D0
(urn)
Dlim.
(um)
mot+)
(f.Lm2 s")
M* 0
0.5
10
95
20 X 109
dispersed inclusions within the weld metal gives rise to a strong austenite grain boundary pinning effect, similar to that documented for TiN in microalloyed steels. However, since no particle coarsening occurs in the present case, it means that a small austenite grain size 95J.lm) is preserved at all relevant peak temperatures, irrespectively of the applied heat input. Accordingly, the weld metal grain growth behaviour is seen to be quite different from that of the base metal, even when the nominal chemical composition has not been significantly changed by the welding process.
372 120
t
E
N
100
cD
::l
80
'w
'(ij
C, 60
'2
(/)
.m 40
~
20
0 CMn steel weld metal Peak temperature: 1350 C
<X:
20
40
60
80
100
-----
120
140
160
Fig. 5.23. Comparison between measured and predicted austenite grain sizes after calibration of model to data reported by Kluken et al. 29 for C-Mn steel weld metal (simulated thick plate heat cycles).
Example (5.6)
Consider deposition of a cap layer (GMAW) on the top of a thick multipass Nb-microalloyed steel weld under the following conditions: 1= 400A, U = 30V, v = 5mm
Estimate on the basis of the nomograms in Figs, 5.21(a) and 5.24(a) the variation in the prior austenite grain size across the fully transformed HAZ at different locations along the periphery of the weld after arc extinction. The situation is illustrated in Fig. 5.25.
Solution
First we calculate the net heat input per unit length of the weld: 11 x I x U q / v=
a
1000
X V
=1.92kJ mm ! Readings from Figs. 5.21(a) and 5.24(a) give the HAZ austenite grain size profiles shown in Fig. 5.26, From this we see that the prior austenite grain size adjacent to the fusion boundary of a cap layer will vary significantly with position along the periphery of the weld, depending on the type of material sampled (i.e. base plate or weld metal). 5.4.2.4 Cr-Mo low-alloy steels Chromium-molybdenum low-alloy steels are widely used in the petroleum industry and in high-power-generating equipment because of their good corrosion and oxidation resistance
373
(a)
10
50
:i
~
0
E E 6
40
en
30
..JO <J
.5
:; 04
Q)
E +=
(5
0
a>
20 .5 0
0)
s:
CD
2 10
900
1000
1100
1200
1300 C
1400
1500
Peak temperature,
(b)
t
(\J
.; <
0
.x:
. ....,
0.15
.s
(tj
Q) .r.
S c..
0.10
CD
0.05
OL---~--~--~--~--~--~--~--~---L---L--~--~-U~
900 1000 1100 1200 1300 Peak temperature, C ----
1400
1500
Fig. 5.24. Grain growth diagrams for reheated C-Mn steel weld metal; (a) Thick plate welding (2-D heat flow), (b) Thin plate welding (l-D heat flow). No preheating (To = 20C).
374
METALLURGICAL
MODELLING
OF WELDING
Base plate
250
1
ai
I I I
I
200
150
I
I
I
I I
'00
I I
I
co
'-
c
0>
('t)
.~ 100
~en
::l
50
900
1000
1100
1200
1300
1400
150
Peak temperature, C
Fig. 5.26. Predicted variation in austenite grain size across the fully transformed HAZ at different locations along the periphery of the weld (Example 5.6).
and their high creep strength. They are readily weldable, although reheat cracking and cold cracking in the HAZ and fusion zone may be a problem. 30 In practice, these difficulties can be overcome by the choice of an appropriate welding procedure (e.g. preheating in combination with a low heat input), which reduces austenite grain growth and maximises the proportion of the HAZ refined by subsequent weld passes.30,31 The grain growth data of Miranda and Fortes'" provide a basis of calibrating the kinetic constants in equations (5-22) and (5-45), as shown in Fig. 5.27. Information about steel chemical composition and parameters used to construct the maps are contained in Table 5.7 and 5.8, respectively. It is seen from the nomograms in Fig. 5.28(a) and (b) that considerable austenite grain growth occurs in Cr-Mo low-alloy steel weldments due to dissolution of molybdenum carbide (Mo2C) at elevated temperatures. However, the maximum HAZ austenite grain size is signifiTable 5.7 Chemical composition of Cr-Mo steel used by Miranda and Fortes''! (in wt%).
C 0.10 Si 0.24 Mn 0.46 P 0.01 S 0.01 Cr 2.19 Mo 0.94 Ni 0.23 Cu 0.1 Al 0.01 V 0.01
375
80
E 60
:::1.
cD
'(j)
N
.~ 40 C)
"C
c::
'0 :0
Q)
Symbol
~ 0-
20
* +
x
00
60
19.1
20
40
80
Measured grain size, urn -----Fig. 5.27. Comparison between measured and predicted HAZ austenite grain sizes after calibration of model to data reported by Miranda and Fortes31 for Cr-Mo low-alloy steel (SAW-thick plates).
Table 5.8 Data used to construct welding maps for Cr-Mo low-alloy steel (from Refs. 24 and 31).
n
Qapp.
sn=
(kJ mol:")
(kJ mol:")
Q/t
300
DO lim.
(urn)
D0
(urn)
M*
(/-Lm
lln
Q*
S-I)
(f.Lm2 S-I)
0.32
180
71
20
1.3 X 109
7.8 X 1017
+Estimated from the solubility product of Mo.C in austenite. tt Activation energy for diffusion of Mo in austenite.
cantly smaller than that observed during welding of Nb-microalloyed steels, in spite of the fact that Mo2C is less stable than NbC. This situation can probably be attributed to drag from alloying and impurity elements in solid solution, which retard grain growth through elastic attraction of the atoms towards the open structure of the grain boundary (indicated by a time exponent n = 0.32 in Table 5.8). 5.4.2.5 Type 316 austenitic stainless steels The 316 series of stainless steels have good general corrosion resistance, but are sensitive to stress corrosion cracking in the presence of chlorides. They are used extensively in the chemical industry and in power plants, particularly in advanced nuclear technology. During welding dissolution of the base metal chromium carbides (e.g. Cr23C6) will occur in parts of the HAZ where the peak temperature of the thermal cycle has been above about 850C. This gives rise to considerable austenite grain growth.i" HAZ grain growth diagrams for this type of steel (with chemical composition as in Table 5.9) can be constructed on the basis of the experimental data of Ashby and Easterling.i" using the parameters listed in Table 5.10. Figure 5.29 shows a correlation between predicted and observed grain sizes after calibration of the model against two experimental data points.
1200
1300
1400
1500
Peak temperature, C ~
Fig. 5.28. HAZ grain growth diagrams for Cr-Mo low-alloy steel; (a) Thick plate welding (2-D heat flow), (b) Thin plate welding (l-D heat flow). No preheating (To = 20C).
377
Table 5.9 Chemical composition of type 316 austenitic stainless steel used by Ashby and Easterling-" (in wt%).
C 0.05 Mn 2.0 Si 1.0 Cr 19 Ni 11 Mo 2.3 P 0.045 S 0.030
Table 5.10 Data used to construct welding maps for type 316 austenitic stainless steel (compiled from miscellaneous sources).
n
Q}
DO
lim.
jj
0
(urn) 18
(urn) 18
Q*
(f.Lm2
S-I)
2.1 X 1014
tEstimated from data quoted by KOU32 for Cr23C6 in 304 stainless steel. :j:Activation energy for diffusion of Cr in austenite.
r
"0
50
E ::i. of
N
'en
~
40
'e
<t$ CD
c:
Cl
:J en
30
:E
20
Calibration point 50
30
40
60
Predicted grain size, um -----Fig. 5.29. Comparison between measured and predicted HAZ austenite grain sizes after calibration of model to data reported by Ashby and Eastcrling-" for type 316 stainless steel (simulated thick plate welds),
It follows from the nomograms in Fig. 5.30(a) and (b) that the presence of Cr23C6has little influence on the HAZ grain coarsening behaviour because it dissolves at a fairly low temperature. This results in a rather coarse austenite grain structure adjacent to the fusion boundary, which in certain cases may exceed 100 urn,
378
METALLURGICAL MODEJ,..LING OF WELDING Thick plate welding (2-D heat flow) Austenite grain size m) .. Relative volume fraction of pinning
4t
. . precipitates
(fIfo) 1
Tm
1
~ ~
o
30
-~
<J
cD
0)
"S
.5
a. 4
~
20 .s
'0
U
C1> .c Q)
ca
10
Boo
1000
1100
1200
Peak temperature,
Thin plate welding (1-D heat flow) Austenite grain size (p.m) - - - Relative volume fraction of pinning
precipitates
(fIfo)
t
C\J
30gm
40 m 50gm
0.20
E ~
<o
.5 s:
~ ~
0.15
~ 0.10 "5 a. i Q)
0.05
1200
1300
1400
1500
Peak temperature,
"C ~
Fig. 5.30. HAZ grain growth diagrams for type 316 austenitic stainless steel; (a) Thick plate welding (2-D heat flow), (b) Thin plate welding (I-D heat flow). No preheating (To = 20C).
379
Example (5.7)
Consider GTA butt welding of a 2mm thin sheet of type 316 austenitic stainless steel under the following conditions: 1= 250 A, U
15 V, v
= 8 mm s', TI = 0.4,
To
= 20C
Provided that the conditions for one dimensional (I-D) heat flow are met, estimate on the basis of the nomograms in Fig. 5 .30(b) the variation in the austenite grain size across the HAZ after welding. Estimate also the total width of the HAZ (referred to the fusion boundary) under the prevailing circumstances.
Solution
First we calculate the net heat input per mm? of the weld: qo / vd = 11 x I x U = 0.4 x 250 x 15 kJ mm " 1000(v x d) 1000(8 x 2)
= 0.094
kJ mm?
Readings from Fig. 5.30(b) give the HAZ austenite grain size profile shown in Fig. 5.31. The total width of the grain growth zone can be estimated from equation (5-49) by using data from Table 1.1 (Chapter 1). Taking the peak temperature for incipient dissolution of Cr23C6 equal to 900C, we obtain:
100
N .Ci)
elf
C
::t. 60
.~ 40 en $
2 .sa ~ 20
<
900
Fig. 5.31. Predicted variation in austenite grain size across the HAZ of a 316 austenitic stainless steel weld (Example 5.7).
380
381
(a)
f
~
'c :::J
-:t 0
e~ :e ~
,...
x
10
3 2
10 ~
12
14
16
1/2
(arbitrary units)
(b)
7.-------.-----------~------------~----~ H 5:1
><
10
2--------~----------~------------~----~ 10 5 15
t1/2 (arbitrary (c)
units)
10
15
----....
t1/2 (arbitrary
units)
Fig. 5.32. Effect of 'thermal pinning' on grain growth; (a) Plot of D vs t for different simulation conditions, (b) Coarsening kinetics of different thermal regions for a mobility gradient of 5: 1, (c) Same as in (b) for a mobility gradient of 10: 1. Data from Fortes and Soares.V
382
5.5.2 Free
surface
effects
In a weld HAZ the fusion line represents a physical barrier against grain growth which cannot be exceeded. In principle, it can be regarded as a free surface, which means that the grain boundaries must meet the fusion line at right angles. The simulation results of Sretre and Ryum''? provide a basis for evaluating to what extent grain growth under isothermal conditions is affected by the presence of a free surface. Figure 5.33 shows the evolution of a 2-D Voronoi structure with time. In Fig. 5.33(a) all grain boundaries are straight lines, but the triple line junctions are not in equilibrium. However, adjustments of the triple line junctions into equilibrium positions lead to local curvatures of the grain boundaries near the junctions, and the grain growth process is thus initiated. Initially, the Voronoi structure contained 485 grains, but this number is gradually reduced during the coarsening process (Figs. 5.33(b), (c) and (d)). A qualitative inspection of the diagrams reveals, on the other hand, no clear difference in the grain size in the radial position, which indicates that the constrain provided by the free surface is only of minor importance in the present context. It should be noted that this does not exclude the possibility that the HAZ grain size is influenced by the presence of a fusion boundary, since other effects such as surface grooving and solute segregation (not studied here) can impose additional restrictions on the system by contributing to physical pinning of the grain boundaries. Consequently, further modelling work is required to explore these possibilities.
References
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. H. Hu and B.B. Rath: Metall. Trans., 1970, 1, 3181-3184. J.W. Cahn: Acta Me tall. , 1962,10,789-798. K. Lucke and H. Stuwe: in Recovery and Recrystallization of Metals, 1963, New York, Interscience. E. Nes, N. Ryum and o. Hunderi: Acta Me tall., 1985,33, 11-22. E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co. C.S. Pande: Acta Me tall., 1987,35,2671-2678. B.R. Patterson and Y. Liu: Metall. Trans., 1992, 23A, 2481-2482. P. Hellman and M. Hillert: Scand. J. Metall., 1975,4, 211-219. R.A. Vandermeer: Acta Metall., 1967, 15,447-458. C. Zener (quoted by C.S. Smith): Trans. AIME, 1948,175,15-51. C.H. Womer and P.M. Hazzledine: JOM, 1992,44 (No.9), 16-20. B.B. Rath and H. Hu: Trans. TMS-AIME, 1969,245, 1243-1252, ibid., 1577-1585. T. Gladman: Proc. Roy. Soc., 1966, 294A, 298-309. G. Grewal and S. Ankem: Acta Metall. Mater., 1990,38,1607-1617. T. Skaland and 0. Grong: University of Trondheim, The Norwegian Institute of Technology, Trondheim Norway (unpublished work). K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 2nd Edn, 1992, Oxford, Butterworth-Heinemann Ltd. S.P. Ringer, W.B. Li and K.E. Easterling: Acta Me tall., 1989,37, 831-841. O.M. Akselsen, 0. Grong, N. Ryum and N. Christensen: Acta Metall., 1986,34,1807-1815. A.M. Brown and M.F. Ashby: Acta Metall., 1980,28,1085-1101. J.M. Lifshitz and V.V. Slyozov: 1. Physics Chem. Solids, 1961, 19, 35-50. C. Wagner: Z. Electrochem., 1961, 65, 581-591. 0. Grong and O.M. Akselsen: Metal Construction, 1986,18,557-562.
383
(a)
(b)
(c)
(d)
Fig. 5.33. The evolution of 2-D grain structures during normal grain growth; (a) Initial Voronoi structure, 485 grains, (b) 456 grains, (c) 349 grains, (d) 251 grains. Data from Satre and Ryum.l"
M.J. Whelan: Metal Sci. J., 1969,3,95-97. M.F. Ashby and K.E. Easterling: Acta Metall., 1982,30, 1969-1978. J.C. Ion, K.E. Easterling and M.P. Ashby: Acta Metall., 1984,32,1949-1962. D. Rosenthal: Trans. ASME, 1946, 68, 849-866. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1990,38,449-460. ASTM Standard, E112-84 (1984). A.O. Kluken, 0. Grong and H. Hemmer: Technical Report STF34 F87093, 1987, Trondheim (Norway), Sintef-Division of Metallurgy.
384
30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45.
METALLURGICAL
MODELLING
OF WELDING
P.J. Alberry and W.K.C. Jones: Metals Technology, 1977,4,45-51; ibid., 360-364; ibid., 557566. R.M. Miranda and M.A. Fortes: Mater. Sci. Eng., 1989, AI08, 1-8. S. Kou: Welding Metallurgy, 1987, Toronto, John Wiley & Sons. V.E. Fradkov and D.G. Udler: in Simulation and Theory of Evolving Microstructures (Eds M.P. Anderson and A.D. Rollett), 1990, TMS, Warrendale, Pa, pp.15-29. R.J. Frost, C.V. Thompson and D.T. Walton: ibid., pp.31-39. K. Kawasaki, T.Nagai and S. Ohta: ibid., pp.65-77. T.O. Sretre and N. Ryum: in Modelling of Coarsening and Grain Growth (Eds C.S. Pande and S.P. Marsh), 1993, The Minerals, Metals & Materials Society, pp.281-294. M.A. Fortes andA. Soares: ibid., pp. 257-270. A.D. Rollett, D.J. Srolovitz and M.P. Anderson: Acta Metall., 1989,37,1227-1240. G.S. Grest, M.P. Anderson, D.J. Srolovitz and A.D. Rollett: Scripta Metall. Mater., 1990,24, 661-665. D.J. Srolovitz, G.S. Grest and M.P. Anderson: Acta Metall., 1986,34, 1833-1845; ibid., 1988, 36, 2115-2128. T.O. Sretre, O. Runderi and E. Nes: Acta Me tall., 1986,34,981- 987. K. Marthinsen, O. Lohne and E. Nes: Acta Metall., 1989,37, 135-145. A.D. Rollett, M.J. Lutony and D.J. Srolovitz: Acta Metall. Mater., 1992, 40, 43-55. A. Kern, W. Reif and U. Schriever: Proc. Int. Conf. on Grain Coarsening, 18-21 June, 1991, Rome, Mater. Sci. Forum, 1992, vols 94-96, 709-714, Trans. Tech Publications. B. Radhakrishnan and T. Zacharia: Proc. Int. Con! on Modeling and Control of Joining Processes, 8-10 Dec., 1993, Orlando, FL, American Welding Society.
S-I)
-0
DUm.
limiting grain size at 12 = 0 (urn or m) element diffusivity (um? s ', mmor m2 s+)
S-1
end temperature of ferrite to austenite transformation (OC or K) nominal alloy composition (wt%, at. % or mol m') various kinetic constants and temperature-dependent parameters plate thickness (mm) grain size (urn or m) mean grain size (urn or m) GMA W
Co
Cl C5
f
fo
particle volume fraction initial particle volume fraction maximum pinning force exerted by a single particle (J nr ') gas metal arc welding gas tungsten arc welding standard enthalpy of dissolution reaction per mole of the diffusate (J mol:' or kJ mol:")
d D D D*
mean linear grain intercept (urn or m) GTAW initial grain size (urn or m)
DUm.
385
activation energy for particle coarsening (J mol-lor kJ mol:") particle radius (nm, urn or m) initial particle radius at t = 0 (nm, urn orm) two-dimensional plane (mm) universial mol ") radius vector in y-z
amperage (A) kinetic strength of thermal cycle with respect to grain growth (flmlln) kinetic strength of thermal cycle with respect to particle coarsening (um'') kinetic strength of thermal cycle with respect to particle dissolution (um'') Zener coefficient constant related to the relative rates of grain growth and particle coarsening in the material (flm3-I/n) constant related to the relative rates of grain growth and particle dissolution in the material (flm2-lIn) grain boundary mobility (variable units) kinetic constant in expression (variable units) for M
Qs
It
12
ro
r*
t,
k m
gas constant
(8.314 J K-1
Sv
m*
SAW
M Mo M*0 n
tI, t2 I1t8/5
integration limits (s) cooling time from 800 to soooe (s) temperature (Oe or K) reference temperature (OC or K) ambient temperature (DC or K) melting point (Oe or K) peak temperature (DC or K) voltage (V) welding speed (mm s-l) grain boundary migration rate (urn or m s')
S-l
T modified kinetic constant in expression for M (um!" S-I) time exponent in grain growth equation number of interacting particles per unit area of grain boundary (m-2) number of particles (nr') dri ving pressure (J rrr ') effective driving growth (J rrr') per unit volume
t;
To Tm
na Nv PG Me Pz
qo
r,
U
v
retardation pressure due to second phase particles (J m-3) net arc power (W) apparent activation energy for grain growth (J mol-lor kJ mol ") x activation energy for element diffusion in dissolution model (J mol-lor kJ mol ') Vb
grain boundary migration rate conforming to the low velocity limit (urn S-1 or m S-I) grain boundary migration rate conforming to the high velocity limit (urn S-1 or m S-I) x-axis/welding y-axis/transverse direction (mm) direction (mm)
o-:
Qd
386 z a
0.
y
Q
interfacial
energy
calibration constant model (!lm3 K S-l) calibration constant model (11m2 s')
in coarsening
Q*
in dissolution
'V
parameter depending on the diffusivity and the interaction energy between the grain boundary and the impurity atoms another complex function of the impurity diffusivity and the interaction energy between the grain boundary and the impurity atoms intrinsic drag coefficient arc efficiency
p p*
density (g em:' or kg m') radius of curvature of a spherical grain (11m or m) volume heat capacity (J
'V/~2
pc
mm='
DC-I)
e 11
6.1 Introduction
The majority of phase transformations occurring in the solid state take place by thermally activated atomic movements. In welding we are particularly interested in transformations that are induced by a change in temperature of an alloy with a fixed bulk composition. Such transformations include precipitation reactions, eutectoid transformations, and massive transformations both in the weld metal and in the heat affected zone. Since welding metallurgy is concerned with a number of different alloy systems (including low and high alloy steels, aluminium alloys, titanium alloys etc.), it is not possible to cover all aspects of transformation behaviour. Consequently, the aim of the present chapter is to provide the background material necessary for a verified quantitative understanding of phase transformations in weldments in terms of models based on thermodynamics, kinetics, and simple diffusion theory. These models will then be applied to specific alloy systems to illuminate the basic physical principles that underline the experimental observations and to predict behaviour under conditions which have not yet been studied.
388
Stable (8)
Stable (a)
Temperature Fig. 6.1. Schematic representation of the molar free energies of two solid a and ~ phases as a function of temperature (allotrophic transformation - no compositional change).
reaction can proceed when the driving force Il.G becomes greater than the right-hand side of equation (~-1). tor an allotropic transformation, in which there is no compositional change, dG will be a simple function of temperature, as illustrated in Fig.o.I. For alloys the situation is slightly more complex, since there is an additional variable, i.e. the composition. 'In such cases the temperature at which the a-phase becomes thermodynamically unstable (TeqJ corresponds to a fixed point on the a-J3 solvus boundary in the equilibrium phase diagram, as shown schematically in Fig: 6.2. Since phase diagrams are available for many of the important industrial alloy systems, it means that the driving force for a transformation reaction can readily be ~obtained from such diagrams in the form of a characteristic undercooling (fl.T).
%B
Fig. 6.2. Schematic representation of the a-~ solvus boundary in a simple binary phase diagram.
389
6.2.2 Heterogeneous nucleation in solids In general, solid state transformations in metals and alloys occur heterogeneously by nucleation at high energy sites such as grain corners, grain boundaries, inclusions, dislocations and vacancy clusters. The potency of a nucleation site, in turn, depends on the energy barrier against nucleation (LlG* het) which is a function of the 'wetting' conditions at thesubstratel nucleus interface.' It can be seen from Fig. 6.3 that nucleation at for instance inclusions or dislocations is always energetically more favourable than homogeneous nucleation (~GZet. < ~G;om.) but less favourable than nucleation at grain boundaries or free surfaces. As a result, the transformation behaviour is strongly influenced by the type and density of lattice defects and second phase particles present within the parent material. 6.2.2.1 Rate of heterogeneous nucleation Whereas every atom is a potential nucleation site during homogeneous nucleation, only those associated with lattice defects or second phase particles can take part in heterogeneous nucleation. In the latter case the rate of nucleation (Nhet) is given by:l,2
Ivhet. = vN v exp(-
Jexp(-
Qd RT
(6-2)
where v is a vibration frequency factor, N; is the total number of heterogeneous nucleation sites per unit volume, ~GZet. is the energy barrier against nucleation, and Qd is the activation energy for atomic migration across the nucleus/matrix interface.
1
C!J'
0 ..c
Inc/USion~
~ <J
..c (!)
05
a>
tJ)
~ctS
32
c 0
tJ)
S
0>
C/)
o+J
(l) C/)
...
~
u..'t::
~~
en Q)a>
C/)
c:~ --(I)
c'a:;: ._ctS
(50
cae:
'-c: (!):::s
ccS"'c
'en :J
<3
~ ~
c: o
c
13
0
::J
c
o
(/)
ctS
.5
~o
Vi
0
>
as
::s
0 .c
(5
Nucleation site
..
nucleation in metals and
Fig. 6.3 Schematic diagram showing the most potent sites for heterogeneous alloys.
390
It follows from the graphical representation of equation (6-2) in Fig. 6.4 that the nucleation rate Nhet.is highest at an intermediate temperature due to the competitive influence of undercooling (driving force) and diffusivity on the reaction kinetics. This change in N het, with temperature gives rise to corresponding fluctuations in the transformation rate, as shown schematically in Fig. 6.5. Note that the peak in transformation rate is due to two functions, growth and nucleation (which peak at different T) whereas peak in N het. is due to nucleation only. 6.2.2.2 Determination of ~G~et. and Qd During the early stages of a precipitation reaction, the reaction rate may be controlled by the nucleation rate Nhet: Under such conditions, the time taken to precipitate a certain fraction of the new phase t* is inversely proportional to Nhet.:
(6-3) where CI and C2 are kinetic constants. By taking the natural logarithm on both sides of equation (6-3), we obtain:
In(t)=ln(c
/ N )+ dGZet. +Qd
v
RT
(6-4)
If the complete C-curve is known for a specific transformation reaction, it is possible to evaluate ~GZet. and Qd from equation (6-4) according to the procedure described by Ryum.' In general, a plot of In t' vs JITwill yield a distorted C-curvewith well-defined asymptotes, as shown in Fig. 6.6. At high undercoolings, when ~GZet. is negligible, the slope of the curve becomes constant and equal to QdIR. The mathematical expression for this asymptote is:
------T------eq.
High undercooJing Low diffusivity A Fig. 6.4. Schematic diagram showing the competitive diffusivity on the heterogeneous nucleation rate.
o
influence of undercooling (driving force) and
391
-----------T
eq.
-------------
log
log
Fig. 6.5. Fraction transformed as a function of time referred to the C-curve (schematic).
+ -Qd
RT
(6-5)
At the chosen reference temperature T, the time difference between the real C-curve and the extension of the lower asymptote amounts to (see Fig. 6.6): * -In(t} * ) = __ flGh*e_. t In(t2) RTr from which (6-7) It follows that equations (6-5) and (6-7) provide a systematic basis for obtaining quantitative information about Qd and flGZet. from experimental microstructure data through a simple graphical analysis of the shape and position of the C-curve in temperature-time space. (6-6)
392
In't* .1
Int* .2
- - -:::- - - -: - - - - - - - Teq. - - - - - - - -
---:-:.~-..:----, ,
'.
:
I I I
-:-..:-:=.;:.-
-=-~--....,
!
. .
I I I
--_ ..--- Tr
'-..
I'
I
I I
: '
....
:---~~ : ~ :
T""" I
I I
Int~
Fig. 6.6. Determination of ~GZet. and Qd from the C-curve (schematic).
6.2.2.3 Mathematical description of the C-curve In order to obtain a full mathematical description of the C-curve, we need to know the variation in the energy barrier ~GZet. with undercooling ~T. For heterogeneous nucleation of precipitates above the metastable solvus, the strain energy term entering the expression for ~GZet. can usually be ignored. In such cases the energy barrier is simply given as: 1 AG* het.
=N
.
(6-8)
where NA is the Avogadro constant, "In is the.interfacial energy per unit area between the nucleus and the matrix, AGv is the driving force for the precipitation reaction (i.e. the volume free energy change associated with the transformation), and Sea) ~e so-called shape factor which takes into account the wetting conditions at the nucleus/substrate interface. For a particular alloy, ~Gv is for small 6.T proportional to the degree of undercooling: 1
(6-9)
where c3 is a kinetic constant. This equation follows from the definition of ~G v in diluted alloy systems and the mathematical expression for the solvus boundary in the binary phase diagram. By substituting equation (6-9) into equation (6-8), we get:
A * ilG het. = Ao (Teq.)
21
(Teq. - T)
(6-10)
.It follows that Ao is a characteristic material constant which is related to the potency of the heterogeneous nucleation sites in the material. The value of Ao is, in turn, given by equations (6-7) and (6-10): . (6-11)
393
t*
In cases where Ao is known, it is possible to obtain a more general expression for substituting equation (6-10) into equation (6-3):
by
(6-12)
Equation (6-12) can further be modified to allow for compositional and structural variations in the parent material by using the calibration procedure outlined in Fig. 6.7. Let denote the time taken to precipitate a certain fraction of ~ at a chosen reference temperature T = T, in an alloy containing N; nucleation sites per unit volume. If we take the corresponding solvus temperature of the ~-phase equal to T;q., the expression for becomes:
t;
t;
(6-13)
(6-14)
Equation (6-14) provides a basis for predicting the displacement of the C-curve in temperature-time space due to compositional or structural variations in the parent material. In genT - - - - - - - - - - - T. = T* - - - - - - - - - - - - eq. eq.
tt
Fig. 6.7. Method for eliminating unknown kinetic constant in expression for t',
log t
394
eral, an increase in N; will shift the nose of the C-curve to the left in the diagram (i.e. towards shorter times), as shown schematically in Fig. 6.8, because of the resulting increase in the nucleation rate. Moreover, in solute-depleted alloys the critical undercooling for nucleation will be reached at lower absolute temperatures where the diffusion is slower. This results in a marked drop in N het which displaces the C-curve towards lower temperatures and longer times in the IT-diagram, as indicated in Fig. 6.9.
Example (6.1)
Isothermal transformation (IT) or continuous cooling transformation (CCT) diagrams are available for many of the important alloy systems." In the case of aluminium, so-called temperature-property diagrams exist for different types of wrought alloys.i-" Suppose that the C-curve in Fig. 6.10 conforms to incipient precipitation of J3'(Mg2Si) particles at manganese-containing dispersoids in 6351 extrusions. Use this information to estimate the values of Ao and Qd in equation (6-3) when the solvus temperature of J3'(Mg2Si) is 520C.
Solution
The parameters Ao and Qd can be evaluated from the C-curve according to the procedure shown in Fig. 6.6. Referring to Fig. 6.11, the value of I1G~et. at the chosen reference temperature T, = 350C (623K) is equal to: I1GZet. (at 623 K) = RTr(lnt; -lnt;)
= 8.314 x 623(0.3 + 2.0)1 mol "
=11 9131 mol:" When I1G~et. is known, the parameter Ao can be obtained from equation (6-11):
T -----------Teq.----------
,. ., T.
r
'I
':
~
..( .................
" , : I ....
""'i
t*
= t*r (N* iN ): v v:
I
..
I I
..
~-- -
..
-log t
space (schematic).
t*
t*r
SOLID
STATE TRANSFORMATIONS
IN WELDS
395
(2)
1)
%8
Nhet.
log t
Fig. 6.9. Effect of solute content on the shape and position of C-curve in temperature-time matic).
space (sche-
A
o
(T
eq.
- T
eq.
(T)2
)2
(793)2
= 547J mol:' Similarly, Qd can be read from Fig. 6.11 by considering the slope of the lower asymptote: 0
Qd
8 314[6.
0
2.1-1.8
This value is in good agreement with the reported activation energy for diffusion of magnesium in aluminium.? 500
~ 400
~ ~
::::::J
Q)
l-
E 300
Q)
200 10 Timers~
Fig. 6.10. C-curve for 99.5% maximum yield strength of an AA6351-T6 extrusion. After Staley.'
10
396
. __ ~
I
. +__
I
-..Solvus tem~rature:
520~C
.-.-.-.l~-=~.~~-~-.-.. . --
---
C}i
2.0
.s
1/
.s
!-C\1
In
t ------...
6.2.3 Growth of precipitates If the embryo is larger than some critical size, it will grow by a transport mechanism which involves diffusion of solute atoms from the bulk phase to the matrix/nucleus interface. 6.2.3.1 Interface-controlled growth When transfer of atoms across the a/f3- interface becomes the rate-controlling step, the reaction is said to be interface-controlled. This growth mode is therefore observed during the initial stage of a precipitation reaction before a large, solute-depleted zone has formed around the particles. In the case of incoherent precipitates, the variation in the particle radius r with time is given by:"
r = Ml
-c ex [c C~ -Co
0
(6-15)
where Ml is a mobility term, Co is the concentration of solute in matrix, Ccx is the concentration of solute at the particle/matrix interface, and C~ is the concentration of solute inside the precipitate. In general, the mobility of incoherent interfaces is high, since the solute atoms can easily 'jump' across the interface and find a new position in the particle lattice, as shown schematically in Fig. 6. 12(a). In contrast, a coherent interface is essentially inmobile because transfer in this case involves trapping of atoms in an intermediate lattice position, as indicated in Fig. 6.12(b). As a result, semi-coherent precipitates are forced to grow by lateral movement of ledges along a low energy interface in a direction where the matrix is incoherent with respect to the particle lattice (see Fig. 6.13). In such cases the thickening rate of the precipitates U~/f3 is given by:3,7
397
Incoherent interface
Coherent interface
(a) (b)
Fig. 6.12. Schematic illustration of atom transfer across different kinds of interfaces; (a) Incoherent interface, (b) Coherent interface.
v,.".~
(6-16)
where M; is a new mobility term, and I is the interledge spacing. 6.2.3.2 Diffusion-controlled growth For growth of incoherent precipitates above the metastable solvus, the rate-controlling step will be diffusion of solute in the matrix. If precipitation of the J3-phase occurs from a
398
supersaturated a, the reaction proceeds by diffusion of solute to the growing ~-particle, as shown schematically in Fig. 6.14. On the other hand, when the ~-phase is formed by rejection of solute from the a-phase, the transformation occurs by diffusion of atoms away from the ~particle, as indicated in Fig. 6.15. Aron et al. 8 have presented general solutions for diffusion-controlled growth of both flat plates and spheres under such conditions. In the former case the half thickness IYZ of the plate is given by: (6-17) The parameter t1 in equation (6-17) is frequently referred to as the one-dimensional bolic thickening constant, and is defined as: para-
1
I
-~-------------------A
0108--...
- - C~
~ C Q)
o
c:
Distance Fig. 6.14. Schematic representation of concentration profile ahead of advancing interface during precipitation of ~ from a supersaturated a-phase.
399
e c __ l_ ~
Q)
Diffusion of solute
o c o
- C~
Distance
Fig. 6.15. Schematic representation of concentration profile ahead of advancing interface during growth of solute-depleted ~ into a metastable ex-phase.
(6-18)
where Dm is the diffusivity of the solute in the matrix, and eifc(u) is the complementary error function (defined previously in Appendix 1.3, Chapter 1). Similarly, for growth of spherical precipitates, the variation in the radius r with time can be written as:"
(6-19)
where as:
2
400
METALLURGICAL
MODELLING
OF WELDING
(2)2 exp(~J[exp(
4Dm
4Dm
-t~
J - '" ~
2-y o;
erfC( 2-y~o; Jl
(6-20)
The parabolic relations in equations (6-17) and (6-19) imply that the growth rate slows down as the J3-phase grows. This is due to the fact that the total amount of solute partitioned during growth decreases with time when the diffusion distance increases. Moreover, the form of equations (6-18) and (6-20) suggests that the maximum in the growth rate is achieved at an intermediate temperature because of the competitive influence of undercooling (driving force) and diffusivity on the reaction kinetics. Consequently, a plot of tl or t2 vs temperature will reveal a pattern similar to that shown in Fig. 6.4 for the nucleation rate, although the thickening constants generally are less temperature-sensitive. In addition to the models presented above for plates and spheres, approximate solutions also exist in the literature for thickening of needle-shaped precipitates, based on the Trivedi theory for diffusion-controlled growth of parabolic cylinders.Pllowever, because of space limitations, these solutions will not be considered here. 6.2.4 Overall transformation kinetics The progress of an isothermal phase transformation may be conveniently represented by an ITdiagram of the type shown in Fig. 6.5. Among the factors that determine the shape and position of the C-curve are the nucleation rate, the growth rate, the density and the distribution of the nucleation sites as well as the physical impingement of adjacent transformed volumes. Due to the lack of adequate kinetic models for diffusion-controlled precipitation, we shall assume that the overall microstructural evolution with time can be described by an Avramitype of equation: 10 x = 1- exp(-ktn) (6-21) where X is the fraction transformed, n is a time exponent, and k is a kinetic constant which depends on the nucleation and growth rates. The exponential growth law summarised in the Avrami equation is valid for linear growth under most circumstances, and approximately valid for the early stages of diffusion-controlled growth.l'' Table 6.1 gives information about the value of the time exponent for different experimental conditions. In general, the value of n will not be constant, but change due to transient effects until the steady-state nucleation rate is reached and n attains its maximum value. Subsequently, the nucleation rate starts to decrease as the sites become filled with nuclei and eventually approach zero when complete saturation occurs. This is because the heterogeneous nucleation sites are not randomly distributed in the volume, but are concentrated near other nucleation sites leading to an overall reduction in n. From then on, the transformation rate is solely controlled by the growth rate. 6.2.4.1 Constant nucleation and growth rates For a specific transformation reaction, the value of k in equation (6-21) can be estimated from
40l
Table 6.1 Values of the time exponent n in the Avrami equation. After Christian. to
Polymorphic changes, discontinuous controlled growth, etc. Increasing nucleation rate Constant nucleation rate Decreasing nucleation rate Zero nucleation rate (saturation of point sites) Grain edge nucleation after saturation Grain boundary nucleation after saturation Diffusion controlled growth All shapes growing from small dimensions, increasing nucleation rate All shapes growing from small dimensions, constant nucleation rate All shapes growing from small dimensions, decreasing nucleation rate All shapes growing from small dimensions, zero nucleation rate Growth of particles of appreciable initial volume Needles and plates of finite long dimensions, small in comparison with their separation Thickening of long cylinders (needles) (e.g. after complete end impingement) Thickening of very large plates (e.g. after complete edge impingement) Precipitation on dislocations (very early stages) precipitation, eutectoid reactions, interface n
>4 4
3-4 3
2 1 n 1 >2 '2 1 2 '2 1 1 '2-21'2 1 1 '2 1-11'2 1 1 1'2 -2/3
kinetic theory, using the classic models of nucleation and growth described in the previous sections. In practice, however, this is a rather cumbersome method, particularly if the base metal is of a heterogeneous chemical nature. Alternatively, we can calibrate the Avrami equation against experimental microstructure data, e.g. obtained from generic IT-diagrams. A convenient basis for such a calibration is to write equation (6-21) in a more general form: (6-22) where k* is a new kinetic constant (equal to k-J/n). In the latter equation the parameter k* can be regarded as a time constant, which is characteristic of the system under consideration. Note that this form of the Avrami equation is mathematically more appropriate, as the dimensions of the k * constant are not influenced by the value of the time exponent n. During the early stages of a transformation reaction, the reaction rate is controlled by the nucleation rate. Let 1*denote the time taken to precipitate a certain fraction of ~ (X = X) at an arbitrary temperature T (previously defined in equation (6-14)). It fqllows from equation (6-22) that the value of k' in this case is given as: k*
=
1*[-ln(1-
X)]-lIn
(6-23)
X= 1-ex
[~
-
t 1*[-ln(1-Xc)]-lIn
n]
(6-24)
1 - (1 -
X)
(tlt*r
(6-25)
Equation (6-25) represents an alternative mathematical description of the Avrami equation, and is valid as long as the nucleation and growth rates do not change during the transformation. It has therefore the following limiting values and characteristics: X = 0 when t = 0, X = X; when t = t*, and X~ 1 when t~ 00. 6.2.4.2 Site saturation If the nucleation rate is considered to be zero by assuming early site saturation, the subsequent phase transformation simply involves the reconstructive thickening of the J3-layer. In the onedimensional case, the process can be modelled in terms of the normal migration of a planar a/J3 interface, as shown schematically in Fig. 6.16. Let Aa/~ denote the interfacial area between a and J3per unit volume and Ua/~ the growth rate of the incoherent a/J3-interface. From Fig. 6.16 we see that the volume fraction of the transformed J3-phase is given as: (6-26) By using the standard Johnson-Mehl correction for physical impingement of adjacent transformation volumes, we may write in the general case:
=X J --(1- X)
which after integration yields: X
dX
= Aa/A
tJ
Ua/~ t
(6-27)
U a/~ t)
(6-28)
This specific form of the Avrami equation is valid under conditions of early site saturation where the al f3-interface is completely covered by f3nuclei at the onset of the transformation. 6.2.5 Non-isothermal transformations
So far, we have assumed that the phase transformations occur isothermally. This is, of course, a rather unrealistic assumption in the case of welding where the temperature varies continuously with time. From the large volume of literature dealing with solid state transformations in
Fig. 6.16. Schematic illustration of the planar geometry assumed in the site saturation model.
SOLID
STATE TRANSFORMATIONS
IN WELDS
403
metals and alloys, it appears that the bulk of the research has been concentrated on modelling of microstructural changes under predominantly isothermal conditions.I=v!' In contrast, only a limited number of investigations has been directed towards non-isothermal transformations.5,10,12-18 However, these studies have clearly demonstrated the advantage of using analytical modelling techniques to describe the microstructural evolution during continuous cooling, instead of relying solely on empirical CCT-diagrams. 6.2.5.1 The principles of additivity From the literature reviewed it appears that there is considerable confusion regarding the application of isothermal transformation theory for prediction of non-isothermal transformation behaviour. These difficulties are mainly due to the independent variations of the nucleation and growth rate with temperature. In fact, it can be shown on theoretical grounds that the problem is only tractable when the instantaneous transformation rate is a unique function of the fraction transformed and the temperature. 10 This leads to the additivity criterion described below. The principles of additivity are based on the theory advanced by Schell.'? He proposed that the start of a transformation under non-isothermal conditions could be predicted by calculating the consumption of fractional incubation time at each isothermal temperature, with the transformation starting when the sum is equal to unity. The Scheil theory has later been extended to phase transformations to predict continuous cooling transformation kinetics from isothermal microstructure data.IO,l7,I8 Let t" again denote the time taken to precipitate a certain fraction of ~ (X = Xc) at an arbitrary temperature T. If the reaction is additive, the total time to reach X; under continuous cooling conditions is obtained by adding the fractions of time to reach this stage isothermally until the sum is equal to unity. Noting that t* varies with temperature, we may write in the general case: t
J~-1 (T)
t*
o
(6-29)
,/
<.
Cooling curve
<.
log
404
6.2.5.2 Isokinetic reactions The concept of an isokinetic reaction has previously been introduced in Section 4.4.2.3 (Chapter 4). A reaction is said to be isokinetic if the increments of transformation in infinitesimal isothermal time steps are additive, according to equation (6-29). Christian 10 defines this mathematically by stating that a reaction is isokinetic if the evolution equation for some state variable X may be written in the form:
dX G(X)
dt
H(T)
(6-30)
where G(X) and H(I) are arbitrary functions of X and T, respectively. For a given thermal history, T(t), this essentially means that the differential equation contains separable variables of X and T. 6.2.5.3 Additivity in relation to the Avrami equation The concept of an isokinetic reaction can readily be applied to the Avrami equation. Differentiation of equation (6-22) with respect to time leads to the following expression for the rate of transformation:
dX = n(k * )-n
dt
i":'
(1- X)
(6-31)
(~)[(X
k*
-1)ln(l[-In(I-X)]lln
X)]
In a typical diffusion-controlled nucleation and growth process, the fraction transformed X will not be independent of temperature, since the equilibrium volume fraction of the precipitates decreases with temperature (e.g. see equation (4-7) in Chapter 4). However, for dilute alloys it is a fair approximation to neglect this variation as the solvus boundary becomes increasingly steeper and in the limiting case approaches that of a straight (vertical) line. Thus, if n is constant and k' depends only on the transformation temperature, the reaction will be isokinetic in the general sense defined by Christian." Because of the independent variations of the nucleation and growth rate with temperature, the transformation rate will not be a simple function of temperature. However, by considering the form of the constitutive equations, it is obvious that the change in the nucleation rate with temperature is far more significant than the corresponding fluctuations in the growth rate. This point is more clearly illustrated in Fig. 6.18, which shows the temperature-dependency of the nucleation and growth rates of grain boundary ferrite in a C-Mn steel. It is evident from these- data that the change in the parabolic thickening constant I is negligible compared with the fluctuations in the nucleation rate. Consequently, in transformations that involve continuous cooling it is sufficient to allow for the variation of IV het. with temperature, provided that site saturation has not been reached. Thus, if n is constant we can apply the Scheil theory directly and rewrite equation (6-25) in an integral form: (6-32) In equation (6-32) II represents the kinetic strength of the thermal cycle with respect to (3precipitation. This parameter is generally defined by the integral:
405 10
4
10
f
ICI)
10
10
3 ~
f
w :::t
ICI)
')J
-'CD
-6 2 .s 10
~
c:
0
iC.s::.
.......-- N*
het.
10
E
1i)
<U
c:
0
0 0)
~CD
Z
c: E
(3
:::s
10
10
.-
:2
~0
800
900
Temperature, C -----...
N ~et.
and
El
mation in a C-Mn steel (0.15 wt% C, 0.40 wt% Mn). Data from Umemoto et al.19
(6-33)
where dt is the time increment at T, and t' is the corresponding hold time required to reach Xc at the same temperature (given by equation (6-14)). The derivation of equation (6-32) is shown in Appendix 6.2 The principles of additivity are also applicable under conditions of early site saturation. If only U a/{3 varies with temperature, it is possible to rewrite equation (6-28) in an integral form:
(6-34)
This equation can readily be integrated by numerical methods when the temperature-time programme is known. 6.2.5.4 Non-additive reactions If the additivity condition is not satisfied, it means that the fraction transformed is dependent on the thermal path, and the differential equation has no general solution. This, in turn, implies that the C-curve concept breaks down and cannot be applied to non-isothermal transforma-
406
tions. Solution of the differential equation then requires stepwise integration in temperature-time space, using an appropriate numerical integration procedure. As already pointed out, this will generally be the case for diffusion-controlled precipitation reactions, since the evolution parameter X is a true function of temperature. Under such conditions, experimentally based continuous cooling transformation (CCT) diagrams must be employed.
The microstructural constituents listed above are indicated in Fig. 6.19, which shows photomicrographs of typical regions within low-alloy steel weldments. 6.3.2 Currently used nomenclature Quantification of microstructures in steel welds is most commonly done by means of optical . microscopy. Several systems have been introduced throughout the years for the classification of the various constituents, with each system reflecting different investigator's views and
407
(a)
(b)
(c)
(d)
Fig. 6.19. Optical micrographs showing various microstructural constituents commonly found in lowalloy steel weldments; (a) Coarse grained HAZ (low heat input welding), (b) Coarse grained HAZ (high heat input welding), (c) As-deposited weld metal (low heat input welding), (d) Reheated weld metal (low heat input welding). Letters in micrographs are defined in the text.
discretions. This controversy in terminology has been a source of confusion, and the work by Sub-Commission IX] of the International Institute of Welding (IIW)23 for developing guidelines for quantification of microstructures is, therefore, an important step towards a standardised system of nomenclature. The IIW recommendations are based on the scheme originally proposed by Abson and Dolby.P The IIW system involves a simplified classification procedure compared with the outline used in Fig. 6.19, since the distinction between acicular ferrite and the various sideplate structures is based on features such as aspect ratio, relative lath size, and number of parallel
408
laths. This has led to the introduction of the FS-constituent (ferrite with aligned second phase), which, in principle, comprises both Widmanstatten ferrite and upper bainite. In contrast to the IIW approach to classifying microstructural elements based on their appearance in the optical microscope, other investigators rank the various constituents solely in terms of their transformation behaviour, according to the scheme originally proposed by Dube et al. 26 From a scientific point of view, this classification system is more correct, since it does not violate common terminology based on thermodynamics and kinetics of transformation reactions. However, with the omission of the FS-constituent grouping utilised by the IIW, the Dube system is more inconvenient to use in practice because the different transformation products often cannot readily be identified on the basis of their transformation characteristics. Consequently, both classification systems appear to have their weaknesses, which, in turn, limit their applicability. 6.3.3 Grain boundary ferrite Grain boundary (or allotriomorphic) ferrite is the first phase to form on cooling below the Ae3temperature. It nucleates preferentially at austenite grain comers and boundaries, since these sites generally provide the lowest energy barrier against nucleation (see Fig. 6.3). The fundamental aspects of grain boundary ferrite have been reviewed in detail by Bhadeshia.I" where many of the original references can also be found. 6.3.3.1 Crystallography of grain boundary ferrite The grain boundary ferrite allotriomorphs nucleate having a Kurdjumow-Sachs tion relationship with one of the austenite grains.?" {lIlly-Fe 11{IIOla_Fe
type orienta-
<IIO>r-Fe
11<III>a_Fe
This orientation relationship, which lies within the so-called Bain orientation region.F is adopted in order to minimise the increase in the strain energy resulting from lattice distortion dGE by formation of a low-energy interface between the ferrite nucleus and the parent austenite phase. 1 Subsequent growth of the ferrite may then occur into the adjacent austenite grain with which the ferrite has a random orientation relationship.P since a disordered (incoherent) interface generally has a higher mobility than an ordered (coherent/semi-coherent) interface at low undercoolings. 6.3.3.2 Nucleation of grain boundary ferrite As shown in Fig. 6.20, allotriomorphic ferrite can nucleate both at grain comers, grain edges and grain faces, the former ones being the most potent sites for ferrite nucleation. 1 If we assume that ferrite nucleation occurs preferentially at austenite grain I N v_* term _ faces, _* the N* .Y..... in equation (6-14) may be replaced by the inverse grain size ratio Dr I Dr, where Dr and Dr refer to the austenite grain size in the actual and the reference material, respectively. This leads to the following expression for t':
(6-35)
409
rner
Fig. 6.20. Sketch of an austenite grain showing different sites for ferrite nucleation.
Based on equation (6-35) it is possible to predict the displacement of the C-curve in temperature-time space due to structural or compositional variations in the parent material. As an illustration, we shall assume that the parameters listed in Table 6.2 are representative of nucleation of grain boundary ferrite in a low-alloy steel with an initial austenite grain size of 10 um, In addition, we need information about the Ae3-temperature in the equilibrium phase diagram. This temperature can readily be obtained from thermodynamic calculations, even for multicomponent systems.i? Alternatively, we can use the empirical relationship quoted by Leslier'? Ae3 (OC) = 910 - 203~%C +31.5(%Mo) - 30(%Mn) -15.2(%Ni) (6-36) - 20(%Cu) + 44. 7(%Si) + 104(% V) + 13.1(% W) + 700(% P) +400(% AI) + 400(%Ti) + 120(% As) where all compositions are given in weight %. By lowering the Ae3-temperature, the austenite stabilising elements C, Mn, Ni and Cu reduce the undercooling and hence, the driving force for transformation of austenite, at any lower temperature. On the same basis, the ferrite stabilising elements will assist the transformation because they increase the undercooling. In particular, the very large ferrite stabilising effects of P, Al and Ti are obvious from the above relationship. Effect of austenite grain size A cl~er i.!!.syectionof equation (6-35) reveals that the positi~l! of th~ C-curve shifts by a factor of (Dy / Dy) when the austenite grain size changes from D; to Dy. This is due to the fact that the austenite grain boundary area per unit volume is inversely proportional to the grain size. Consequently, grain growth which occurs during welding will have a profound effect on the subsequent transformation behaviour of the weld. An illustration of this point is given below.
Table 6.2 Input data used to construct C-curve for allotriomorphic
Parameter Tr (K) t, (s)
T;q.
(K)
A0
(J mol:')
o:
* Dy
(J.1m)
Value
823
0.6
1108
700
135
10
410
Example (6.2) ,
Consider SA welding on a thick plate of a Nb-microalloyed steel under the following conditions: 1= 550A, U = 32V, v = 4mm S-I, T) = 0.95, To = 20C Evaluate on the basis of the grain growth diagram in Fig. 5.21(a) (Chapter 5) and equations (6-32), (6-33), and (6-35) the conditions for ferrite formation at two different positions within the HAZ corresponding to a peak temperature Tp of 1350C and 1000C, respectively. Assume in these calculations that the ferrite may form within the temperature range from 800 to 600C, and that the equilibrium volume fraction of ferrite (fr;q~Fe) in the fully transformed steel is 0.9.
Solution
First we calculate the net heat input per unit length of the weld:
Iv=
qo
TlxlxU
= 0.95x550x32kJmm-1
1000xv
1000x4
~4.2kJmm-'-1 Readings from Fig. 5.21(a) give the following HAZ grain sizes:
Tp = 1000C:
T =
1350C:
Dr ~
By substituting data from Table 6.2 into equation (6-35), we obtain: Grain refined region ( Dr = 10J1m):
t*
10
=0
6ex [700 ( (1108)2 _ (1108)2 + 135 X 10 P R T(1108 - T)2 823(1108 - 823)2 R region ( Dy
(1.__
T
1_)] 823
Grain coarsened
150 J1m):
* = tlO * (150) tl50
10
* = 15tlO
As expected, the theoretical C-curves in Fig. 6.21 reveal a strong effect of the austenite grain size on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it is possible to estimate the volume fraction of ferrite fa.-Fe which forms in each case from equations (6-32) and (6-33). Taking the time exponent n in the Avrami equation equal to 5/2 for nucleation of ferrite at austenite grain boundaries'? and 1 - X c = 0.98, we obtain: . 0 Grain refined region ( Dr = 10 11m):
411
o o
~ 600
/'
,....... -._.-.-.-
/'
~ ~
:::l
Q)
"
"
.,.......
.....................
~ 500
Q)
400
oy = 101lm~
0.1 0.3
.>.,
3
,.,
300~--~~~~~--~~~~~--~~~~~--~~~~~
10 Time, s
30
100
300
1000
Fig. 6.21. Effect of austenite grain size on the HAZ transformation kinetics (Example 6.2). The superimposed cooling curve corresponds to a cooling time, t,.tS/5 , of 21s.
and
r:"
= fr:q~Fe
(1- 0.981) )
= 0.90(1-
0.98623)
0.90
= 150 um):
5/2
II = It2
[
t)
-$t150 ]
= 0.65
and
fa-Fe
= fr:q~Fe (1- 0.981) )
= 0.90(1-
0.98.65) ~ 0.01
From this we see that polygonal ferrite dominates the microstructure within the grain refined region, whereas ferrite hardly forms within the grain coarsened HAZ under the prevailing circumstances. Although experimental data are not available for a direct comparison, the predicted effect of the austenite grain size on the HAZ transformation kinetics is reasonable and consistent with general experience (e.g. see experimental CCT-diagrams in Fig. 6.22). Effect of austenite and ferrite stabilising elements In practice, the transformation behaviour of steel weldments is affected both by the prior austenite grain size and by alloying additions. In welding metallurgy, the combined effects are often discussed byconsidering their influence on hardenability. Broadly speaking, additions of hardenability elements may serve two purposes: (i) (ii) To ensure the desired strength level by solid solution or precipitation strengthening; To control the microstructure through modification of the nucleation and growth rates of proeutectoid ferrite.
412
t
Pi ;::,
Q)
800
o 700 o
~ 600
E Q) r-
0-
500
400
Martensite
-1
300
10
Fig. 6.22. CCT-diagrams for a low-carbon Cu-Ni containing steel. Superimposed on the CCT-diagrams are two cooling curves corresponding to /).t8/5 of 10 and 100 s, respectively. Austenitising conditions; Heavy solid lines: 900C for 5 min, Heavy broken lines: 1300C for 5 s. Data from Cross et al."
In the latter case the effect is related to a shift in the Ae3-temperature of the steel, which alters the undercooling and hence, the driving force for the austenite to ferrite transformation. This point is illustrated by the following example.
Example (6.3)
Consider SA welding on a thick plate of a Nb-microalloyed steel under conditions similar to those employed in Example (6.2). Based on equations (6-32), (6-33), and (6-35) estimate the volume fraction of grain boundary ferrite in the grain coarsened HAZ (Tp = 1350C) after welding when the Ae3-temperature of the steel is 863C (1136K).
Solution
By substituting data from Table 6.2 into equation (6-35), we arrive at the following expression for t~50:
t* =0 6(150)ex 150 . 10
[700( P R
135'><10 (1 + R
T - 823
1)]
It is evident from the graphical representation of the above equation in Fig. 6.23 that an increase in the Ae3-temperature (e.g. from 1108 to 1136K) displaces the C-curve towards higher temperatures and shorter times in the IT-diagram. This, in turn, gives rise to improved conditions for ferrite nucleation. Taking the time exponent n in the Avrami equation equal to 5/2 and (1 - X) = 0.98 as in the previous example, we obtain after integration:
SOLID
STATE TRANSFORMATIONS
IN WELDS
413
800
r
o
0
700
~ 600 .3
J-
E 500
Q) Q)
0.3
10
30
100
300
1000
Time,s -----Fig. 6.23.Effect of steel chemical composition (Aej-temperature) on the HAZ transformation (Example 6.3). The superimposed cooling curve corresponds to a cooling time, I1t8/5, of 2Is.
5/2
kinetics
II =
[
St2
t\
-$t150 ]
= 4.0
and
Although the calculated value offa-Fe is rather uncertain, the trends predicted in the present example are reasonable and consistent with general experience. Effect of boron alloying The role of boron in steel transformation kinetics has been a subject of research for many years. A number of hardenability mechanisms have been proposed to explain the behaviour of boron in steels. Of these, only four have survived to the present.F All assume that boron influences hardenability by increasing the energy barrier against ferrite nucleation at austenite grain boundaries * and that it does not influence the thermodynamic properties of the austenite and ferrite phases, i.e. reduces the Ae 3-temperature of the steel. 32 Moreover, it is now generally accepted that impurity-boron interactions are responsible for most of the observed erratic behaviour of boron steels with respect to hardenability.P'<" In particular, if the steel contains small amounts of dissolved oxygen and nitrogen, these elements may combine with boron to reduce the free, diffusible boron content. 32 The hardenability effect of boron is therefore sensitive to variations in the applied steel deoxidation and alloying practice.32-37
*Based on classic nucleation theory it can be argued that a decrease in the 'Y-'Ygrain boundary energy due to boron segregations will suppress the formation of ferrite at these sites. Further enhancement of the energy barrier against ferrite nucleation through an increase in the total strain energy of the embryo is possible if the free grain boundary volume becomes filled with either boron atoms or borocarbide precipitates (e.g. Fe23(B,C)6)'
414
During welding, quantitative information about the extent of boron segregation which occurs to the austenite grain boundaries under various thermal programmes can be obtained on the basis of a well established theoretical model for quench-induced segregation of boron in stee1.38,39 At peak temperatures above 1000 to 1100C, borocarbides and -nitrides present in the base plate will rapidly dissolve in the matrix,40,41leading to a significant increase in the amount of free diffusible boron. Generally, solute atoms in a crystal lattice will have an associated strain energy," which implies that it is energetically feasible to pair the solute boron atom with a vacancy. Since the formation of vacancies is a thermally activated process, it follows that the fraction of boron occupying such sites, [Blv, increases exponentially with temperature."
[B] =m[B]exp(-Ef
v
RT
+~J
RT
(6-37)
Here m contains various geometric and entropy terms, Ef is the vacancy formation energy, Eb is the vacancy-boron binding energy, and [B] is the bulk concentration of free boron. By substituting reasonable average values for m, Ef, and Eb into equation (6-37), we arrive at the following expression for [B lv:39,42
[BJvCppm)::::48[B]exp ( -~86850)
(6-38)
During cooling, the vacancy concentration initially established at elevated temperatures tends to readjust by elimination of excess point defects at grain boundaries through diffusion. This, in turn, results in an associated flow of solute boron to the austenite grain boundaries (the extent of which is controlled by the peak temperature and the bulk boron concentration), provided that the diffusi vity of the boron-vacancy complexes is higher than that of the vacancies and the boron atoms at all relevant temperatures. Consider now the limiting case where the boron-vacancy complex diffusion to the grain boundaries occurs sufficiently rapid to keep pace with the falling temperature; i.e. the equilibrium concentration, [B]v, is maintained from the peak temperature Tp down to the start temperature of the austenite to ferrite transformation Ae3. In view of the high diffusivity of boronvacancy complexes in austenite, this is not an unrealistic assumption when the cooling rate is of the order of 50C S-1 (representative of a weld HAZ).39 Under such conditions, the amount of boron which segregates to the austenite grain boundaries within the temperature interval from Tp to Ae3 on cooling, [B]gb' can approximately be written as:42
when Tp Ae3. In addition to the diffusible boron content, the grain size is also an important variable in steel hardenability. A quantitative estimate of the combined effect of boron segregations and
415
austenite grain size on the HAZ transformation kinetics can be obtained by assuming that the ferrite nucleates primarily on grain faces. In such cases the total number of heterogeneous nucleation sites per unit volume N; is given as: (6-40) where na is the number of nucleation sites per unit grain boundary area, and S; is the grain boundary surface area per unit volume (equal to 2IDy). If we, as a second approximation, assume that na is inversely proportional to the amount of boron which diffuses to the grain boundaries on cooling, the N: I N; ratio in equation (6-14) can be written as: (6-41) Equation (6-41) predicts that the fosition of the ferrite C-curve in temperature-time space depends on the ([ B]gb Dr ) I ([B];b D~ ratio, as shown schematically in Fig. 6.24. Consequently, this ratio can be regarded as a measure of the HAZ hardenability during welding of boroncontaining steels. In Fig. 6.25 the microstructure data of Akselsen et al. 42 have been replotted vs the hardenability parameter [B]gb Dy, taking the product [B];b D~ in the reference steel equal to unity for a direct comparison between theory and experiments. It is evident from the graph that the HAZ martensite content of the two boron-containing steels can be represented by one single curve under the prevailing circumstances. This result is to be expected if the displacement of the ferrite C-curve in temperature-time space is determined by a relationship of the type shown in equation (6-41).
I I
~~
'Y
""
...
Cooling curve
log time
Fig. 6.24. Effect of boron alloying on the shape and position of ferrite C-curve in temperature-time according to the site blocking mechanism (schematic). space
416
100
1
Q)
75
'Ci)
50 25
~ ~ ~
10 100 Hardenability parameter ([8) between the volume fraction of martensite in urn, [B];b
is ) gb 'Y
..
1000
Dr for simulated
Dr
D~ = 1).
parameter
Example (6.4)
Consider GMA welding on a thick plate of a boron-containing steel under the following conditions: I = 300A, U = 30V, v = 4mm s', 11 = 0.8, To = 20C Suppose that the free (diffusible) boron content of the steel at elevated temperatures is 40 ppm. Estimate on the basis of the theory outlined in the previous sections the conditions for ferrite/martensite formation in the grain refined HAZ (Tp = 1100C) when the austenite grain size is 15urn, In these calculations we shall assume that the [B];b D~ product in the reference steel (with thermodynamic properties as in Table 6.2) is close to unity.
Solution
First we estimate the amount of boron which segregates to the austenite grain boundaries during cooling from equation (6-39). When Tp = 1100C (1373K), we get:
[B]
::::48[40]exp
The resulting displacement of the ferrite C-curve can now be calculated from equation (614), using input data from Table 6.2:
417
__ I_)]
T 823
t*
[B]
=06(143)ex
.
[700(
R
(1108)2
T(1108T)2
(1108)2 823(1108-823)2
J+135XI03(~
R
As expected, the theoretical C-curves in Fig. 6.26 reveal a strong effect of boron alloying on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it is possible to estimate the volume fraction of ferrite [ex-Fe which forms from equations (6-32) eq . and (6-33). Taking the time exponent n in the Avrami equation equal to 5/2 for nucleation of ferrite at austenite grain boundaries and (1 - X) = 0.98 as in the previous examples, we obtain after integration within the temperature range from 800 to 60Q C:
O
5/2
1=
1
JI'..!f!--[
t, t[B] ]
~O.09
and
09
)::=
[B]gb
If the same calculations are performed for the reference steel in Table 6.2 (characterised by = [B];b)' we get:
= --t[B]
1.5 * 14.3
= 0.10t[B]
and
26)
::=
0.37
From this we see that boron, even in small quantities, can have a dramatic effect on the HAZ transformation kinetics by promoting the formation of bainite and martensite at the expense of grain boundary ferrite. This is in good agreement with general experience (see experimental CCT-diagrams in Fig. 6.27). Effect of solidification-induced segregation As shown in Chapter 3 of this textbook, the characteristic solidification pattern of mild and low-alloy steel weld metals leads to extensive segregation of alloying and impurity elements to grain boundaries and interdendritic spaces. Of particular interest in this respect is phosphorus segregations at columnar austenite grain boundaries. Referring to Figs. 3.38 and 3.39 (Chapter 3), the phosphorus-rich zone is seen to extend about 7J.lm on either side of the grain boundaries. Since phosphorus is among the strongest ferrite stabilising elements in steel (see equation (6-36)), the existence of such solidification-induced segregations would be expected
418 800
METALLURGICAL
MODELLING
OF WELDING
0 0
1
Q)
700
600 ~ ::l ~
c..
I-
E Q)
500
400
300
0.1
Time, s -----
Fig. 6.26. Predicted displacement of ferrite C-curve in temperature-time space due to segregation of boron to prior austenite grain boundaries (Example 6.4). The superimposed cooling curve corresponds to a cooling time, fltS/5' of 9.2s.
to enhance the nucleation rate of grain boundary ferrite at these sites due to the associated increase in the Ae3-temperature. The above phenomenon should not be confused with equilibrium segregation of phosphorus to austenite grain boundaries during heat treatment of steel, which stems form attraction of the atoms towards the open structure of the boundary. In the latter case phosphorus may act as a hardenability element by occupying favourable sites for ferrite nucleation at the austenite grain boundaries analogous to that documented for boron in steel. 43,44
Example
(6.5)
Suppose that the local phosphorus content adjacent to the columnar austenite grain boundaries in a low-alloy steel weld metal is 500 ppm, whereas the bulk concentration of phosphorus is 100 ppm. Estimate on the basis of the theory developed in the previous sections the resulting displacement of the ferrite C-curve in temperature-time space when the columnar austenite grain size is 80J.lm. In these calculations we shall assume that the transformation characteristic of the bulk metal is similar to that of the reference steel in Table 6.2.
Solution
First we estimate the actual Ae3-temperature within the phosphorus-rich region adjacent to the columnar austenite grain boundaries from equation (6-36). Taking the Ae3-temperature of the bulk phase equal to 835C (1108K), the local phase boundary temperature becomes:
r.; =
T;q.
X 0.04)K
1136K
By substituting data from Table 6.2 into equation (6-35), we arrive at the following expres. c * sion lor t80 :
t*
80
J + 135
10
(!__
T
1_)] 823
419
700 ~ 600
.......................................................
F .
1U 500 Ci3 c.
~ r400
~ ~
- I'
M
---Tp !:l.!1350C
............ 10
Cooling
Tp ~1000 C
100
time,l:l.tB/5,
5 ~
(b)
800
0 a
1
~
700
600
PI .a
a.
Q)
WF/UB
500
M
Tp ~1350 C Tp !'l1000 C
200~----~--------~--------------~--~--~
10
100
Cooling time, A tB/s, 5 Fig. 6.27. CCT-diagrams for boron-containing Data from Akselsen et al. 42 steels; (a) Steel A (llppm B), (b) Steel B (26ppm B),
It is evident from the graphical representation of the above equation in Fig. 6.28 that the observed increase in the Ae3-temperature from 1108 to 1136K displaces the ferrite C-curve towards higher temperatures and shorter times in the IT-diagram. The resulting effect on the weld metal transformation kinetics is obvious, since an increase in the nucleation rate of ferrite will favour early site saturation at the austenite grain boundaries. On this basis it is not surprising to find that allotriomorphic ferrite in low-alloy steel weld metals tends to form continuous veins of blocky ferrite along the columnar austenite grain boundaries, as shown in Fig. 6.19(c). Outside the fusion zone the conditions for early site saturation are less favourable, since modern steelmaking practice implies that solidification-induced segregations are removed by prolonged high-temperature annealing prior to the welding operation. Hence, ferrite veining of the type shown in Fig. 6.19(c) is not commonly observed within the reheated regions of the base plate, unless the heat input is extremely large (see Fig. 6.19(a) and (bj),
o
0
- 600 ~ ::s ~
CJ.)
E 500
Q)
t-
0.3
10
30
100
300
100
Time,s---Fig. 6.28. Effect of solidification-induced phosphorus segregations on the austenite to ferrite transformation in low-alloy steel weld metals (Example 6.5).
Effect of cooling rate onferrite grain size In steel metallurgy it is well accepted that accelerated cooling refines the ferrite grain size, and thus improves both strength and toughness of the parent material. 19,20 If the austenite is allowed to recrystallise before it transforms, the final ferrite grain size Da in the base plate will be an unique function of the prior austenite grain size Dr and the cooling rate through the critical transformation temperature range for ferrite formation (C.R.). The fundamental aspects of ferrite grain refinement by accelerated cooling have been considered by Umemoto et ale 19 By allowing for the variation in the nucleation and growth rates of allotriomorphic ferrite with temperature (see Fig. 6.18), they arrived at the following theoretical expression for Do. when austenite grain faces are the dominant nucleation site of ferrite:
-
(6-42)
where Dr is in urn and (C.R.) is in C s'. In practice, Umemoto et al.19 observed a discrepancy between theory and experiments due to competitive nucleation of ferrite at austenite grain edges and corners (characterised by a theoretical grain size dependence of Dr2/3 and Dr, respectively). Consequently, the real ferrite grain size in the steel varied with austenite grain size and cooling rate as:
Da.(Jlm)=c4 (-
Dr
)0.46
(C.R.)
-026
(6-43)
where C4is a kinetic constant which is characteristic of the alloy system under consideration. An indication of the applicability of equation (6-43) to welding can be obtained from the microstructure data of Evans,45 reproduced in Fig. 6.29. In this plot, the reported heat inputs have been converted into an equivalent cooling rate at 700C via equation (1-71) (Chapter 1). It is evident from the figure that the observed variation in the ferrite grain size with cooling rate is consistent with calculations based on equation (6-43). Hence, the model ofUmemoto et
421
10
'Ci)
C
'~
cD
:::i
C')
$ .;::
LL
Q)
o
0 20
wt% wt%
C, 0.58 C, 1.33
60
80
Cls
Fig. 6.29. Variation of ferrite grain size with cooling rate in reheated C-Mn steel weld metals, Data from Evans.P
al.19 can readily be employed for prediction of the ferrite grain size within the grain refined region of both single pass and multipass steel welds, provided that the austenite grain size and the cooling rate can be estimated with a reasonable degree of accuracy.
Example (6.6)
Consider multipass welding with covered electrodes (SMAW) on a thick plate of low-alloy steel under the following conditions: I
= 230A,
= 23V, v = 4mm
S-I,
'Tl
= 0.8,
To
= 20C
Previous experience has shown that polygonal ferrite forms within the low-temperature reheated regions of the weld metal, typically 1.0 to 1.7 mm beneath the surface (fusion boundary) of subsequent weld passes. Estimate on the basis of the grain growth diagram in Fig. 5.24(a) (Chapter 5) and equation (6-43) the maximum variation in the ferrite grain size across the weld HAZ under the prevailing circumstances. Thermal data for low-alloy steels are given in Table 1.1 (Chapter 1).
Solution
First we need to convert the depths at which polygonal ferrite appears beneath the fusion boundary to an equivalent (characteristic) peak temperature range. If we neglect the contribution from heat flow in the welding direction, this conversion can be done on the basis of equation (5-47) in Chapter 5:
~T
III
p -
20
.)1500
422 Taking
~r:
and
Since the cooling rate (C.R.) at a given temperature is essentially the same across the weld HAZ, the maximum variation in the ferrite grain size can be evaluated directly from equation (6-43) without further background information:
-
Da(max) Da(min)
=(lZ)' 10
046
=1.3
From this we see that the variation in the ferrite grain size is significantly smaller than the corresponding change in the prior austenite grain size. This result is in good agreement with general experience. 6.3.3.3 Growth of grain boundary ferrite If the austenite grain boundaries become rapidly decorated with a continuous layer of ferrite (so that the subsequent transformation involves the reconstructive thickening of these layers), the evolution of allotriomorphic ferrite is determined solely by its growth kinetics.F As shown in Example 6.5, this is a realistic assumption in the case of as-deposited steel weld metals, where the presence of solidification-induced phosphorus segregations at the columnar austenite grain boundaries favours an early site saturation. Several investigators have modelled the evolution of allotriomorphic ferrite in low-alloy steel weld metals along the lines indicated above.46-48 The most thorough analysis is probably that of Bhadeshia et al. 46 who were able to account for the combined effect of temperature and steel chemical composition on the growth kinetics. However, in order to illustrate the competition between the different variables that contribute to the formation of grain boundary ferrite in as-deposited steel weld metals, the simplified treatment of Liu and Olson'? and Fleck et al. 48 has been adopted here. As a starting point the Avrami equation'? is considered for the limiting case of zero nucleation rate by assuming early site saturation. If subsequent growth of the grain boundary ferrite allotriomorphs occurs in both directions perpendicular to the austenite grain boundaries at a rate which is controlled by diffusion of carbon in austenite, the time dependence of the ferrite thickness, L\Zu, is given by equation (6-17): (6-44)
423
where 1 is the one-dimensional parabolic thickening constant (defined in equation (6-18)). The corresponding growth rate of allotriomorphic ferrite U a is then obtained by differenti. ating equation (6-44) with respect to time:
Ua
(6-45)
From a stereological standpoint, the grain boundary surface area per unit volume S; cannot be calculated without further assumptions regarding the shape of the columnar austenite grains. Bhadeshia et ale 37,46,49 solved this problem by representing the grain morphology by a uniform, space-filling array of hexagonal prisms. However, with the precision aimed at here, it is sufficient to assume that S; is equal to the surface area per unit volume of an inscribed cylinder whose volume is equivalent to that of the hexagonal prisms (see Fig. 6.30). Noting that each grain boundary is shared by two adjacent grains, the expression for S; (in the absence of end effects) becomes:47,48 (6-46)
where dy and l-yare the diameter and length of the inscribed cylinder, respectively. By substituting the above expressions for Ua and S; into equation (6-34), we obtain:
(6-47)
In practice, the one-dimensional parabolic thickening constant in equation (6-47) varies both with temperature and steel chemical composition, as shown by the data of Bhadeshia et al. 46 reproduced in Fig. 6.31. However, if we instead use a reasonable average value for tl within the characteristic transformation temperature range for allotriomorphic ferrite (i.e. tl = 1), the integral 12 in equation (6-47) has the following analytical solution:
x Growth direction of columnar grains
Fig. 6.30. Hexagonal prism model for the columnar austenite grain morphology in low-alloy steel weld metals. The inscribed cylinder in the figure has approximately the same surface to volume ratio as the hexagonal prisms.
424
800
Temperature, "C -----Fig. 6.31. Effect of temperature and steel chemical composition on the one-dimensional parabolic thickening constant for allotriomorphic ferrite. (1): 0.03 wt% C, (2): 0.06 wt% C, (3): 0.08 wt% C and (4): 0.10 wt% C. Data from Bhadeshia et al.46
dy
f
t
tl ()-1/2 t
d t------- 21 {i
dy
(6-48)
r:" /r:"
= =
X 1- 2~;J
exp( -
(6-49)
Equation (6-49) predicts that the volume fraction of grain boundary ferrite in the as-deposited weld metal depends on the combined action of the following three main variables: (i) The one-dimensional parabolic thickening constant 1 which is determined by the weld metal chemical composition (i.e. the content of austenite and ferrite stabilising elements).
(ii) The columnar austenite grain size dy which is controlled by the weld metal solidification microstructure (i.e. the weld metal chemistry, the weld pool geometry, and the thermal conditions under which solidification occurs). (iii) The retention time within the critical transformation temperature range for allotriomorphic ferrite, as determined by the applied heat input and the mode of heat flow (i.e. thick plate, medium thick plate, or thin plate welding, respectively).
SOLID
STATE TRANSFORMATIONS
IN WELDS
425
It can be seen from the microstructure data of Grong et al. 50 reproduced in Fig. 6.32 that the influence of these variables are adequately accounted for in the present model. However, the calculated volume fractions of grain boundary ferrite are consistently lower than the measured ones. This discrepancy can probably be attributed to the use of constant values for the onedimensional parabolic thickening constant. Consequently, if proper corrections are made for the inherent variation in tl with temperature and steel chemical composition, the agreement between theory and experiments is significantly improved, as shown by the data of Bhadeshia et al.46 reproduced in Fig. 6.33.
Example (6.7)
Consider GMA welding on a thick plate of a low-alloy steel under the following conditions: I
= 400A,
= 30V, v = 4mm
= 200C
Suppose that the grain boundary allotriomorphs form within the temperature range from 750 to 600C at a constant rate 1 of 3 urn s-1I2. Estimate on the basis of the theory developed in the previous section the volume fraction of allotriomorphic ferrite in the weld deposit when the columnar austenite grain size is 80Jlm. In these calculations we shall assume that the equilibrium volume fraction offerritef~~Fe in the fully transformed steel is 0.9. Thermal data for low-alloy steels are given in Table 1.1 (Chapter 1).
Solution
The situation is described in Fig. 6.34. In this case the problem is to estimate the retention time ~tr within the critical temperature range for ferrite formation from the Rosenthal thick plate solution (Chapter 1). From equation (1-66), we have:
u,
C!J
0.4
Regression line
0.3
15 ~o 0.2 ~ ....
E :J (5
Q)
>
0.1
Mn content: 0.33-2.72
wt%
0.15
0.05
X=1-exp(
-2
dy
Vf
0.10
)
---II
Fig. 6.32. Experimental verification of equation (6-49). Data from Grong et al.50
426
u,
0.6
0
c
~ ~~
0.4
(I)
(5 >
:::J
"0
~ :J en ('($
Q)
0.2
Q)
'::s
Q)
Co
..-
E Q)
~Lltr~
I
I I I I I
log time
Fig. 6.34. Conditions for allotriomorphic
SOLID
STATE TRANSFORMATIONS
IN WELDS
427
= (400X30
xO.8 / 4)(
750-600
(750 - 200)(600 - 200)
21t x 0.025
JS = 10.4s
Although experimental data are not available for a direct comparison, the calculated value of fa-Fe is reasonable and of the expected order of magnitude. 6.3.4 Widmanstiitten ferrite With increasing degree of undercooling the redistribution of carbon becomes insufficient to maintain a planar growth mode, and hence, further growth of the ferrite can only take place by lateral movement of ledges along a low-energy interface *.1,20 This, in turn, implies a Kurdjumow-Sachs-type orientation relationship between the austenite and the ferrite, i.e. {Ill }-y-Fe parallel with {110}a_Fe and <110>-y_Fe parallel with <lll>a_Fe, which is a characteristic feature of the Widmanstatten ferrite structure. The ferrite sideplates, once nucleated, grow very rapidly under the prevailing conditions because of an efficient redistribution of carbon to the sides of the advancing interface. 1 As a result, parallel arrays of ferrite laths of high aspect ratios (typically 10: 1 to 20: 1) are often found in the areas adjacent to the austenite grain boundaries, as shown by the optical micrographs in Fig. 6.19(b) and (c). Following the treatment of Bhadeshia et al.,46 the lengthening rate of Widmanstatten ferrite can be estimated using the Trivedi theory? for diffusion-controlled growth of parabolic cylinders with correction for the assumed displacive character of the Widmanstatten ferrite transformation. As shown in Fig. 6.35, the pertinent growth rates are normally so large that the formation of Widmanstatten ferrite is essentially complete within a fraction of a second. This implies that the transformation, for all practical purposes, can be treated as being isothermal. 37 When the growth rate is known, the volume fraction of Widmanstatten ferrite in the asdeposited weld metal may be estimated by assuming that the ferrite sideplates nucleate at a constant rate at the 'Y Fe /uFe boundaries and subsequently grow into the interior of the columnar austenite grains until they physically impinge with intragranularly nucleated acicular ferrite or allotriomorphic ferrite." It is seen from the microstructure data of Bhadeshia et al. 46 reproduced in Fig. 6.36 that the calculated volume fractions of Widmanstatten ferrite are in reasonable agreement with experiments, although the discrepancy in certain cases is admittedly large.
*A different view is suggested by Bhadeshia et ale 37,46,51,52 who claim that growth of Wi dmanstatten ferrite occurs in a displacive manner analogous to that documented for martensite in steel, with the exception that carbon must diffuse during growth.
428
600--------------~--------------~
t
lL
~lb
-
500 ...
~ ;. 400 r0-
3:
Q)
15 300 ~
..c
~e 200
C!J
I-
2 3 2
100'"""
-4~
3 ---4
I
I
540
580
620 540
Temperature,
580
620
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in C-Mn steel weld metals; (a) Growth rate calculations for weld compositions listed in Fig. 6.31 , (b) Growth rate calculations after modifying compositions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia et al.46
6.3.5 Acicular ferrite in steel weld deposits Simultaneously with or immediately after the formation of Widmanstatten ferrite at the austenite grain boundaries, acicular ferrite may start to nucleate intragranularly at non-metallic inclusions. This phase is commonly observed in low-alloy steel weld metals, where the fine dispersion of oxide inclusions provides favourable sites for heterogeneous nucleation.36,53 There seems to be general agreement that microstructures primarily consisting of acicular ferrite provide optimum weld metal mechanical properties, both from a strength and toughness point of view, by virtue of its small lath size and high dislocation density.36,53,54Consequently, the formation of this particular microconstituent is of significant commercial importance and has therefore attracted substantial research interest over the years.36,37,46-58 In spite of all this effort, the acicular ferrite transformation in low-alloy steel weld metals is still a subject of considerable controversy. 6.3.5.1 Crystallography of acicular ferrite It is well established that acicular ferrite nucleates in the transformation temperature range between Widmanstatten ferrite and lower bainite. 55-59Based on conventional diffraction pattern analyses in the transmission electron microscope (TEM), Bhadeshia et al. 54,59 have shown that the acicular ferrite plates exhibit an orientation relationship with the austenite grain in which they grow. The observed orientation relationship lies within the Bain orientation region?? and can approximately be described as:
429
o
u.. 0.6
c
$: ~ o
o
~ ro ~ 0.4
Q)
::l (5 >
"0
E ~
ca ~
CJ)
:J
0.2
Q)
0.2
0.4
Calculated volume fraction of WF ---Fig. 6.36. Comparison between measured and predicted volume fractions of Widmanstatten ferrite in CMn steel weld metals. Data from Bhadeshia et al. 46 (IOI)a-Fe 11(1Il)y_Fe
[lIlla-Fe II[IIO]y_Fe
This corresponds to the well-known Kurdjumow-Sachs (K-S) orientation relationship * and suggests that growth of the AF plates occurs either by a ledge mechanism or by a pure shear transformation similar to that reported for upper bainite in steel. 1,20,60-62 6.3.5.2 Texture components of acicular ferrite Kluken et al.63 have studied the development of transformation textures in as-deposited steel weld metals containing acicular ferrite, using the electron backscattering pattern (EE SP) technique. Referring to the (200) stereo graphic projection in Fig. 6.37, the measured acicular ferrite orientations form a symmetrical and consistent pattern and disclose evidence of a coupled solidification/solid state transformation texture in the weld metal. A closer examination of the data reveals the existence of three major texture components within the acicular ferrite microstructure, i.e. one <100> component and two complementary <111> components.v' The former component includes acicular ferrite plates which have a < 100> direction approximately
*The orientation relationship can alternatively be described by the Nishiyama-Wasserman (N-W) correspondence which also lies within the Bain orientation region. However, since the K-S and the N- W orientation relationships only differ from each other by a 5.26 rotation of the close packed planes, they can be regarded as equivalent.
430
METALLURGICAL
MODELLING
OF WELDING
(a)
(b)
Fig. 6.37. The development of transformation textures in as-deposited steel weld metals containing acicular ferrite; (a) (200) pole figure showing the crystallographic orientations of acicular ferrite referred to the original cell/dendrite growth direction, (b) Backscattered electron channeling contrast image of delta ferrite/austenite columnar grain (the metallographic section is normal to the cell/dendrite growth direction). After Kluken et al. 63
parallel with the cell/dendrite growth direction (Fig. 6.38(a)). The other plates have a <111> direction aligned in the same crystal growth direction (Fig. 6.38(b)). These data can be represented by two sub-components which are displaced with respect to each other by a 60 rotation about a common <111> axis. It is evident from the measurements of Kluken et al. 63 that the acicular ferrite plates in asdeposited steel weld metals exhibit an orientation relationship with both the austenite and the prior delta ferrite columnar grains in which they grow. This 'memory' effect arises from the characteristic solidification pattern and transformation behaviour of low-alloy steel welds (e.g. see discussion in Section 3.8.2, Chapter 3). As shown schematically in Fig. 6.39, the columnar grain region will exhibit a sharp < 100> solidification texture which has its origin in the phenomenon of preferred crystal growth. At the onset of the peritectic reaction, the austenite adopts a K-S type of orientation relationship with the delta ferrite in order to minimise the energy barrier against nucleation.P" The austenite subsequently grows around the periphery of the primary phase until impingement occurs on neighbouring columnar grain boundaries. During the ~Fe to CiFe transformation, this memory effect gives rise to the formation of acicular ferrite plates which have a < 100> direction approximately parallel with the original cell/dendrite growth direction. The presence of the two other texture components within the weld metal is thus a result of complementary crystal rotations taking place within the same orientation region. The proposed sequence of reactions is in excellent agreement with the texture analysis of Hu65 who made theoretical calculations of the resulting orientations of iron after 8Fe to ~ Fe and ~Fe to CiFe transformations in succession according to the scheme outlined in Fig. 6.39. 6.3.5.3 Nature of acicular ferrite Currently, the mechanisms of acicular ferrite formation in low-alloy steel weld metals are not
431
I\"d
o
o
T""'
A"d
(a)
(b)
Fig. 6.38. Schematic diagrams showing the three main texture components in acicular ferrite according to the Kurdjumow-Sachs orientation relationship; (a) The <100> texture component, (b) The two complementary <111> texture components. After Kluken et al. 63
fully understood. However, detailed TEM studies performed by Bhadeshia et al. 37,59,62,66 have clearly demonstrated that acicular ferrite is a form of intragranularly nucleated bainite. In practice, this means that the micro constituent may be present either as 'upper' or 'lower' acicular ferrite in the weld deposit (depending on the carbon concentration), as shown schematically in Fig. 6.40. In general, the ferrite component of upper bainite is composed of groups of thin parallel laths (subplates) with a well-defined crystallographic habit.20,61 Although the growth mechanism of upper bainite is still a subject of considerable controversy, it has been postulated that the subplates advance into the austenite with their own tip configurations. One model is shown in Fig. 6.41, where each subplate forms as a ledge upon the adjacent subplates through a nucleation and growth process. These ferrite laths possess the same variant of the K-S orientation relationship, which means that they are separated by low-angle grain boundaries. A typical austenite grain will contain numerous sheaves of bainitic ferrite exhibiting different variants of the K-S orientation relationship. This implies that the boundary between adjacent plates of acicular ferrite should alternately be of the low-angle and high-angle type, a feature which also has been observed experimentally (see data in Fig. 6.42). Hence, both the morphology and the crystallography of acicular ferrite bear a close resemblance to upper bainite.
432
liq. ----
0Fe
Iiq.+OF ~y
e
Fe
'YFe----aFe
Columnar grain Fig. 6.39. Schematic diagram showing the sequence of reactions occurring during cooling of a low-alloy steel weld through the critical transformation temperature ranges. After Kluken et al.63
6.3.5.4 Nucleation and growth of acicular ferrite Since the thickening rate of plate-like precipitates is very high, the acicular ferrite laths, once nucleated, will grow into the austenite phase until they physically impinge with neighbouring plates. At present, there exist no models which allow the acicular ferrite content to be calculated from first principles. The reason for this is quite obvious, since the final weld metal microstructure generally depends on complex interactions between a number of different variables, includingr'" (i) (ii) (iii) (iv) (v) The total alloy content. The concentration, chemical composition, and size distribution of non-metallic inclusions. The solidification microstructure. The prior austenite grain size. The weld thermal cycle.
433
Fig. 6.40. Schematic illustration of the transition from 'upper' to 'lower' acicular ferrite in low-alloy steel weld metals. The diagram is based on the ideas of Bhadeshia and Svensson.F
parallel plates
of
t
en
Q) Q) Q)
80
70
I 60
0, 50
"'0
.Q
C 40
Cti
.g ~
E 30 Q)
.~20 10
o
5 10 15 20 25 30 35 40 45 50
Plate number ~
Fig. 6.42. Measured spatial misorientation between adjacent plates of acicular ferrite in a low-alloy steel weld. Data from Kluken et al.63
434
METALLURGICAL
MODELLING
OF WELDING
A more realistic approach would be to estimate the volume fraction of acicular ferrite f~FFe via the equation.F (6-50) where fgFFe and ffitFFe are the corresponding volume fractions of grain boundary ferrite and Widmanstatten ferrite, respectively (note that in equation (6-50) the formation of microphases has been disregarded). The method outlined above has shown to work well for numerous welds (e.g. see Fig. 6.43), but fails when the primary microstructure consists of a mixture of acicular ferrite and martensite, as is the case in high strength steel weld deposits67,68 In spite of this shortcoming, equation (650) expresses in an explicit manner the real essence of the problem, namely that the evolution of the acicular ferrite microstructure depends on the interplay between several competing nucleation and growth processes which occur consecutively during cooling from the Ae3-temperature. This important point is often overlooked when discussing the conditions for acicular ferrite formation in low-alloy steel weld metals. Size effects in heterogeneous nucleation The important influence of second-phase particles on the austenite to ferrite transformation has been examined theoretically by Ricks et al.69 using classical nucleation theory. Byassuming inert, incoherent, and non-deformable inclusions and constant values for the volume free energy change and the surface free energy of both the "YFe /"YFe and "YFe /UFe boundaries, the normalised energy barrier to nucleation has been calculated and plotted against the particle radius as shown in Fig. 6.44. It can be seen from the figure that the nucleation of ferrite at inclusions is always energetically more favourable than homogeneous nucleation, but less favourable than nucleation at austenite grain boundaries, irrespectively of the inclusion size. The most potent nucleation sites are particles of a radius greater than about 0.2-0.5 urn, which
1.0
0.8
t
U
o
c 0.6
~ 0.4
:::J
0.2
0.04
0.06
Carbon content,
0.08
0.10
wt%
---.
in C-Mn steel
Fig. 6.43. Predicted variation in microstructure as a function of carbon concentration weld deposits (SMAW-lkJ mm-l). Data from Bhadeshia and Svensson."
SOLID
STATE TRANSFORMATIONS
IN WELDS
435
1
. ..c
0.75
E o 0.50
.-------Inclusion
--- Q3.
. ..c
~
0.25
8.01
0.1
10
Inclusion radius, urn
100
.
1000
Fig. 6.44. Effect of particle radius on the normalised energy barrier against ferrite nucleation at inclusions. The corresponding energy barrier against nucleation of ferrite at austenite grain boundaries is indicated by the horizontal broken line. Data from Ricks et al. 69
is within the typical size range of most weld metal inclusions (see Figs. 2.57-2.61 in Chapter 2). This finding is in excellent agreement with the results of Barbaro et al. 70reproduced in Fig. 6.45, showing that a certain minimum inclusion size (say 0.2-0.3 urn) is required for acicular ferrite nucleation in steel weld deposits. It should be noted, however, that Ricks et al. 69 omitted a consideration of the effects of plastic strain produced as a result of differences in thermal contraction between the austenite and the particles as well as the possibility for the ferrite to adopt reasonable orientation relationships with both the austenite and the catalyst particles. Based on nucleation theory it can be argued that these factors will influence the transformation process. 1 This, in tum, may explain why certain types of inclusions appear to be more favourable nucleation sites for acicular ferrite than other (to be discussed below). Catalyst effects in heterogeneous nucleation As mentioned above, there is considerable circumstantial evidence available in the scientific literature that intragranular nucleation of acicular ferrite is preferentially associated with specific types of non-metallic inclusions (i.e.1'-AI203, MnOAI203, TiN).36,55,71-78 Different mechanisms have been proposed over the years to explain these phenomena, including:36,55,73,76,77 (i) Nucleation resulting from a small lattice disregistry between the inclusions and the ferrite. Nucleation in the vicinity of inclusions caused by local compositional inhomogeneity in the steel matrix. Nucleation in the vicinity of inclusions resulting from favourable strain or dislocation arrays due to differences in the thermal contraction between the particles and the matrix.
(ii)
(iii)
436 1.0
.2
(1)
"
/
""
0.8
I I I I I I I I I I I I I I I I
/ A
/~
0
I
(U
P I
/ /
Q
I I I I I I I I I
:J 0.6 C .....
0
(3
/A
:0
~ 0.4 ca ..c
n,
>-
*
0
I I I I I I I
0.2
0 0.1
,
I
I
,
I
* 1100
~t8/5 (5) 40 4
4
0.2
0.5
Fig. 6.45. Effect of inclusion size on the probability of acicular ferrite nucleation in steel weld deposits. Data from Barbaro et al.7o
Because of the complexity of the weld metal inclusions, and the experimental difficulties involved in performing controlled in situ measurements, it cannot be stated with certainty which of these three mechanisms that are operative during the acicular ferrite transformation. However, based on simple theoretical calculations it can be argued that the contribution from the elastic strain fields around the particles due to differential contraction effects probably is too insignificant to influence the free energy of transformation and that the resulting punchingstress at the particle/matrix interface is well below the critical value required to generate new dislocations in the austenite.47,81,82 Moreover, detailed STEMIEDS microanalyses have failed to reveal detectable variations in the matrix composition in the vicinity of the inclusions.P Hence, nucleation resulting from a small lattice disregistry between the inclusions and the ferrite appears to be the most likely explanation to the observed effects of deoxidation practice on the weld metal transformation behaviour. From a theoretical standpoint, the development of a faceted ferrite nucleus which exhibits a rational orientation relationship with both the austenite and the inclusions would require that the substrate and the austenite have similar crystal structures and identical lattice orientations. The catalyst particles must therefore be cubic and bear an orientation relationship with the austenite which lies within the Bain region."? However, considering the fact that the weld metal inclusions form in the liquid state prior to the solidification process, the latter requirement cannot generally be met. 37 Nevertheless, even if the orientation of the inclusions were perfectly random, it is apparent that orientation relations within the Bain region would be observed purely by chance. In view of the high symmetry of the cubic system, the probability of this happening must be calculated. An approximative estimate is given below for single phase cubic inclusions, based on the method described by Ryder et al.84 Figure 6.46 contains a standard stereographic projection of the austenite crystal, showing the <100>-y-Fe-poles (squares) and the <110>-y-Fe-poles (dots). The Bain orientation region is represented by small circles of radius 110 centered on the <100>-y-Fe- and the <110>-y_Fe-poles.84
SOLID
STATE TRANSFORMATIONS
IN WELDS
437
Fig. 6.46. Stereographic projection of Bain regions represented by 11 circles, round <IOO>'Y_Fe-poles (squares) and <IIO>'Y_Fe-poles (dots) of the austenite lattice. After Ryder et al. 84
0
The austenite/inclusion(i) orientation relationship is within the orientation region derived from the Bain correspondence if one <100>cpole lies within a <100>'Y_Fe-region and the other two <100>cpoles lie within <110>'Y-Fe-regions. Assuming a random inclusion orientation, the probability PI that a given <100>cpole lies within a given <100>'Y-Fe-region is wI41t, where w is the solid angle enclosed by one <100>'Y-Fe-region (equal to 2n(1-cos 11) = 0.115). Since there are three <100>i directions and six <100>'Y_Fe-regions, the probability that at least one <100>i-pole lies within a <100>'Y_Fe-region is equal to:
PI =3x6x-=0.165
41t W
(6-51)
Imagine now that the inclusion lattice is rotated through 360 about the <100>caxis which lies within a <100>y_Fe-region. The other two <100>rPoles will then lie within <110>'Y_Fe-regions for at the most 4 X 22 = 88 of this rotation, since the diameter of the <100>'Y_Fe-regions is 22. Hence, the probability P2 that, if one <100>cpole lies within a <100>'Y_Fe-region, the other two will lie within <110>'Y_Fe-regions is given in the upper limit by: P ~~=0.244 (6-52)
2
360
The total probability that a given orientation relationship lies within the Bain region purely by chance is thus:
P~PI xP2 =0.165xO.244=0.04
(6-53)
Therefore, assuming random orientation, about 4% of the weld metal inclusions would lie within the Bain region purely by chance if they were single phase cubic crystals. In practice, however, inclusions commonly found in low-alloy steel weld metals are of a very heterogeneous chemical and crystalline nature. As shown in Fig. 2.72 in Chapter 2, a typical inclusion may contain up to six different constituent phases, including the three cubic phases ,),-AI203, MnOAI203, and TiN. This implies that at least 12% of the inclusions may contain a cubic phase which lies within the Bain orientation region.
438
Measurements of orientation relationships between specific inclusion constituent phases (i.e. ),-AI203, MnOAI203, and TiN) and contiguous acicular ferrite plates performed by Grong et al. 85, support the above interpretation (see Table 6.3). Referring to the standard stereographic projections in Fig. 6.47, a very high proportion of the ferrite/inclusion orientations falls within the Bain region. In addition, two other variants (i.e. No.5 and 6) are indicated for TiN which do not meet this requirement. They are therefore regarded as spurious (in the sense that the observed orientation relationships do not stem from a catalyst nucleation event) and should be ignored. Hence, it may be concluded that the observed orientation relationships between acicular ferrite and specific inclusion constituent phases are not fully reproducible in the true meaning of the word, since only those combinations which satisfy the inherent crystallography of the acicular ferrite microstructure are acceptable. An interesting observation from the data in Table 6.3 is that nucleation of acicular ferrite on inclusions is always associated with low-index planes of the {IOtl] or the {110} type, which indicates a faceted growth morphology of the inclusions. Faceted growth may occur as a result of anisotropy in the growth rates between high-index and low-index crystallographic planes, and can in the extreme case lead to a morphology of the type shown in Fig. 6.48. Consequently, formation of faceted inclusions in the liquid steel during deoxidation appears to be an intrinsic feature of low-alloy steel weld metals. Simple verification on the basis of classic nucleation theory shows that the associated reduction of the energy barrier to nucleation, ilGZet., is the primary cause for the ferrite nucleus to develop epitaxial orientation relationships with the substrate and the austenite.86,87 Referring to Fig. 6.49, a qualitative ranking of the different inclusion constituent phases with respect to nucleation potency of acicular ferrite can be made from the data presented in Table 6.3. It is evident that both ),-AI203, MnOAI203, and TiN reveal a good lattice matching with the ferrite phase in one crystallographic direction. In addition, nucleation of acicular ferrite at TiN offers
Table 6.3 Observed orientation relationships between acicular ferrite and different inclusion constituent phases in a SA low-alloy steel weld. Data from Grong et al. 85
Substrate (s) 'Y-AI203 (distorted spinel) MnOA1203 (spinel) Orientation Relationship Variant (100)s - II (OI1)a_Fe [011]s - II [533]a-Fe (200)s - II (110)a_Fe (Ol1)s II (OIO)a_Fe roiu, II [OOl]a_Fe (110) s II (lOO)a_Fe (112)s 1/ (Ol1)a_Fe [111]s II [Ol1]a-Fe TiN (NaCl) (101) s II (103)a-Fe (320)s II (l12)a_Fe [232]s II [351]a_Fe (101) s II (133) a-Fe [221]s II [200]a_Fe 6 {133} a-Feand {550}s {2001a_Fe and {221 J, 1.10 1.01 5 {3101a_Feand {330}s {1l2}a_Fe and {320}s 0.91 0.99 2 3 {11O}a-Feand {400 1s {2001a_Fe and {4401s 0.99 0.99 No. I Interplanar Spacing Plane Combinations {Oil} a-Feand {400}s Ratiot 1.02
0.97 1.17
t Defined
as d(hkl)a_Fe /dthkl),
439
010
010
100 100
100
Fig. 6.47. Standard (100) stereographic projections of the orientation relationships listed in Table (j.3; (a) Variant (1) and (3), (b) Variant (4), (5) and (6). The-Bain orientation region is indicated by the 110 circles in the graphs (see Fig. 6.46 for details).
the advantage of partial lattice coherence in a second (independent) direction, which further contributes to a reduction of ACZet. through a minimisation of the interfacial energy between the two phases. This makes TiN an extremely efficient nucleant for acicular ferrite. Microstructure data available for submerged arc (SA) steel weld deposits clearly support the above findings that nucleation of acicular ferrite occurs preferentially at inclusions which contain aluminium or titanium. As shown in Fig. 6.50, a high volume fraction of acicular ferrite is always achieved when sufficient amounts of titanium are added either through the filler wire or the flux, irrespectively of the aluminium and oxygen concentrations. This is in sharp contrast to welds produced with welding consumables containing low levels of titanium, where the acicular ferrite content drops rapidly with decreasing [~%Al]weld/[%O]anal. ratios
440
Fig. 6.48. Example of a faceted crystal delimited by {IOO} and {IIO} planes (schematic).
z
F
Nucleation site ----Fig. 6.49. Qualitative ranking of different inclusion constituent phases with respect to nucleation potency of acicular ferrite.
due to the presence of lower fractions of )'-AI203 and MnOAl203 in the inclusions (see Fig. 2.72 in Chapter 2). Similar observations have also been made by other investigators.36,7276,78,79 It should be noted that the weld metal transformation behaviour in practice depends on complex interactions between a number of important variables, including alloying and deoxidation practice, the solidification microstructure, the prior austenite grain size, and the weld thermal cycle.36,53,58This means that the presence of 'Y- A1203, MnOAl203 or TiN at the surface of the inclusions is perhaps a necessary but not a sufficient criterion for formation of acicular ferrite in steel weld metals.
Example (6.8)
Consider a partly Ti-AI deoxidised steel weld metal which contains a total number of 4 X 107 inclusions per mm". Based on Fig. 2.72 in Chapter 2 and the theory developed above, estimate an upper limit for the volume of a typical plate of acicular ferrite when the weld metal [~%AI]weld/ [%OJanal. ratio is 0.80.
SOLID
STATE TRANSFORMATIONS
IN WELDS
441
SAW
80
u.. 60
<
?f!"0
>
40
20
O~----~~~~~~~--~--~~~~~L---~--~~~
0.01 0.1 [L1%AI]
we
1.0
anal.
Id / [%0J
~ in
Fig. 6.50. Effect of deoxidation practice (inclusion chemistry) on the acicular ferrite transformation low-alloy steel weld metals. Data compiled from miscellaneous sources. Solution
From Fig. 2.72 it is seen that the total number of constituent phases in the inclusions is six, including the three cubic phases ),-AI203, MnOAI203, and TiN. If we assume a random orientation and only one nucleation event per inclusion, the following upper limit for the acicular ferrite plate volume is obtained:
VAF =
1 O.04x3x4x10
7
mm
~2x10
-7
mm
The above volume corresponds to an acicular ferrite plate which has the shape of a square lath of side 10J.lmand thickness 2J.lm. Although this estimate is in reasonable agreement with experimental observations.F the prediction is conservative in the sense that it assumes only one nucleation event per inclusion. In practice, an oxide inclusion which is orientated within the Bain region has the capability of nucleating several acicular ferrite plates, as shown by the SEM micrograph in Fig. 6.51. In addition, the acicular ferrite plates may nucleate autocatalytically at UFe /)'Fe boundaries, a process which also is referred to as sympathetic nucleation in the literature.36,37,70,83 At present, it is not clear to what extent autocatalytic nucleation plays a role in the development of the acicular ferrite microstructure. Hardenability effects Since acicular ferrite is one of the last phases to form after the growth of allotriomorphic and Widmanstattcn ferrite, it is bound to be influenced by the prior transformation products, as indicated by equation (6-50). The strong dependence of the acicular ferrite content on the austenite grain size must therefore be understood on this basis.37,47,48,70,88-9o It is evident from the data of Barbaro et al. 70 reproduced in Fig. 6.52 that a coarse austenite grain size favours intragranular nucleation of acicular ferrite at the expense of formation of allotriomorphic and
442
Fig. 6.51. SEM micrograph of a carbon extraction replica showing evidence of multiple nucleation of acicular ferrite at a weld metal inclusion.
Widmanstatten ferrite. This effect is most pronounced during slow cooling, since the combination of a small austenite grain size and a slow cooling rate implies that much of the ~ Fephase already has transformed to allotriomorphic ferrite before the temperature for intragranular nucleation of acicular ferrite is reached.
Example (6.9)
Consider a low-alloy steel weld metal which contains a total number of 5 X 107 inclusions per mm ' with an average radius of 0.25 urn. Use this information to evaluate the conditions for acicular ferrite formation within the as-deposited weld metal and the low-temperature reheated region of the weld when the austenite grain size is 100 and 10 urn, respectively. In these calculations we shall assume that the equilibrium volume fraction of ferrite fc;q~Fe in the fully transformed steel is 0.9.
Solution
First we need to estimate the total surface area per unit volume available for ferrite nucleation at austenite grain boundaries, S; (GB), and non-metallic inclusions, S; (I), respectively. As-deposited weld metal
Sv(GB)= Sv(I) = 4nNv (r)2 _2 =~mm2permm3
0.1
dr
= 20mm2permm3
= 39mm2
per mm '
SOLID
STATE TRANSFORMATIONS
IN WELDS
443
LL
'0
0 :g
c:
0.8
e--------
u. S/.:' - - -
At
- 4S
- --
0.6
jg
(])
E 0.4
::l
(5
>
0.2
100
Fig. 6.52. Effect of austenite grain size on the acicular ferrite transformation deposits. Data from Barbaro et al.l"
From the above calculations it is apparent that the conditions for intragranular nucleation of acicular ferrite at inclusions are particularly favourable within the as-deposited weld metal, since Sv(l) > Sv( GB), whereas nucleation and growth of ferrite at austenite grain boundaries will dominate within the reheated region of the weld (Sv(GB> Sv(l). This conclusion is also consistent with predictions based on the Avrami equation (equation (6-49. If we, as an illustration, assume that the volume fraction of grain boundary ferrite in the as-deposited weld metal is 0.3, the corresponding fraction of GF in the reheated weld metal becomes:
fa-Fe =ja-Fe
GF eq.
dYIDY
=0.9 1- 1- 0.9
[ ( 0.3JI0]
~O.9
On this basis it is not surprising to find that the microstructure within the grain refined region of low-alloy steel welds is usually polygonal ferrite, while the as-deposited weld metal also contains high proportions of acicular ferrite (see Fig. 6.19(c) and (d), respectively). It is important to realise, however, that the presence of allotriomorphic ferrite at the austenite grain boundaries has the benefical effect of suppressing the formation ofbainitic sheaves at the austenite grain boundaries, which, in tum, allows the acicular ferrite to develop on intragranular nucleation sites.90,91 Consequently, due to the number of competing nucleation and growth processes involved, the volume fraction of acicular ferrite is often seen to pass through a maximum when the weld metal hardenability is successively increased by additions of alloying elements, as shown in Fig. 6.53.
444
?ft. 'E
(1)
100
80
c ~
o o o :;
60
(ij
t5 2
U5 ~
40
20
0.5
1.0
1.5
2.0
After Babu
Fig. 6.53. Effect of chromium additions on the C-Mn primary weld metal microstructure. and Bhadeshia. 91
6.3.6 Acicular ferrite in wrought steels In recent years a new class of low-carbon microalloyed steels has emerged to meet the need for improved weldability, particularly with respect to the HAZ toughness. 70,92These steels are not aluminium-killed (they usually contain less than 30 ppm AI), but are instead deoxidised with titanium (120-140 ppm) to produce a relatively coarse distribution of submicroscopic Ti oxide inclusions within the base metal (presumably Ti203). During high heat input welding (> 5 kJ mm "), the oxides will not retard austenite grain growth but instead act as favourable nucleation sites for acicular ferrite within the interior of the austenite grains, as shown schematically in Fig. 6.S4(a). However, since the inclusion number density is quite low 106 particles per mm '), the microstructure is significantly coarser than that normally observed in steel weld deposites.'? Nevertheless, their presence contributes to suppress the formation of Widmanstatten ferrite and upper bainite within the grain coarsened HAZ, which is a common problem with the traditional AI-Ti microalloyed steels22 (see Fig. 6.S4(b)). 6.3.7 Bainite Bainitic microstructures (besides acicular ferrite) are frequently observed during welding, particularly in the HAZ of low-carbon microalloyed steels,22,93-95but also within the fusion region of the weld if the nucleation conditions are favourable.37,55,59,91Two main forms can be identified, i.e. upper and lower bainite, as indicated in Fig. 6.19(a). 6.3.7.1 Upper bainite In general, the morphology of upper bainite bears a close resemblance to Widmanstatten ferrite (both are typical sideplate structures), while its crystallography is more like that of lowcarbon martensite (the K-S orientation relationship is usually less precise than for WF).20 In spite of this similarity, the formation of upper bainite cannot be fully understood in terms of
445
HAZ
"'Y" HAZ Fig. 6.54. Schematic illustration of the HAZ transformation behaviour during high heat input welding; (a) Ti-oxide containing steel, (b) Conventional AI-Ti microalloyed steel. The diagrams are based on the ideas of Homma et al. 92
the classic theory of martensite nucleation assuming a pure invariant plane strain deformation.1,20,61 HAZ transformation behaviour As shown in Fig. 6.19(a), upper bainite is frequently observed within the grain coarsened HAZ of low-carbon microalloyed steels. The bainite sheaves nucleate preferentially at austenite grain boundaries and subsequently grow into the interior of the grains until they physically impinge with other transformation products such as martensite or Widmanstatten ferrite. Because of these competitive nucleation and growth processes, the highest volume fractions of upper bainite are normally attained at an intermediate cooling rate, as indicated by the microstructure data in Fig. 6.55. Weld metal transformation behaviour Upper bainite may also form within the columnar grain region of high heat input SA steel welds, as shown by the optical micrograph in Fig. 6.56. The indications are that the observed
446
~ 100
eft
Q)
"i
c: o Q.
75
E
"-
o o
(ij
t5 ::s
1i)
"-
::s
50
e o
25
20
40
Cooling time,
60
80
.
Ilt8/5'
Fig. 6.55. Effect of cooling time, I1t8/5' on the grain coarsened HAZ transformation behaviour (simulated thick plate heat cycles, Tp 1350C). Data from Akselsen et al.96
Z
Fig. 6.56. Optical micrograph showing formation of upper bainite within the columnar grain region of a SA steel weld.
shift in the weld metal transformation behaviour is related to a change in the deoxidation practice which alters the kinetics of the subsequent solid state transformation reactions through a modification of the solidification microstructure.?" Solidification induced phosphorus segregations are of particular interest in this respect, since previous examinations have shown that phosphorus can strongly enhance the formation of grain boundary ferrite by raising the local Ae3 temperature of the steel?" (see also Example 6.5). Following the discussion in Section 6.3.5.4, the evolution of allotriomorphic ferrite at the austenite grain boundaries has the beneficial effect of suppressing the formation of bainitic sheaves at these sites, which, in turn, allows the acicular ferrite to develop on intragranular
SOLID
STATE TRANSFORMATIONS
IN WELDS
447
nucleation sites. However, in fully aluminium-killed SA steel welds the presence of pure Al203 inclusions ahead of the advancing solid/liquid interface will give rise to a change in the mechanisms of the peritectic reaction (see discussion in Section 3.8.2.3, Chapter 3), which facilitates growth of the austenite grains across the primary solidification boundaries of the delta ferrite phase. Since the austenite grain boundaries are no longer situated within phosphorus-rich regions, the local steel harden ability is raised to a point where the bainitic sheaves may nucleate directly from the austenite grain boundaries and subsequently grow across the entire austenite grains. As a result, a dramatic drop in the weld metal acicular ferrite content is observed when the stoichiometric composition for precipitation of pure aluminium oxide in the weld pool is reached, as shown by the microstructure data of Kluken et ai.64 reproduced in Fig. 6.57. 6.3.7.2 Lower bainite At higher undercoolings, the transformation product will be lower bainite. In the light microscope, lower bainite closely resembles low-carbon martensite, and can therefore not readily be identified. However, transmission electron microscopy has shown that the lower bainite plates have a similar lath substructure to upper bainite, but the crystallography of the plates is different and slightly more complex.i" In addition, due to the lower temperature of transformation, precipitation of the carbides (which either can be cementite or e-iron carbides) occurs intragranularly, as shown by the TEM micrograph in Fig. 6.58. These carbides grow in the <lll>a_Fe direction and reveal an orientation relationship with the ferrite which is identical to that found in tempered martensite. However, in contrast to tempered martensite, the precipitates exhibit only one variant of this orientation relationship, such that they form parallel arrays at about 60 to the axis of the bainite plate.i" This suggests that the lower bainite reaction is controlled by precipitation of carbides at the advancing 'Y Fe /c.Y..Fe interface, which then decreases the carbon content of the austenite and enhances the driving force for further transformation.
100
?f!.
C
Q)
80 60
E o
o
~::l t5 2
U)
40
e o
0.4 O.S
1.2
1.6
ana. I
2.0
2.4
[A %AI] we Id / [%0]
Fig. 6.57. Effect of deoxidation practice on the as-deposited weld metal transformation behaviour (SAWE:::::5.2 kJ mnr '). Data from Kluken et al.64
448
METALLURGICAL
MODELLING
OF WELDING
I Martensite
Fig. 6.58. TEM micrograph showing the formation of lower bainite within the HAZ of a low-carbon micro alloyed steel.
6.3.8 Martensite At very high undercoolings, the austenite decomposes to martensite by means of an invariant plane strain deformation, which implies that there is no change in the steel chemical composition. The reaction product will either be lath or plate (twinned) martensite, depending on the alloy level. Lath martensite is commonly found in plain carbon and low-alloy steels up to about O.5wt% C, and is formed by a slip mechanism, as shown schematically in Fig. 6.59(a). When the carbon content exceeds this threshold, the martensite transformation occurs rather by formation of deformation twins (Fig. 6.59(b)). The crystal structure of plate martensite is bet (body-centred tetragonal), while lath martensite reveals a bee (body-centred cubic) strueture,61 which becomes increasingly distorted with increasing steel carbon contents." Both transformation products exhibit the characteristic Kurdjumow-Sachs orientation relationship with the austenite, but this relationship tends to be less precise at high carbon levels.1,20 6.3.8.1 Lath martensite Lath martensite is frequently observed within the grain coarsened HAZ during low heat input welding of micro alloyed steels. As shown in Fig. 6.60, the microstructure will be fully martensitic for values of ~t8/5 up to about 5-10s, depending on the steel harden ability. The hardness of the martensite formed is usually below 400 VPN, which is significantly lower than the corresponding peak value achieved after water-quenching." This means that considerable autotempering occurs during cooling from the Ms-temperature, which in the case of low-carbon micro alloyed steels can be as high as 500C. 6.3.8.2 Plate (twinned) martensite In the intercritical HAZ (i.e. partly transformed region), carbon-rich austenite, formed by decomposition of pearlite or isolated carbides, may transform to high carbon (twinned) martensite on cooling, as shown schematically in Fig. 6.61. However, since the Mf temperature in this case will be far below room temperature, significant amounts of retained austenite may be present in the areas adjacent to the martensite islands. This transformation product is therefore
449
Fig. 6.59. Mechanisms of martensite formation in steel (schematic); (a) Slip along parallel planes, (b) Generation of deformation twins. The diagram is based on the ideas of Verhoeven."
>R. 0 75
I
> 'E Q) 'E
'(i)
100
I
75
,, ,
\
Steel A}
-.-.-
Steel B Tp~1350C
,
\ \ \
\ \
\ \
\
\ \
Steel C
S
c:: Q) t::
50
\
\ \ \
ro
25
, , ,
20
" -,
<,
<,
<,
""-
10
30
Cooling time,
40
50
,s
60
70
Ilt8/5
Fig. 6.60. Effect of cooling time, ~t8/5' on the volume fraction of lath martensite within the grain coarsened HAZ of microalloyed steels (simulated thick plate heat cycles). Data from Akselsen et al.98
Kinetics of austenite formation In general, the formation of austenite during intercritical annealing of low-carbon microalloyed steels can be separated into three main stagesr'" (i) (ii) Rapid growth of austenite into pearlite until the dissolution process is completed. Slower growth of austenite into ferrite at a rate which is either controlled by carbon diffusion in the austenite or by diffusion of substitutional elements such as manganese in the ferrite, depending on the applied annealing temperature.
450
-----------AS---------
-0.8
Fig. 6.61. Schematic illustration of the formation of plate (twinned) martensite within the intercritical HAZ of low-carbon micro alloyed steels; Heating leg of thermal cycle: Pearlite ~ ~ Fe (>0.5 wt% C), Cooling leg of thermal cycle: ~Fe (> 0.5 wt% C) ~ UFe (twinned martensite),
(iii)
Very slow final equilibration of solute concentration gradients within the austenite or the ferrite through diffusion.
At temperatures above say 770C, the rate of pearlite decomposition is sufficiently high for complete dissolution, even during low-heat input welding. Hence, in hot rolled and normalised steels all pearlite will transform to austenite when the peak temperature of the thermal cycle exceeds this threshold. However, if the starting microstructure is a mixture of martensite and upper bainite (as often will be 'the case in multi-pass welding), the austenite will nucleate both at prior austenite grain boundaries and along the interfaces between laths of bainite or martensite, as shown by the optical micrographs in Fig. 6.62. A similar pattern has also been observed during intercritical annealing of dual-phase steels.104,105 Considering the kinetics, carbides precipitated within autotempered low-carbon martensite or between the laths of bainite would be expected to dissolve at a rate comparable with that of pearlite. 105 Hence, it is reasonable to assume that no retained carbides will be present intragranularly after reheating to say Tp ~ 770C when the starting microstructure is a mixture of martensite and bainite. Further growth of the austenite into the ferrite requires, however, that the peak temperature of the thermal cycle is raised significantly above 770C in order to reach the kinetic (paraequilibrium) stage where the reaction is controlled by diffusion of carbon in austenite.f" Consequently, the possibilities of obtaining growth of the austenite colonies within the low peak temperature regions of the intercritical HAZ are strongly limited, which implies that the carbon content of the austenite islands in these regions should be close to the saturation level of carbon in austenite at the temperature of dissolution. Conditions for M-Aformation On subsequent cooling, the carbon-rich austenite can decompose to a variety of microstructures (ranging from twinned martensite to pearlite) or remain untransformed, depending on the base plate hardenability, the weld cooling programme, and the initial size of the austenite islands.102,103,106 When the starting microstructure is a mixture of ferrite and pearlite (as fre- ' quently observed in normalised steels), twinned martensite forms readily along the prior base
SOLID
STATE TRANSFORMATIONS
IN WELDS
451
(a)
(b)
Fig. 6.62. Optical micrographs showing favourable sites for austenite formation during two-pass weld thermal simulation. (First cycle: Tp ~ 1350C, Dat6/4 ~ 12 s, Second cycle: Tp ~ 775C, Dat6/4 ~ 12 s); (a) Intergranular, (b) Intragranular. After Akselsen et al.103
metal pearlite bands or within isolated pearlite colonies at high cooling rates, as shown by the TEM micrographs in Fig. 6.63(a) and (b). With decreasing cooling rates, however, the transformation product shifts from twinned martensite to predominatly pearlite (see Fig. 6.63( c) and (d)) in the absence of strong hardenability elements such as molybdenium and boron which can stabilise the M-A constituent. 102,107-109 In controlled rolled and accelerated cooled steels, where the carbides are mainly present in the form of submicroscopic colonies or isolated particles located at ferrite/ferrite grain boundaries (see Fig. 6.64(a)), the situation becomes slightly more complex. Under such conditions, the majority of the austenite islands formed within the low peak temperature region of the intercritical HAZ will be of a size below 1 urn, It has been verified experimentally that such
452 (a)
(b)
(c)
(d)
Fig. 6.63. Examples of transformation products formed after intercritical thermal cycling of a ferriticl pearlitic starting microstructure to a peak temperature of 775C; (a) M-A islands (black areas) surrounded by ferrite (~t6/4 ~ 5 s), (b) Close-up of twinned martensite within a M-A island (~t6/4 ~ 5 s), (c) Isolated pearlite colony formed at an intermediate cooling rate (!:it6/4 ~ 12 s), (d) Pearlite colonies formed during slow cooling (!:it6/4 ~ 35 s). After Akselsen et al.103
small particles do not readily transform to martensite, but will largely remain in the steel in the form of retained austenite. 100,103,110,11 1 This can be explained by the lack of nucleation opportunities for martensite or by the volume restraint provided by the surrounding ferrite matrix. III Consequently, twinned martensite in controlled rolled and accelerated cooled steel is seen to form within a few, relatively large austenite islands which stern from decomposition of single carbide colonies, as illustrated by the TEM micrograph in Fig. 6.64(b). During multipass welding, the initial carbide distribution in the base plate will be of less importance because of the transformations imposed by the heat of previous weld passes. Hence, after full reaustenitising of the steel the M-A constituent may form both intergranularly and intragranularly, depending on the starting microstructure, as shown previously in Fig. 6.62(a) and (b). An example ofintergranularly nucleated twinned martensite is contained in Fig. 6.65. Volume fraction of M-A constituent At present, an adequate kinetic model for partial reaustenitising of steel during continuous heating and cooling is lacking. However, in view of the previous discussion it is obvious that the volume fraction of the M-A constituent within the intercritical HAZ of single pass steel weldments must be closely related to the initial base plate pearlite content. This is also in agreement with experimental observations (see Fig. 6.66).
453
(a)
(b)
Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate carbide distribution, (b) TEM micrograph showing evidence of twinned martensite within a 1urn large austenite colony formed during intercritical thermal cycling to Tp ~ 775C (I1t6/4 ~ 35 s). After Akselsen et al.103
454
Fig. 6.65. TEM micrograph showing evidence of intergranularly nucleated twinned martensite in double-cycled specimens (First cycle: Tp::::: 1350C, ~t6/4 :::::12s, Second cycle: Tp:::::775C, ~t6/4 :::::12s). After Akselsen et al. 103
"0
eft. >
!
c
Q)
24
Tp~ 775C
20
At
6/4
~ 5s
...;'
+-'
:l
16
C 0
I
+-'
12
-...0
c
Q)
+-' C 0
'>~
0
~-v~
.~Cj
~o~
12
16
20
24
Fig. 6.66. Effect of base plate pearlite content on the volume fraction of M-A constituent in thermally cycled specimens (simulation conditions: Tp :::::775C, ~t6/4::::: 5s). Data from Akselsen et al.106
455
.II' .II'
1400
.II'
0 0
I
cD
1200
1000
~ ...
~ 800 .a ~
c.
Q)
E 600 Q) ~
400 118 Cr - 8 Ni 1
--
for different types of carbides in 304 austenitic stainless steel as a function of carbon content. Data from Kou.I12 Cr depleted zone Grain boundary
Cr carbide precipitate Fig. 6.68. Schematic illustration of Cr23C6 precipitation at grain boundaries in austenitic stainless steels. The diagram is based on the ideas of Kou.Il2
456
6.4.1 Kinetics of chromium carbide formation Precipitation of Cr23C6in austenitic stainless steels is a typical nucleation and growth process. The temperature-time transformation curve is therefore C-shaped, and can be modelled according to the general principles outlined in Section 6.2.2.3. As an illustration, we shall assume that the parameters listed in Table 6.4 are representative of nucleation of Cr23C6in a 304 type austenitic stainless steel containing 0.07 wt% C. If we at the same time allow for the variation in austenite grain size with distance from the fusion boundary, the shape and position of the C-curve in temperature-time space at different locations within the HAZ is given by equation (6-35):
t*
~60(DY 18 )exp[18.04(
(Teq)2 2 T(Teq. - T)
O.0212J+24056(-T1 -O.OOIIJ]
(6-54)
A graphical representation of equation (6-54) in Fig. 6.69 shows that the nose of the C-curve for the reference steel (characterised by Teq. = T;q. andDy = D~) is strongly shifted to the left in the IT-diagram, thereby providing favourable conditions for Cr23C6 precipitation in the heat affected zone during welding. The problem becomes less imminent if the base plate carbon content is reduced from say 0.07 to 0.04 wt%. In that case the associated reduction in the solvus temperature Teq. from 920 to 800C will displace the C-curve towards longer times in the diagram which by far exceed the duration of a typical weld thermal cycle, as shown in Fig. 6.69. 6.4.2 Area of weld decay Experience shows that corrosion attact within the HAZ of single pass austenitic stainless steel welds usually starts in regions where the peak temperature of the thermal cycle has been close to the nose of the C-curve in Fig. 6.69. This observation cannot be explained by just considering the HAZ temperature-time pattern, since the retention time within the critical temperature range for chromium carbide precipitation is virtually the same for both high and low peak temperature thermal cycles, as indicated in Fig. 6.70. The explanation is, of course, that the C-curve becomes increasingly shifted to the right in the IT-diagram as the peak temperature increases due to austenite grain growth occurring during welding. This point is more clearly illustrated below.
Example (6.10)
Consider single pass butt welding of 2mm sheets of 304 austenitic stainless steels with covered electrodes under the following conditions: 1= 100A, U
= 25V, v = 5mm
S-1,
11 = 0.8, To = 200e
Evaluate on the basis of equations (6-32), (6-33), and (6-54) in combination with the grain growth diagram in Fig. 5.30(b) (Chapter 5) the conditions for chromium carbide formation within the HAZ during welding when the base plate carbon content is 0.07 and 0.04 wt% C, respectively. Thermal data for austenitic stainless steels are given in Table 1.1 (Chapter 1).
457
900
t
o o
800
<Ii' 700
:s
Q)
~ 600
JQ)
500
10
30
100
Time,s
300 ---....
1000
3000
10000
Fig. 6.69. Effect of carbon content on the isothermal precipitation of Cr23C6 in 304 austenitic stainless steels. Broken curve: 0.07 wt% C, Solid curve 0.04 wt% C. The diagrams are constructed on the basis of equation (6-54).
T
Fig. 6.70. Schematic illustration of thermal cycles and sensitisation unstabilised austenitic stainless steel.
Table 6.4 Input data used to construct C-curve for Cr23C6 precipitates in reference steel.
Parameter Tr (K)
t
r
(s)
* r.;
(K)
Ao (J mol ')
(kJ mol+)
Q}
-* Dy
(J.Lm)
Value
923
60
1193
150
200
18
458
Solution
If we, as a first approximation, neglect the contribution from heat flow in the welding direction, the temperature-time pattern is given by equation (1-100) in Chapter 1.Taking n = 4 and (1 - X) = 0.98 (assuming interface-controlled growth), the extent of chromium carbide precipitation occurring within the HAZ of the weld can be calculated from equations (6-32) and (6-33) by numerical integration over the actual thermal cycles. The results of such computations are presented graphically in Fig. 6.71. As expected, precipitation of chromium carbides occurs readily within the low peak temperature region of the weld when the carbon content is 0.07 wt%. For points located within the grain growth zone, the resulting displacement of the C-curve towards longer times in the CCT-diagram will gradually retard the reaction, which implies that the fraction transformed starts to decline when the peak temperature of the thermal cycle exceeds about 1150C. In contrast, the low-carbon steel reveals no sign of chromium carbide precipitation within the HAZ. This result is in good agreement with general experience, and explains why the carbon content of modern austenite stainless steels has been gradually lowered to values below 0.03 wt% in step with the progress in the steel manufacturing technology.2o,112
1.0
250 co
Cl> CG .0
"0
1
.
c
"0
1 0.8
0.6 0.4
200
(5
Cl>
Q) E
en
(j)
cD 150 .~ .~
0>
~ u,
t5
2
(ij
'0 c
Cl>
100 2
50
2 Ci5
Cl>
0.2
::>
600
800
1000
C -----
1200
Peak temperature,
Fig. 6.71. Conditions for Cr23C6 precipitation within the HAZ of a single pass austenitic stainless steel butt weld (Example 6. 10).
459
6.5.1 Quench-sensitivity
in relation to welding
High strength alloys such as AA 6082 contains manganese in addition to magnesium and silicon.' Manganese is added to control recrystallisation and grain growth in the material during hot forming. The disadvantage is that it increases the quench sensitivity of the alloy.II5 The reason for this is that the Mn-bearing dispersoids (which form during homogenisation) provide favourable nucleation sites for the non-hardening ~'(Mg2Si) phase, as shown by the TEM micrograph in Fig. 6.72. The resulting reduction in the solute content leads to a reduced HAZ strength both in the naturally aged T4 and peak aged T6 conditions.5,6,II6 6.5.1.1 Conditions for ~ "( Mg 2Si) precipitation during cooling Myhr and Grong" have shown how the quench-sensitivity concept can be applied to welding of AI-Mg-Si alloys. Their model has later been refined and extended to heat treatment of AA 6082 alloys by Shercliff et al.1l7 and Bratland et al.1l8,1l9 As an illustration, we shall assume that the parameters listed in Table 6.5 are representative of nucleation of ~'(Mg2Si) particles in an AA 6082 alloy homogenised at 580C. If we further assume that the ~'(Mg2 Si) particles only form at Mn-bearing dispersoids, the (N; / Nv) term in equation (6-14) may be replaced by the corresponding ( S; / Sv) ratio, where S; and S; refer to the total surface area per unit volume of dispersoids in the reference and the actual aluminium alloy, respectively. This leads to the following expression for t":
Fig. 6.72. TEM micrograph showing nucleation of non-hardening dispersoids in an AA 6082 alloy.
Table 6.5 Input data used to construct C-curve for (3' (Mg2Si) precipitates in reference aluminium alloy.
Parameter Tr (K)
t*
r
(s)
* Teq. (K)
766
(J mol:")
A0
(kJ mol+)
Q/
S:
Value
573
20
350
130
100
460
A graphical representation of equation (6-55) in Fig. 6.73 shows that the nose of the Ccurve for the reference alloy (characterised by T eq. = T* and Sv = S*) is strongly shifted to eq. v the left in the IT-diagram, thereby providing favourable conditions for (3'(Mg2Si) formation during welding. In general, an increase in Teq. or Sv will enhance the quench-sensitivity of the material because of the resulting increase in the nucleation rate. This will be the case if the alloy contains large amounts of excess silicon in solid solution or is homogenised at a temperature lower than 580C.118J19
Example (6.11)
Evaluate on the basis of equations (6-32), (6-33), and (6-55) the conditions for J3'(Mg2Si) precipitation within the high peak temperature region of the HAZ during welding (Tp > Tel/J In these calculations we shall assume that the transformation behaviour of the base metal is similar to that of the reference alloy in Table 6.5. Relevant thermal data for AI-Mg-Si alloys are given in Table 1.1 (Chapter 1).
Solution
Referring to Fig. 1.43 in Chapter 1, the mode of heat flow becomes essentially one-dimensional if the net arc power is kept sufficiently high compared with the plate thickness (i.e.
450
t
~ ~
f-
400
~ 350 :::J
~ 300 E Q)
250
10
30
100 Time,s
300 ~
1000
3000
Fig. 6.73. C-curve for precipitation of J3'(Mg2Si) in the reference AA 6082 alloy. The diagram is constructed on the basis of equation (6-55).
461
qold>0.05 kW mm "), Since this requirement is met in the present case, the HAZ temperaturetime pattern is given by equation (1-100). Taking n = 0.75 and (1 - XC> = 0.84 for precipitation of ~'(Mg2Si) in AA6082 aluminium alloys.!" we obtain after integration of equations (6-32) and (6-33) over the weld cooling cycle:
I. =
and
j [J~;
O.75
=0.94
The above calculations show that precipitation of ~' (Mg2Si) particles at dispersoids is, indeed, a significant process under the prevailing circumstances. Since the mode of heat flow during single pass butt welding of aluminium plates is essentially one-dimensional, it is possible to construct general transformation diagrams which give the fraction transformed as a function of the applied heat input. An example of such a diagram is contained in Fig. 6.74. 6.5.1.2 Strength recovery during natural ageing Myhr and Grong" have shown how equation (6-15) can be applied for modelling of the HAZ room temperature ageing characteristics. Figure 6.75 shows a typical natural ageing curve for AA 608~ alloys after full solution heat treatment. Due to enhanced precipitation of GP-zones, the hardness will increase from about 42 VPN to a maximum of 80 VPN after a period of 5 to 7 days. However, this ultimate hardness is significantly reduced if solute is lost during the weld cooling cycle because of precipitation of non-hardening ~'(Mg2Si) particles at dispersoids. 6,117 The relationship between the solute content, Co, and the maximum fraction of hardening precipitates which form during natural ageing, ~P' can be obtained from a simple 0.5
-0
1
en c:
0
0.4
0.3
Q)
.E
c ro !:;
~ ~ u..
0.2 0.1
logt
00
0.05
0.10
0.15
o
0.20
2
0.25
-------
0.30
Net heat input (q /vd), kJ/mm Fig. 6.74. Conditions for ~'(Mg2Si) precipitation
o> Teq).
462
1
c
ctS
Water-qlJenched specimens
N80
"E
:c -42
5-7 days
Fig. 6.75. Typical ageing curve for an AA 6082 aluminium alloy at room temperature (schematic).
2-D kinetic (cell) model, assuming that the reaction is interface-controlled. Let r denote the radius of the growing precipitates (defined in Fig. 6.76(a. Since we are only interested in the terminal value of r at a fixed temperature, the time t in the expression for r can be regarded as constant. Hence, we may write (when Co Ca and C~ Co): (6-56) where C4 is a kinetic constant. If the distribution of the precipitates is approximated by that of a 2-D face-centered cubic space lattice (see Fig. 6.76(b, the parameter, flXp, is simply given as:
(6-57)
where rm denotes the maximum particle radius which forms within the system if all alloying elements are present in solid solution at the onset of the ageing reaction (Co = C;). Because of the stoichiometry of the precipitation reaction, Co and in the expression for 11Xp may be taken proportional to the magnesium concentration in solid solution. Hence, we may write:
C;
LlX
Mg =[CC* ]2 Mg
(6-58)
Equation (6-58) predicts that 11Xp is a simple power function of CMg The magnesium concentration in solid solution can be obtained from electrical resistivity measurements, and is given by the following equationr-!"
CMg = C* [ Pm Mg
Pss - Po
Po]
(6-59)
where Pm is the measured resistivity, Pss is the resistivity in the as-quenched condition, while Po
463
is the corresponding resistivity in the fully annealed condition (i.e. when all Mg and Si are tied up in precipitates). A comparison between equation (6-58) and the electrical resistivity data in Fig. 6.77 confirms the relevance of this power-law-relationship, although the deviation in certain cases in admittedly large.
Example (6.12)
Consider plasma-MIG butt welding of a 10 mm thick AI-Mg-Si plate under conditions similar to those employed in Example 6.11. Estimate on the basis of equation (6-58) the relative (a) (b)
c o
'E Q)
o c
r------co
cex. Distance
Fig. 6.76. Simplified 2-D kinetic (cell) model for precipitation of hardening particles in AI-Mg-Si during natural ageing; (a) Particle/matrix concentration profile, (b) Cell geometry.
alloys
I
<J
1.0~----------------------~----------------------~
0.8
11~ (max)
== 1
a. 0.6
0.2 0.2
0.4
0.6
0.8
1.0
-----
464
fraction of hardening precipitates which forms within the fully reverted region of the HAZ after prolonged room temperature ageing.
Solution
Referring to Example (6.11), the relative fraction of J3'(Mg2Si) precipitates which forms at dispersoids during the weld cooling cycle amounts to:
C~
z
X(~')
= f"
A'
/ fPeq.
A'
-CM g C*
Mg
. 1 - (1 - X c )11
0.17
from which
This gives:
Since the resulting precipitation strength increment is directly proportional to dXp' 117 loss of solute in the form of 13' (Mg2Si) particles during the weld cooling cycle will inevitably lead to a reduced HAZ strength in the naturally aged (T4) condition." We shall return to this question in Chapter 7 (Section 7.4.3). 6.5.2 Subgrain evolution during continuous drive friction welding Continuous drive friction welding is a solid-state joining process that produces coalescence by the heat developed between two surfaces by mechanically induced rubbing motion. When the appropriate rotation velocity is reached, the two workpieces are brought together and an axial force is applied. The two surfaces are held under pressure, and due to the heat developed, a plasticised layer forms at the interface. After a predetermined time the rotation stops and the pressure is increased to facilitate forging or local upsetting of the heated metal. Filler metal, flux, or shielding gas is not required with this process. The structural changes within the fully plasticised region of friction welded AI-Mg-Si alloys arise from the combined action of work hardening and softening due to dynamic recovery. When the steady state conditions for deformation in friction welding are reached, the rate of dislocation generation in the plasticised material balances the dislocation annihilation rate, which means that large strains can be imposed without any changes in the flow stress or sub grain size. Referring to Fig. 6.78, the subgrain structure within the fully plasticised region consists of sheaves of virtually equiaxed grains (2-3 J.Lm in size) that are separated from each other by low-angle grain boundaries. This type of microstructure is characteristic of hot worked aluminium alloys.P" In general, the Zener-Hollomon (Zh) parameter provides a basis for evaluating the evolution of the subgrain structure during hot working. As shown by McQueen and Jonas.P' the following relationship exists between the sub grain diameter ds and the Zh-parameter in aluminium alloys:
ds
= [-0.60 +O.081ogZh
]-1
(6-60)
~
(a)
..
465
(b)
Fig. 6.78. Micrographs showing the subgrain structure within the fully plasticised and partly deformed region of a friction welded AI-Mg-Si alloy; (a) Overview, (b) Close-up of the subgrain structure (EBSP image).
where
(6-61)
The peak temperature Tp and the strain rate E distributions within the fully plasticised region of a friction 'weld can be computed on the basis of the generic models developed by Midling and Grong.P? Examples of such calculations are shown in Fig. 6.79(a). Plots of the resulting Zener-Hollomon parameter and sub grain diameter at different locations within the HAZ are contained in Fig. 6.79(b).
466
10
580
I
~CIl
+-J
' 1<f
10
1
c 'j
cD ca ~
'-
560
0 0
I
cD
'::l
Q)
540 ~
00
10
E Q)
10.1
520 ~ttS o,
Q)
<+oJ
16 16
500
0.2
0.4
0.6
0.8
Axial distance, mm
.,
480 1.0
1 1012-'------_~_-_-_~___ -_-_~_-_-_~------------~--10
Radial position: Smm -----4 mm - _.- _.1mm
14
5.0
E
t.-"'
:1
s --- ..----,.._
: .. ---:.
3.0 Ci) E ca 20 .~ . 0 .c
C/)
CD
:.0 c
:::J
1.0
Axial distance, mm
Fig. 6.79. Modelling of the subgrain evolution in the fully plasticised region of a friction welded AA 6082 aluminium alloy; (a) Predicted peak temperature Tp and strain ra.!.e(e) distributions, (b) Plots of the Zener-Hollomon Zh parameter and resulting subgrain diameter ds at different locations within the HAZ. Data from Midling and Grong.F?
467
It is evident from the graphs that the value of the Zener-Hollomon parameter is of the order of 1014_1012 S-1 within the fully plasticised region of the HAZ during continuous drive friction welding of AI-Mg-Si alloys. The predicted range in Zh corresponds to a subgrain size of 2-3 urn, in agreement with experimental observations (see EBSP image in Fig. 6.78(b)). Outside the fully plasticised region the Zener-Hollomon parameter drops from about 1012 to 1010 s' due to a sudden change the strain rate from 102 to 10 S-1 as the contribution from the material flow field in the rotational direction ceases. This value is outside the validity range of equation (6-60), and leads to the formation of coarse subgrains at the boundary between the fully plasticised and the partly deformed region, as shown by the TEM micrographs in Fig. 6.80.
Fig. 6.80. TEM micrographs showing evidence of coarse subgrains at the boundary between the fully plasticised and partly deformed region of a friction welded AA 6082 aluminium alloy.
References
1. 2. 3. 4. D.A. Porter and K.E. Easterling: Phase Transformations in Metals andAlloys, 1981, Wokingham (England), Van Nostrand Reinhold. M.E. Fine: Phase Transformations in Condensed Systems, 1964, New York, The Macmillian Company. N. Ryum: Compendium in Physical Metallurgy, 1973, Trondheim (Norway), The University of Trondheim, The Norwegian Institute of Technology. G.F. Vander Voort (ed.): Atlas of Time-Temperature Diagramsfor Nonferrous Alloys, 1991, Materials Park (Ohio), ASM International.
468
5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.
J.T. Staley: Mater. Sci. Technol., 1987,3,923-935. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1991,39,2693-2702; ibid., 2703-2708. R.D. Doherty: In Physical metallurgy, 3rd Edn (Eds R.W. Cahn and P. Haasen), 1983, Amsterdam, 934-1 030, North-Holland Physics Publishing. H.B. Aaron, D. Fainstein and G.R. Kotler: f. Appl. Phys., 1970,41,4404-4410. R. Trivedi: Me ta ll. Trans., 1970, 1, 921-927. J .W. Christian: The Theory of Phase Transformations in Metals and Alloys -Part I, 1975, Oxford (England), Pergamon Press. Atlas of isothermal transformation and cooling transformation diagrams, 1977, Materials Park (Ohio), ASM International. E. Scheil: Arch. Eisenhiittenwes., 1934/35, 8, 565-571. J.S. Kirkaldy: Metall. Trans., 1973,4,2327-2333. M.B. Kuban, R. Jayaraman, E.B. Hawbolt and J.K. Brimacombe: Metall. Trans., 1986, 17A, 1493-1503. LA. Wierszykkowski: Me tall. Trans., 1991, 22A, 993-999. R.G. Kamat, E.B. Hawbolt, L.C. Brown and J.K. Brimacombe: Metall. Trans., 1992, 23A, 24692480. M.Avrami: f. Chem. Phys., 1939,7,1103-1112; ibid., 1940,8,212-224; ibid., 1941,9,177-184. J.W. Chan: Acta Metall., 1956, 4, 572-575. M. Umemoto, Z.R. Guo and L Tamura: Mat. Sci. Technol., 1987,3,249-255. R.W.K. Honeycombe: Steels - Microstructure and Properties, 1980, London, Edward Arnold (Publishers) Ltd. H. Suzuki: Weld. World, 1982,20, 121-148. 0. Grong and O.M. Akselsen: Metal Construction, 1986,18,557-562. R.C. Cochrane: IIW Doc. IXJ-102-85 (1985). D.J. Abson and R.E. Dolby: IIW Doc. IXJ-29-80 (1980). C.A. Dube, H.I. Aaronson and R.F. Mehl: Rev. Met., 1958,55,201-210. H.K.D.H. Bhadeshia: Progress in Mater. Sci., 1985,29,321-386. E.C. Bain: Trans. Am. Inst. Min. Eng., 1924,70,25-46. S.S. Babu and H.K.D.H. Bhadeshia: Mater. Sci. Eng., 1991, A142, 209-219. H.K.D.H. Bhadeshia and D.V. Edmonds: Acta Metall., 1980,28, 1265-1273. W.C. Leslie: The Physical Metallurgy of Steels, 1981, London (England), McGraw-Hill. C.E. Cross, 0. Grong, S. Liu and J.F. Capes: In Applied Metallography, (Ed. G.F. Vander Voort), 1986, New York, Van Nostrand Reinhold, 197-210. J.E. Morral and T.B. Cameron: Proc. Int. Symp. on Boron in Steels, Milwaukee (Wisconsin), September, 1979, 19-32, Publ. The Metall. Soc. AIME (1980). S. Pakrasi, E. Just, J. Betzold and F. Hollrigl-Rosta: ibid., pp. 147-164. L. F. Porter: ibid., pp. 199-211. J.H. Devletian: Weld. i.. 1976,55, 5s-12s. 0. Grong and D.K. Matlock: Int. Met. Rev., 1986,31,27-48. H.K.D.H. Bhadeshia and L.E. Svensson: In Mathematical Modelling of Weld Phenomena, (Eds H. Cerjack and K.E. Easterling), 1993, London, The Institute of Materials, 109-180. E.D. Hondros and M.P. Seah: In Physical metallurgy, (Eds R.W. Cahn and P. Haasen), 1983, Amsterdam, North Holland Physics Publishing B.V., 855-931. T.M. Williams, A.M. Stoneham and D.R. Harris: Met. Sci. i; 1976,10,14-19. R.W. Fountain and J. Chipman: Trans. AIME, 1962, 224, 599-606. Ph. Maitrepierre, D. Thivellier and R. Tricot: Metall. Trans., 1975, 5A, 287-301. O.M. Akselsen, 0. Grong and P.E. Kvaale: Metall. Trans., 1986, 17A, 1529-1536. C.J. McMahon: Metall. Trans., 1980, l1A, 531-535. F.A. Jackobs and G. Krauss: f. Heat Treating, 1981,2, 139-146.
469
45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59.
60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.
74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86.
G.M. Evans: IIW Doc. II-A-469-79 (1979). H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: Acta Metall., 1985,33, 1271-1283. S. Liu and D.L. Olson: Weld. t.. 1986,65, 139s-150s. N.A. Fleck, 0. Grong, G.R. Edwards and D.K. Matlock: Weld. i., 1986,65, 113s-121s. H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: i. Mat. Sci., 1986,21,3947-3951. 0. Grong, T.A. Siewert and G.R. Edwards: Weld. i.. 1986, 65, 279s-288s. H.K.D.H. Bhadeshia: Mater. Sci. Forum, 1990, 56-58, 263-274. A. Ali and H.K.D.H. Bhadeshia: Mater. Sci. Techno!., 1990,6,781-784. D.J. Abson and R.J. Pargeter: Int. Met. Rev., 1986,31, 141-194. J.R. Yang and H.K.D.H. Bhadeshia: Weld. i., 1990,69,305-307. P.L. Harrison and R.A. Ferrar: Int. Mater. Rev., 1989,34, 35-51. R.C. Cochrane: Weld. in the World, 1983,21, 16-29. J.M. Dowling, J.M. Corbett and H.W. Kerr: Me tall. Trans., 1986, 17A, 1611-1623. D.J. Abson: Weld. in the World, 1989,27, 11-28. M. Strangwood and H.K.D.H. Bhadeshia: Proc. Int. Conf. on Advances in Welding Science and Technology, Gatlinburg, TN, May 1986, pp. 209-213. Publ. ASM INTERNATIONAL, Metals Park, OH, 1986. R.F. Hehemann, K.R. Kinsman and H.1. Aaronson: Metall. Trans., 1972,3, 1077-1094. J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons. H.K.D.H. Bhadeshia and J.W. Christian: Me tall. Trans., 1990, 21A, 767-797. A.D. Kluken, 0. Grong and J. Hjelen: Me tall. Trans., 1991, 22A, 657-663. A.D. Kluken, 0~Grong and G. Rervik: Me ta ll. Trans., 1990, 21A, 2047-2058. H. Hu: Trans. AIME, 1965,233, 1071-1075. A.A.B. Sugden and H.K.D.H. Bhadeshia: Metall. Trans., 1989, 20A, 1811-1818. P. Deb, K.D. Challenger andA.E. Therrien: Metall. Trans., 1987, 18A, 987-999. H.K.D.H. Bhadeshia and L.E. Svensson: J. Mater. Sci., 1989,24,3180-3188. R.A. Ricks, G.S. Barritte and P.R. Howell: Proc. Int. Conf. on Solid-Solid Phase Transformations, Pittsburg (Pennsylvania), August 1981, pp. 463-468. Pub!. The Metall. Soc. AIME (1982). F.J. Barbaro, P. Krauklis and K.E. Easterling: Mater. Sci. Techno!., 1989,5, 1057-1068. S. Liu: In Ferrous Alloy Weldments, (Eds D.L. Olson and T.H. North), 1992, Zurich (Switzerland), Trans. Tech. Publications, pp. 1-18. 0. Grong and A.D. Kluken: ibid., pp. 21-46. L. Devillers, D. Kaplan, B. Marandet, A. Ribes and P.V. Ribound: Proc. Int. Conf. on Effects of Residual, Impurity and Alloying Elements on Weldability and Weld Properties, London, November 1983, Paper 1, Publ. The Welding Institute (England). C. Bonnet and F.P. Charpentier: ibid., Paper 8. M.E. Saggese, D.N. Hawkins and J.A. Whiteman: ibid., Paper 15. A.R. Mills, G. Thewlis and J.A. Whiteman: Mater. Sci. Techno!., 1987,3, 1051-1062. R.A. Farrar and P.L. Harrison: Mat. Sci., 1987,22,3812-3820. G.M. Evans: Weld. t., 1992, 71, 447s~454s. M. Es-Souni, P.A. Beaven and G.M. Evans: Mater. Sci. Eng., 1990, A130, 173-184. S. Liu and D.L. Olson: r. Mater. Eng., 1987,9,237-251. S. Liu: PhD thesis, 1984, Dept. of Metall. Eng., Colorado School of Mines, Golden, Colorado (USA). 0. Grong and A.D. Kluken: Unpublished work. R.A. Ricks, P.R. Howell and G.S. Barritte: Mat. Sci, 1982,17,732-740. P.L. Ryder, W. Pitsch and R.F. Mehl; Acta Metall., 1967,15,1431-1440. 0. Grong, A.O. Kluken, H.K. Nylund, A.L. Dons and 1. Hjelen: Metall. Mater. Trans., 1995, 26A, 525-534. W.C. Johnson, C. L. White, P.E. Marth, P.K. Ruf, S.M. Tuominen, K.D. Wade, K.C. Russell and
s.
r.
470
87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102.
103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117.
118.
119. 120.
P.E. Marth, H.1. Aaronson, G.W. Lorimer, T.L. Bartel and K.C. Russell: Metall. Trans., 1976, 7A, 1519-1528. C.M. Dallam and D.L. Olson: Weld. J., 1989, 68, 198s-205s. G. Thewlis: IIW Doc. IXJ-165-90 (1990). R.A. Farrar, Z. Zhang, S.R. Bannister and G.S. Barrite: J. Mater. Sci., 1993, 28, 1385-1390. S.S. Babu and H.K.D.H. Bhadeshia: Mater. Sci. Techno!., 1990, 6, 1005-1020. H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 30 I s-309s. R.E.Dolby: Weld. J., 1979,58, 225s-238s. B.C. Kim, S. Lee, N.J. Kim and D.Y. Lee: Meta II. Trans., 1991, 22A, 139-149. R.H. Phillip: Weld. J., 1983, 62, 12s-18s. O.M. Akselsen, J.K. Solberg, G. Rervik andA.J. Paauw: Technical ReportSTF34 F87013, 1987, Trondheim (Norway), Sintef-Division of Metallurgy. A.O. Kluken and 0. Grong: Proc. Int. Can! on Recent Trends in Welding Science and Technology, Gatlinburg, TN, May, 1989, pp. 781-786. Publ. ASM International, Materials Park, OH, 1990. O.M. Akselsen, G. Rorvik, M.1. Onseien and 0. Grong: Weld. J., 1989, 68, 356s-362s. G.R. Speich, V.A. Demarest and R.L. Miller: Me ta II. Trans., 1981, 12A, 1419-1428. C.A.N. Lanzillotto and F.B. Pickering: Metal Sci., 1982,16,371-382. J.H. Chen, Y. Kikuta, T. Araki, M. Yoneda and Y. Matsuda: Acta Meta II., 1984,32,1779-1788. M. Ramberg, O.M. Akselsen and 0. Grong: Proc. Int. Con! on Advances in Welding Science and Technology, Gatlinburg, TN, May, 1986, pp. 679-684. Publ. ASM International, Materials Park, OH,1986. O.M. Akselsen, 0. Grong and J.K. Solberg: Mater. Sci. Technol., 1987,3,649-655. X, -L. Cai, A.J. Garratt-Reed and W.S. Owen: Me ta II. Trans., 1985, 16A, 543-557. J.J. YI, I.S. Kim and H.S. Choi: Me ta II. Trans., 1985, 16A, 1237-1245. O.M. Akselsen, J.K. Solberg and 0. Grong: Scand. J. Metall., 1988,17, 194-200. A.R. Marder: M eta II. Trans., 1981, 12A, 1569-1579. X.P. Shen and R. Priestner: Meta II. Trans., 1990, 21A, 2547-2553. N.J. Kim and Y.G. Kim: Mater. Sci. Eng., 1990, A129, 35-44. K.R. Kinsman, G. Das and R.F. Hehemann: Acta Me ta II., 1977,25,359-365. N.K. Balligerand T. Gladman: Met. Sci., 1981,15,95-108. S. Kou: Welding Metallurgy, 1987, Toronto (Canada), John Wiley & Sons. J.E. Hatch (ed.): Aluminium - Properties and Physical Metallurgy, 1984, Metals Park (Ohio), American Society for Metals. F.M. Mazzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publish. Inc. O. Lohne andA.L. Dons: Scand. J. Metall., 1983,12,34-36. R.P. Martukanitz: Proc. Int. Con! on Advances in Welding Science and Technol., Gatlinburg (Tenn.), May, 1986, pp. 193-201. Publ. ASM International, Materials Park (OH). H.R. Shercliff, 0. Grong, O.R. Myhr and M.F. Ashby: Proc. 3 rd. Int. Conf. on Aluminium Alloys -Their Physical and Mechanical Properties, Trondheim (Norway), June, 1992, Vol. III, pp. 357369. Publ. The Norwegian Institute of Technology, Trondheim, Norway. D.H. Bratland, 0. Grong, O.R. Myhr and H.R. Shercliff: Proc. Int. Conf. on Computer-Assisted Materials Design and Process Simulation, Tokyo (Japan), September, 1993, pp. 135-141. Publ. The Iron and Steel Institute of Japan. D.H. Bratland, 0. Grong, H.R. Sherc1iff, O.R. Myhr and S. Tjotta: Acta Metall. Mater., Overview ~o.124,1997,45,1-22. C.M. Sellars: Proc. 3rd Int. Conf on Aluminium Alloys-Their Physical and Mechanical Properties, Trondheim (Norway), June, Vol. III, pp. 89-106. Publ. The Norwegian Institute of Technology, Trondheim, Norway. H.I McQueen and 11Jonas: Treatise on Mater. Sci. Techno I., 1975,6,393-493. O.T. Midling and 0. Grong: Acta Metall. Mater., 1994,42,1595-1609: ibid., 1611-1622.
121. 122.
471
ACI
start temperature of ferrite to austenite transformation (OC or K) end temperature of ferrite to austenite transformation (OC or K) equilibrium ",lex solvus temperature (OC or K)
Ca.
concentration of solute at particlelmatrix interface (at. % or wt%) concentration of solute inside the precipitates (at. % or wt%) concentration of solute in matrix (at. % or wt% ) maximum concentration of solute in matrix (at. % or wt% )
AC3
C~
Co Ae3
C*
Aa./~
interfacial area per unit volume (um? per Jlm3 or m2 per m ') material constant related to the potency of the heterogeneous nucleation sites (J mol ") difference between total and acid soluble aluminium in weld metal (wt%) acicular ferrite integration parameter bulk concentration of free boron in steel (ppm) amount of boron which segregates to austenite grain boundaries on cooling (ppm) amount of boron which segregates to austenite grain boundaries in reference steel on cooling (ppm)
Ao
CCT
C.R. d
or K
S-I)
(d%Al)weld
plate thickness (mm) sub grain diameter (urn or m) average diameter of columnar austenite grains (urn or m) ferrite grain size (urn or m) austenite grain size (urn or m)
ds dy
Da Dy
-*
AF
b [B]
[B]gb
Dy
[B];b
Dm
S-1
or
[B]v
fraction of boron atoms associated with vacancies (ppm) various kinetic constants and temperature-dependent parameters
eifc(u) Eb
error function
vacancy-boron binding energy (J mol-lor kJ mol:') vacancy formation mol-lor kJ mor") energy (J
clc4
Ef
472
EBSP EDS
electron backscattering
pattern
H(T)
I
arbitrary function of (T) amperage (A) kinetic strength of thermal cycle with respect to (3-precipitation modified II integral integral in equation (6-47)
electron dispersive system volume fraction of transformation product equilibrium volume fraction of transformation product ferrite with aligned second phase molar free energy of a-phase (J mol-lor J rrr ')
L;
FS
kinetic constant in Avrami equation (variable units) time constant in Avrami equation (s) Kurdjumow-Sachs relationship orientation
G~
molar free energy of ~-phase (J mol-lor J m') arbitrary function of X molar free energy of transformation (J mol-lor J rrr ') free energy donated to the system when nucleation occurs heterogeneously (J mol-lor J m') increase in strain energy resulting from lattice distortion (J mol-lor J rrr ') increase in surface energy due to nucleus formation (J mol-lor J
rrr ')
k*
G(X)
K-S
~G
IT
I",
austenite
LB
m
proportionality constant in boron diffusion model martensite mobility parameters related to the jump-frequency of atoms across a particle/matrix interface (variable units) end temperature of martensite transformation (OC or K) start temperature of martensite transformation (OC or K)
volume free energy change associated with transformation (J mol-lor J m-3 ) energy barrier against heterogeneous nucleation (J or J mol ") energy barrier against homogeneous nucleation (J or J mor ') GF GMAW grain boundary ferrite gas metal arc welding M-A MIG
constituent
473
time exponent in Avrami equation number of nucleation sites per unit grain boundary area (m ? or Jlm-2) Avogadro constant 1023 mor ') (6.022
X
r*
na
~r~
R
isothermal zone width referred to the fusion boundary (mm) universal gas constant (8.314 J mol-l K-1) grain boundary or particle surface area per unit volume (m2 per m3 or mm? per mm ' ) shape factor related to the wetting conditions at nucleus/ substrate interface submerged arc welding shielded metal arc "welding scanning transmission microscope time (s) electron
NA
Sv N
het.
rate of heterogeneous nucleation (m-3 s') rate of grain boundary nucleation (rrr? s ' or cm? S-I) total number of heterogeneous nucleation sites per unit volume
(rrr ')
. *
S(8)
het.
Nv
N* v
total number of heterogeneous nucleation sites per unit volume in reference material (rrr-' ) Nishiyama-Wasserman tion relationship orienta-
N-W
t*
[%O]anal.
analytical weld metal oxygen content (wt%) probability polygonal ferrite net arc power (W or kW) activation energy for diffusion (J mol-lor kJ mor ') radius of precipitates (nm, urn or m) maximum particle radius following prolonged room temperature ageing (nm, urn or m)
time taken to precipitate a certain fraction of (3 at an arbitrary temperature T (s) time taken to precipitate a certain fraction of ~ at a chosen reference temperature T, (s)
* t;
PF
q() Qd
tl , t2
integration limits (s) cooling time from 600 to 4000e (s) cooling time from 800 to (s) retention time (s) temperature (Oe or K) undercooling or temperature interval (OC or K)
~t6/4
~t8/5
soooe
~tr
T
rm
~T
474
t.;
sz
Zh
To
Tp
t;
T4 T6 TEM
v
chosen reference (Oe or K) temperature
"In
frequency factor (s-l) interfacial energy between nucleus and matrix (J m-2) wetting angle (degrees) non-hardening (metastable) precipitates in AI-Mg-Si alloys hardening (metastable) precipitates in AI-Mg-Si alloys strain rate (S-I) parabolic thickening constant for plate-like precipitates (urn s-112 or m S-1I2) average value of 1 within a range specific temperature (urn S-112 or m S-112) parabolic thickening constant for spherical precipitates (urn s-1I2 or m S-112) arc efficiency solid angle enclosed stereographic Bain region by one
e
~'(Mg2Si) micro~"(Mg2Si)
electron
U Uex/f3
migration rate of incoherent interface (urn s-1) migration rate of semi -coherent interface (urn s-1) upper bainite
" 1
U~/f3
VB
v
V VPN WF X
1 welding speed (mm S-1) volume (urn", mm", or rrr') 2 Vickers pyramid number Widmanstatten ferrite 11 fraction transformed (corrected for physical impingement of adjacent volumes) fraction transformed (not corrected for physical impingement of adjacent volumes) fraction transformed defined by the C-curve maximum fraction of hardening precipitates which form during natural ageing
w
X'
Pm
measured
(nfzrn)
electrical
resistivity
Xc
Po
electrical resistivity in solutedepleted matrix (nfzm) electrical resisti vi ty in asquenched condition (nfzm)
tlXp
Pss
475
(A6-1)
= - (1 -
X
C
)(tlt*)n
In (1 -
X)
C
.!L
dt
(~)n t:
!!:..
t
=Substituting
(1 - X
C
)(tI t*)n
1n(1 _ X)
C
(~)n
t*
- (1 - X)(tlt*t
(X -1)
(A6-2)
(~)n
t*
= [ 1n(1-X) ] In (1 - X)
(A6-3)
_
t- t
*[
In (1 - X)
1n(l- X)
] lin
(A6-4)
dX
dt
In (1 - X)
1n(I-X)
]I!n
(A6-5)
Provided that equation (A6-5) contains separable variables of X and grated as follows:
n[ln (1 -
t' (T),
it can be inte-
- 1
X)]lln
fX
0
(1- X)[ln(l-
dX
X)]l-lln
= ft
0
dt
t*
= 1*
I
(A6-6)
Taking b = 1 - lin, y
= I-X
and dy
=-
dX, we get:
-1
n[ln (1 -
X)]lIn
fY
1
dy
y[ln y]b
-1
n[ln (1 -
X)]lln
1 (b -1)[ln y]b-1
476
-1
n[In (1 -
X) ]lin
lin
[In (1 - XJ]
[In (1 - X)
]lIn
*
lin
= II
(A6-7)
(A6-8)
where (A6-9)
Properties of Weldments
7.1 Introduction
Weldments are prime examples of components where the properties obtained depend upon the characteristics of the microstructure. Since failure of welds often can have dramatic consequences, a wealth of information is available in the literature on structure-property relationships. However, in order to fit some of the apparently conflicting results into a more consistent picture, a theoretical approach is adopted here rather than a review of the literature. This procedure also involves the use of phenomenological models for the quantitative description of structure-property relationships in cases where a full physical treatment is not possible.
478
dislocation density and small lath size. The formation of large proportions of upper bainite, Widmanstatten ferrite, or grain boundary ferrite, on the other hand, are considered detrimental to toughness, since these structures provide preferential crack propagation paths, especially when continuous films of carbides are present between the ferrite laths or plates. Attempts to control the weld metal acicular ferrite content have led to the introduction of welding consumables containing complex deoxidisers (Si, Mn, AI, Ti) and balanced additions of various alloying elements (Nb, V, Cu, Ni, Cr, Mo, B). The final weld metal microstructure will depend on complex interactions between several important variables such as:3-5 (i) (ii) (iii) (iv) (v) The total alloy content. The concentration, chemical composition, and size distribution of non-metallic inclusions. The solidification microstructure. The prior austenite grain size. The weld thermal cycle.
Although the microstructural changes taking place within the weld metal on cooling through the critical transformation temperature range in principle are the same as those occurring during rolling and heat treatment of steel, the conditions existing in welding are significantly different from those employed in steel production because of the characteristic strong nonisothermal behaviour of the arc welding process. For example, in steel weld deposits the volume fraction of non-metallic inclusions is considerably higher than that in normal cast steel products because of the limited time available for growth and separation of the particles. Oxygen is of particular interest in this respect, since a high number of oxide inclusions is known to influence strongly the austenite to ferrite transformation both by restricting the growth of the austenite grains as well as by providing favourable nucleation sites for various types of microstructural constituents (e.g. acicular ferrite). Moreover, during solidification of the weld metal, alloying and impurity elements tend to segregate extensively to the centre parts of the interdendritic or intercellular spaces under the conditions of rapid cooling.v ' The existence of extensive segregations further alters the kinetics of the subsequent solid state transformation reactions. Accordingly, the weld metal transformation behaviour is seen to be quite different from that of the base metal, even when the nominal chemical composition has not been significantly changed by the welding process."? This, in turn, will affect the mechanical integrity of the weldment. 7.2.1.1 Weld metal strength level In low-alloy steel weld metals there are at least four different strengthening mechanisms which may contribute to the final strength achieved. These are: (i) (ii) (iii) (iv) Solid solution strengthening. Dislocation strengthening. Grain boundary strengthening. Precipitation strengthening.
The relative contribution from each is determined by the steel chemical composition and
PROPERTIES OF WELDMENTS
479
the weld thermal history. Because of the number of variables involved, a full physical treatment of the problem is not possible. Consequently, the simplified treatment of Gladman and Pickering" has been adopted here. Figure 7.1 shows the individual strength contributions in low-carbon bainite, which is the dominating microconstituent in as-deposited steel weld metals (includes both upper and lower bainite as well as acicular ferrite). Firstly, there are the solid solution strengthening increments from alloying and impurity elements such as manganese, silicon and uncombined nitrogen, which in the present example correspond to a matrix strength of about 165 MPa. Secondly, the grain size contribution to the yield stress is shown as a very substantial component, the magnitude of which is determined by the bainite lath size. Finally, a typical increment for dispersion strengthening is indicated. This contribution is negligible at large lath sizes typical of upper bainite, but becomes significant at small grain sizes because of a finer intralath carbide dispersion." Hence, in steel weld deposits containing high proportions of acicular ferrite or lower bainite carbides will make a direct contribution to strength, even at relatively low carbon levels. The results in Fig. 7.1 are of significant practical importance, since they show the inherent limitations of the system with regard to the maximum strength that can be achieved through control of the microstructure. As shown in Section 6.3.5.4 (Chapter 6), the typical lath size (width) of acicular ferrite in low-alloy steel weld metals is about 2 urn, According to Fig. 7.1, this corresponds to a maximum yield strength of approximately 650 MPa, which is in good agreement with the observed threshold strength of acicular ferrite containing steel weld deposits.? If higher strength levels are required, it is necessary to decrease the grain (lath) size through a refinement of the microstructure, i.e. by replacing acicular ferrite with either lower
1000
o
800
N:/4 (mm-1!2)
30 40
10
20
200
10
15
20
1/2
25
-
30
Fig. 7.1. Contributions to strength in low-carbon bainite. Data from Gladman and Pickering."
480
bainite or martensite. Development along these lines has led to the introduction of a new generation of high strength steel weld metals with a yield strength in the range from 650 to 900
MPa.lO,ll
7.2.1.2 Weld metal resistance to ductile fracture It is well established that the weld metal resistance to ductile fracture is strongly influenced by the volume fraction, shape, and size distribution of non-metallic inclusions.l+P Although a verified quantitative understanding of the fracture process is still lacking, there seems to be general agreement that it involves the following three basic steps;" (i) (ii) (iii) Nucleation of internal cavities during plastic flow, preferentially at non-metallic inclusions. Growth of these cavities with continued deformation. Final coalescence of the cavities to produce complete rupture.
Details of these three stages may vary widely in different materials and with the state of stress existing during deformation. Similarly, the fractographic appearance of the final fracture surface is also influenced by the same factors. Effect of inclusion volume fraction The primary variables affecting the true strain at fracture 1 are the inclusion diameter d., and the inclusion volume fraction Vv. The relation between 1 and Vv has been determined experimentally for a wide variety of materials, and can most simply be expressed as:l?
(7-1)
where Cl is an empirical constant. The tensile test data of Widgery12 reproduced in Fig. 7.2 reveal a strong dependence of 1 on Vv, but the relationship appears to be linear rather than non-linear, as predicted by equation (7-1). Due to a similar fracture mechanism, a correlation also exists between the Charpy Vnotch (CVN) upper shelf energy and the true fracture strain in tensile testing, as shown in Fig. 7.3. For this reason, the weld metal impact properties are normally seen to decrease with increasing oxygen concentrations when testing is performed in the upper shelf region. From Fig. 7.4 we see that the CVN upper shelf energy is a linear function of the weld metal oxygen content. This observation is not surprising, considering the fact that the inclusion volume fraction is directly proportional to the oxygen level (see equation (2-75) in Chapter 2). Effect of inclusion size distribution Void nucleation may occur both by cracking of the inclusions and by interface decohesion. the former case, the critical stress for particle cracking (Jp is given by: 16
In
(7-2)
where 'Yp is the surface energy of the particle, Ep is the Young's modulus of the particle, A is the stress concentration factor at the particle, and d; is the particle diameter.
PROPERTIES
OF WELDMENTS
I
481
1.2
1.11-
I~ Ci.i
jg
~ ::J ts
C1>
1.0
<,
0.9 ~
.,
<,
I-
0.8~
., . .'
<,
<,
I
<, <,
0.7 ~
<,
<,
<,
<,
<,
0-8.003
0.005
Ej with
0.007
0.009
t
~
300 250
e>
<1>
Q)
200
150 100 50 0
~ ..c
(J)
a. a. ::J Z
Q5
> o
0.5
1.0
1.5
2.0
True fracture strain -----Fig. 7.3. Correlation of CVN upper shelf energy with true fracture strain in tensile testing. Akselsen and Grong.i" Data from
Equation (7-2) predicts that large inclusions will tend to form voids first as the stress required for initiation is proportional to (l/dv )112. This result is also in agreement with experimental observations. As shown in Fig. 7.5, the size distribution of inclusions located in the centre of voids at the fracture surface is significantly coarser than the corresponding particle size distribution in the material. In particular, large, angular shaped aluminium oxide (AI203)
482
140
1
J
:>,
C)
120
Q5 c
CD
'+CD ..c
'Q)
::l Z
en
100
c..
0..
> o
80
60
0.02
0.04
0.06
0.08
0.10
Fig. 7.4. Correlation of CVN upper shelf energy with analytical weld metal oxygen content. Data from Devillers et al. 13
inclusions appear to be preferential nucleation sites for microvoids in low-alloy steel weld metals (see Fig. 7 .5(b ). Although the combined effect of particle size and local stress concentration on the ductile fracture behaviour cannot readily be accounted for in a mathematical simulation of the process, the CVN data in Fig. 7.6 suggest that the content of large inclusions (e.g. of a diameter greater than about 1.5 urn) should be minimised in order to maintain a high resistance against dimpled rupture. In practice, this requires careful control of the weld metal aluminium-oxygen balance and the heat input applied during welding (see Section 2.12 in Chapter 2). Effect of strength level The toughness of a material reflects its ability to absorb energy in the plastic range. One way of looking at toughness is to assume that it scales with the total area UT under the stress-strain curve. Several mathematical expressions for this area have been suggested. For ductile materials we may write: 19 (7-3) where Rm is the ultimate tensile strength (UTS). If U T is regarded as a material constant, one would expect that Rm and cf are reciprocal
PROPERTIES OF WELDMENTS
483
--------~
60
d, = 0.64
20
o
(a)
0.5
1.0
----1.~
1.5
2.0
i
~ >.
0
40
20
c Q) :::J
0" Q)
d, 10
= 0.67
~ >.
0
i
o
Q)
u:
20 :::J 0"
~ u,
o
(b)
o
Inclusion diameter, Jlm
0.5
1.0
1.5
--......:JI
2.0
Fig. 7.5. Histograms showing the size distribution of non-metallic inclusions in the weld metal and in the centre of microvoids at the fracture surface, respectively; (a) Low aluminium level (Al-containing manganese silicate inclusions), (b) High aluminium level (A1203 inclusions). Data from Andersen. 18
484
150
t ..,
C> s....
.&
0
125
s: CIJ
Q) ::J Z
Q)
Q)
c:
Q5
100
a. c. 75 SAW 50
>
6
5
10
Nv (d, >1.5 urn) 10 -----Fig. 7.6. Correlation of CVN upper shelf energy with number of particles per mm ' greater than 1.5 Jim, N; (d; > 1.5 Jim). Data from Grong and Kluken.15
2.5
'e
+-I
(f)
c:
2.0
1.5 1.0 0.5
'"tel)
'0 as
~ ::s
'I-
:J
~OO
600
800
.
1000
Fig. 7.7. Correlation of true fracture strain with ultimate tensile strength (low-alloy steel weld metals). Data from Akselsen and Grong.I''
quantities, i.e. an increase in Rm is always associated with a corresponding according to the equation:
decrease in 1, (7-4)
where
C2
PROPERTIES OF WELDMENTS
485
It is evident from the tensile test data in Fig. 7.7 that the fracture strain is a true function of Rm, although the relationship appears to be linear rather than non-linear, as predicted by equation (7-4). These results are of considerable practical importance, since they imply that the upper shelf energy absorption, and hence, the shape of the CVN transition curve is strongly affected by the weld metal strength level. Accordingly, control of the weld metal microstructure becomes particularly urgent at high strength levels to avoid problems with the cleavage fracture resistance (to be discussed below). 7.2.1.3 Weld metal resistance to cleavage fracture Cleavage fracture is characterised by very little plastic deformation prior to the crack propagation, and occurs in a crystallographic fashion along planes of low indicies, i.e. of high atomic density. 1 Body-centred cubic (bee) iron cleaves typically along {laO} planes, which implies that the cracks must be deflected at high angle grain (or packet) boundaries, as shown schematically in Fig. 7.8. Consequently, in steel weld metals the ferrite grain size and the bainite packet width are the main microstructural features controlling the resistance to cleavage crack propagation. Since the microstructure which forms within each single austenite grain will not be uniform but a complex mixture of two or more constituent phases, it is difficult, in practice, to define a meaningful grain size or packet width. For this reason, most investigators have attempted to correlate toughness with the presence of specific microconstituents in the weld metal. 3-5 For example, an increase in the volume fraction of acicular ferrite will result in a corresponding increase in toughness (i.e. decrease in the CVN transition temperature), as shown in Fig. 7.9.
(a)
(b)
Fig. 7.S. Schematic diagrams showing cleavage crack deflection at interfaces; (a) High angle ferriteferrite grain boundaries, (b) High angle packet boundaries (bainitic microstructures).
486
o Pi
0
100 50 0
-50 -100
SAW (E
= 5.2~6.2 kJ/mm)
:::J
(I)
.sa
a. E c
0 .~ '(j)
. ----------------- .-----------40 0 20
~ I-
c:
~I
..
.--------.-80 .
60
100
Fig. 7.9. Correlation between the weld metal 35J CVN transition temperature and the acicular ferrite content. Data from Grong and Kluken.P
This observation is not surprising, considering the extremely fine lath size of the acicular ferrite microstructure (typically less than 5J..Lm). Moreover, results obtained from fractographic examinations of SMA and FCA steel weld metals have demonstrated that large non-metallic inclusions (> IJ..Lm) can strongly influence the cleavage fracture resistance, either by acting as cleavage cracks themselves of by providing internal sites of stress concentration which facilitate carbide-initiated cleavage in the adjacent matrix.21,22 In the former case, the critical stress required for crack propagation in the matrix, a/(M), is given by the Griffith's equation: 19 2E (
a (M) =
/
m Y eff TCC
)112
(7-5)
where Em is the Young's modulus of the matrix, ~eff. is the effective surface energy (equal to the sum of the ideal surface energy and the plastic work), and c is the half crack length. Since c is proportional to the particle diameter d., equation (7-5) predicts that welds containing large inclusions should be more prone to cleavage cracking than others. This result is also in agreement with general observations. For example, in self-shielded FCA steel weld metals it has been demonstrated that cleavage crack initiation is usually associated with large aluminium-containing inclusions which form in the molten pool before solidification (see Fig. 7.10). Consequently, control of the inclusion size distribution is essential in order to ensure an adequate low-temperature toughness. 7.2.1.4 The weld metal ductile to brittle transition In addition to the parameters mentioned above, there are several other factors, some interrelated, which play an important part in the initiation of cleavage fracture. These are: 1 (i) The temperature dependence of the yield stress.
(a)
487
(b)
(c)
Fig. 7.10. Initiation of cleavage fracture in a self-shielded FCA steel weld from an aluminium-containing inclusion; (a) Initiation site short distance ahead of the notch, (b) Detail of initiation site showing cracked inclusion, (c) Detail of cracked inclusion (remnants of particle are left in the hole).
Dislocation locking effects caused by interstitials or alloying elements in solid solution (e.g. nitrogen and silicon). Nucleation of cracks at twins. Nucleation of cracks at carbides.
488
In general, this picture is too complicated to establish a physical framework within which the various theoretical models for the ductile to brittle transition in steel can be embedded. We are therefore forced to base our judgement and understanding of how key parameters affect the position and shape of the CVN transition curve solely on scattered phenomenological observations and empirical models (e.g. see the reviews of Grong and Matlock' or Abson and Pargeter"), An example of how far the latter approach has been developed is given below. Akselsen and Grong20 have established a series of empirical equations which relates toughness to the weld metal acicular ferrite content and the ultimate tensile strength (UTS). Figures 7.11 and 7.12 show how each of these parameters influences the CVN transition curve. It is evident from the diagrams that control of the weld metal acicular ferrite content becomes particularly important at high strength levels to avoid problems with the fracture toughness. In cases where undermatch is aimed at (Le. a weld metal to base plate strength ratio less than unity), the weld metal tensile strength is typically of the order of 450 to 550 MPa. Within this range a volume fraction of acicular ferrite beyond 25 vol % will generally be sufficient to meet current toughness requirements (35 J at -40C). If overmatch is desired, the volume fraction of acicular ferrite becomes more critical, partly because of a higher weld metal strength level and partly because of more stringent toughness requirements (e.g. 45 J rather than 35 J at -40C). Process diagrams of the type shown in Figs. 7.11 and 7.12 can therefore serve as a basis for proper selection of consumables for welded steel structures. It should be noted that Akselsen and Grong-" in their analysis omitted a consideration of the important influence of free (uncombined) nitrogen and non-metallic inclusions on the CVN transition curve. Based on the experimental data in Fig. 7.13 it can be argued that such compositional variations can be equally detrimental to toughness as a decrease in the acicular ferrite content. Consequently, further refinements of the models are required if a verified quantitative understanding of the ductile to brittle transition in low-alloy steel weld metals is to be obtained.
Example (7.1)
Consider multipass FCA steel welding with two different electrode wires, one with titanium additions and one without. Table 7.1 contains a summary of weld metal chemical compositions. Provided that the microstructure and the inclusion size distribution are similar in both cases, use this information to evaluate the low-temperature toughness of the welds, as revealed by CVN testing.
Solution
Since the nitrogen content of both welds is quite high (0.011 wt%), the risk of a toughness deterioration due to strain ageing is imminent, particularly at low Ti levels. Taking the atomic weight of titanium and nitrogen equal to 47.9 and 14.0 g mol:", respectively, the stoichiometric amount of titanium that is necessary to tie-up all nitrogen as TiN can be calculated as follows:
[wt%Ti]
= 0.038
WeldA In weld A most of the nitrogen is free (uncombined) due to an unbalance in the titanium content. This means that the risk of a toughness deterioration due to strain ageing is high.
PROPERTIES OF WELDMENTS
489
(a)
t
~
:>.
c:
Q)
CD
"0
-e
<
Q)
(/J ..0
100
50
-160
O~--~~~----~--~----L----L--~~--~
-120 ..80 -40
40
80
120
160
Test temperature, C
(b)
t
Q) Q)
~ 200 E>
c
"0
150
75
-e
(J)
Q)
<
o .c
100
50
-160
160
Fig. 7.11. Predicted effect of weld metal acicular ferrite content on the CVN transition curve at two different tensile strength levels; (a) Rm = 600 MPa, (b) Rm = 800 MPa. Data from Akselsen and Grong.r"
WeldB Weld B contains 0.030 wt% Ti, which is not far from the stoichiometric amount of titanium necessary to tie-up all nitrogen. Although some titanium also is bound as Th03, it is reasonable to assume that the free nitrogen content in this case is sufficiently low to eliminate problems with strain ageing. Consequently, weld B would be expected to exhibit the highest toughness (i.e. the lowest CVN transition temperature) of the two, as indicated in Fig. 7.14.
490 (a)
......,
I
E>
c
CD
Q)
300
25 vol% acicular ferrite 250
UTS
200
150 100
/
"'0
CD .0 0 C/) .0
I-
/ .I/.':,;" .
'/',
50
.._..- .._..- :.. ~;;;'f.~~f.;..~~ .. .,.._.._.._..- .~!? .J9y.I.~l?- ..- ..- .._ .._ ..
160
Test temperature,
'''~~:.' ....
./:;'
M160 -120
(b)
300
......,
~ Q)
CD
:>,
UTS
c:
-c
.,-....".....---------- 500
.. 600
-e
Q) C/) .a
50
.._ .._ ..- ~~?f: ~.::.~:.::._ .._ .._ .._ .._ .._ ..- ..3.~ .J.o..ule.$.- .._ ..- ..- .._ ..
160
'l"
'l.::'.....
.'
~/:" .... ..
,
.I/./~,"~
-160
Fig. 7.12. Predicted effect of weld metal ultimate tensile strength (UTS) on the CVN transition curve at two different volume fractions of acicular ferrite; (a) 25 voI% AF, (b) 75 voI% AF. Data from Akselsen and Grong.i'' Table 7.1 Chemical composition of FCA steel weld metals considered in Example (7.1).
Element Weld A B wt%C 0.10 0.10 wt% Si wt%Mn 1.50 1.50 wt%Al 0.005 0.005 wt%Ti 0.006 0.030 wt%S 0.008 0.008 wt%N 0.011 0.011 wt%O 0.031 0.031
DAD DAD
PROPERTIES OF WELDMENTS
491
t
c ~
Q) (J)
150
~ e>
100
-e
o tJ) .c
50
300.---------~--------~--------~-----------
-.,200 ~ 0>
'Q)
Q)
"C
!
c:
Q)
...... 9.---0/0
/ / ~
inClusions ( > 1J.L m )
-e
~
o CI)
100
9'
."",.,. -0
/
-100
Fig. 7.13. Effect of impurities on weld metal CVN toughness; (a) Nitrogen content, (b) Inclusion level. Data from ESAB AB (Sweden) and Grong et al. 22
7.2.1.5 Effects of reheating on weld metal toughness In principle, improvement of weld properties can be achieved through a post-weld heat treatment (PWHT). This may have the benefits of:3 (i) Enhancing the fatigue strength through a general reduction of welding residual stresses.
492
>.
Cl)
/~
./'
c
Cl)
"0
e o
Q)
en .c
___ ~~~
__ I /: 1--/)
/~WeldA
Test temperature,
C ~
Fig. 7.14. Schematic drawings of the CVN transition curves for welds A and B (Example (7.1)). (a)
(b)
Fig. 7.15. Typical low-temperature fracture modes of Ti-B containing steel weld metals; (a) Quasicleavage (as-welded condition), (b) Intergranular fracture (after PWHT).
PROPERTIES OF WELDMENTS
493
(ii)
Increasing the toughness by recovery (i.e. removal of strain-aged damage) and martensite tempering.
For these reasons local PWHTs are commonly required for all structural parts above a specified plate thickness (e.g. 50 mm according to current North Sea offshore specifications). Post-weld heat treatment is usually carried out in the temperature range from 550 to 650C. In practice, however, the toughness achieved will depend on the weld metal chemical composition, and in some cases deterioration rather than improvement of the impact properties is observed after PWHT. In such cases the reduction in toughness can be ascribed to:3,4 (i) Precipitation hardening reactions. Present experience indicates that elements such as vanadium, niobium, and presumably titanium can produce a marked deterioration in toughness because of precipitation of carbonitrides in the ferrite, provided that these elements are present in the weld metal in sufficiently high concentrations. Segregation of impurity elements (e.g. P, Sn, Sb and As) to prior austenite grain boundaries, which, in tum, can give rise to intergranular fracture. The indications are that this type of embrittlement is strongly enhanced by the presence of second phase particles at the grain boundaries.
(ii)
Experience shows that Ti-B containing steel weld metals often fail by intergranular fracture in the columnar grain region after PWHT,23 as evidenced by the SEM fractographs in Fig. 7.15. The observed shift in the fracture mode is associated with a significant drop in toughness (Fig. 7.16) and arises from the combined action of solidification-induced phosphorus segregations and borocarbide precipitation along the prior columnar austenite grain bounda-
60r---------~1~--------~1----------~,--------~
SAW Open symbols: 5 - 8 ppm B Filled symbols: 20 - 25 ppm B -
> o <l
\ o ~._._.\._._._._._._._.!!~~J.i.!!~_._._._._._._._._._._._
-20 -
-40 -
-, ---- ----.~
--- ~
494
Fig. 7.17. TEM micrograph showing precipitation of borocarbides, Fe23(B,C)6, along prior austenite grain boundaries in a Ti-B containing steel weld metal after PWHT (600C-l h).
ries (Fig. 7.17). Since borocarbides are brittle and partly incoherent with the matrix, they can be regarded as microcracks (of length dp) ready to propagate. In such cases there is virtually no plastic deformation occurring before crack propagation, which implies that the intergranular fracture stress is given by the Griffith's equationr'" (7-6)
where 'Yeff. is again the effective surface energy (equal to the sum of the ideal surface energy and the plastic work), and dp is the particle diameter. Although the value of 'Yeff. would be expected to be low in the presence of solidificationinduced phosporus segregations.P' this alone is not sufficient to initiate intergranular fracture in the weld metal. However, during PWHT the borocarbides will start to grow from an initially small value to a maximum size of about 0.1 to O.2J.lm(Fig. 7.17), following the classic growth law for grain boundary precipitates dpat1l4 25 This implies that the intergranular fracture stress, all), will gradually decrease with increasing annealing times, as indicated in Fig. 7.18. When the matrix fracture strength, aiM), is reached, the fracture mode shifts from predominantly quasi-cleavage in the as-welded condition (Fig. 7.1S(a to intergranular rupture after PWHT (Fig. 7.1S(b. This is observed as a marked reduction in the CVN toughness, as shown previously in Fig. 7.16. 7.2.2 HAZ mechanical properties The last twenty years have seen a revolution in the metallurgical design of steel. Whereas old steels relied on the use of carbon for strength, the trend today is to rely more on grain refinement in combination with microalloy precipitation to meet the current demand for an improved weld ability. This includes both the sensitivity to weld cracking and the HAZ mechanical properties required by service conditions and test temperatures. The latter aspect is of particular interest in the present context and will be discussed later.
PROPERTIES OF WELDMENTS
495
..
(M)
:- - - - - - - - - - - - af
Quasi-cleavage fractu re mode
Particle diameter
[Annealing time]1/4
Fig. 7. IS. Schematic illustration of the mechanisms of temper embrittlement in Ti-B containing steel
weld metals (aj(M): matrix fracture strength, aj(l): intergranular fracture strength).
7.2.2.1 HAZ hardness and strength level The HAZ hardness and strength level is of significant practical importance, since it influences both the cracking resistance and the toughness. Although the peak strength is mainly controlled by the martensite content (see Fig. 7.19), the relationship is generally too complicated to allow reliable predictions to be made from first principles. This implies that our understanding of the HAZ strength evolution, at best, is semi -empirical. 100
15 > 'E
0 0
Q)
t ~
0
.. ....
'
600
-"
1400
500
70
E
50 2
c Q) t::
---
t
-..
E .x::
I
0.
1200
.....
'. -. ~--'., ..
400 (\IE
1000 ~
as
' '--.:.:~.
" It
'(i.i
" .ee.?..... --
:.:.:.:.: .::':.:.:.::.::':'::'::.:....
HVs
E
0.
C\I
>t[)
ci
as
25
10
20
30
40
50
S ----...
60
70
Fig. 7.19. Structure-property relationships in the grain coarsened HAZ of low-carbon microalloyed steels (vol% M: martensite content, RpO.2 : 0.2% proof stress, Rm: ultimate tensile strength, HV 5: Vickers hardness,
fl.tS/5:
496
METALLURGICAL
MODELLING
OF WELDING
A number of different empirical models exist in the literature for prediction of HAZ peak hardness and strength.26-31 However, the aptness of some of these models is surprisingly good, which justifies construction of iso-hardness and iso-strength diagrams for specific grades of steels. 32 Examples of such diagrams are given in Fig. 7.20. It is evident from Fig. 7.20 that the HAZ peak strength is controlled by two main variables, i.e. the steel chemical composition and the weld cooling programme. The compositional effect is allowed for by the use of an empirical carbon equivalent, which ranks the influence of the various alloying elements on the steel hardenability. According to Yurioka et al., 28 the CEIl-equivalent is given as:
C %Si 24 %Mn 5 %Cu 10 %Ni 18 %Cr 5 %Mo 2.5
CE II = -10
01.
+--+--+--+--+--+--+-+-5
%V
%Nb 3
(7-7)
where all compositions are given in wt%. Moreover, the cooling time from 800 to 500C, /},l8/5, is used as an abscissa in Fig. 7.20. This parameter is widely accepted as an adequate index for the weld cooling programme, and can be read from nomograms of the type shown in Fig. 1.49 (Chapter 1). The axes of Fig. 1.49 are dimensionless, but they can readily be converted into real numbers through the use of the following conversion factors.P Ordinate:
~'t8/5 ~
1.57 X 10-4
~t8/5
n3 [118500 -1/8800]
11E
(7-8)
Abscissa:
n3 2[1/8 500
+1/8800]
11E ~ 636622
[1
(500-To)
+----
(800-To)
1]
(7-9)
The different parameters in equations (7-8) and (7-9) are defined in Appendix 7.1. The results in Fig. 1.49 are interesting, since they show that the cooling time, I1t8/5, depends on the mode of heat flow during welding. In this case the transition from 'thick' to 'thin' plates, corresponding to an abscissa of about 0.64, is clearly not represented by a single plate thickness d, but will be a function of both the net heat input 11E and the ambient temperature To. Accordingly, the HAZ strength level is seen to vary between wide limits, depending on the steel chemical composition and the operational conditions applied (Fig. 7.20).
Example (7.2)
Consider stringer bead deposition (GMAW) on two low-alloy steel plates of similar composition but different thickness under the following conditions: I
= 250A,
= 30V,
= 5mm
According to the steel mill certificate the CEIl carbon equivalent is equal to 0.46 wt%. Use this information together with the diagrams in Figs. 1.49 and 7.20 to estimate the peak HAZ strength level when the plate thickness is 10 and 30 mm, respectively.
PROPERTIES OF WELDMENTS
497
(a)
0.55
~ ~
0
0.50
w0
0.45
0.40
25
' S
30
.
35
(b)
0.55
1
ur
o
0.50
0.45
10
15
20
25
' S
30
35
Fig. 7.20. HAZ iso-property diagrams for HSLA steels; (a) Iso-hardness contours, (b) Iso-yield strength contours. Data from Kluken et ai.32
498
Solution
First we calculate the net heat input per unit length of the weld 11: 250x30 1000 x 5
11 = 0.8
ns -2
8
[1/8
500
(780)
257 ::::::-2 d
= 10 mm:
-4 fit8/5
11E
1 ] ( 4~O) _ ( 7~O) - 4
= 30mm:
-4 fit 8/5
11E
1 (4~O)-( 7~O)
] -1
We can now use the diagrams in Fig. 7.20(a) and (b) to obtain the peak HAZ hardness and yield strength, respectively. This gives: d
= 10mm:
HV ~ 230VPN and R
PO.2
~ 470MPa
~ 740MPa
It is evident from the above calculations that the HAZ strength level is sensitive to variations in the welding conditions. Normally, the HAZ hardenability is high enough to facilitate
PROPERTIES OF WELDMENTS
499
a local strength increase adjacent to the fusion boundary, as shown in Fig. 7.21. An exception is high heat input welding on quenched and tempered steels (Fig. 7.21(b)), where the presence of large amounts of Widmanstatten ferrite and polygonal ferrite within the grain coarsened and grain refined region, respectively can lead to a severe HAZ softening. This type of mechanical impairment represents a problem in engineering design, since it puts a restriction on the use of high strength steels in welded structures.
(a) 1200
I I I 1 Ir, I I ,I
I I I 1
~\
I'~ 1_-I-:ro.IR
1 1 1 I
I 1 I 1
(b)
urn strength
1 1 I I 1 I I I I I E I 1 1 I 1 1 I 1 1 I lieu 0I I I ~ I 1,':""',1 1 -1' I ~ ,'1 I I II: I I I
steels
-g
a:
cu
"!
R -.rn..... t
I I
I I
Fig. 7.21. Effects of steel chemical composition and welding conditions on the HAZ strength level (BM: base metal, SR: subcritical region, IR: intercritical region, GRR: grain refined region, GCR: grain coarsened region); (a) Low heat input, (b) High heat input. Data from Akselsen and Rervik.I"
500
7.2.2.2 Tempering of the heat affected zone Certain regulations for offshore structures require that no part of the welded joint shall be harder than a specified limit, e.g. 280,300 or 325 VPN, to reduce the risk of hydrogen cracking. Such requirements cannot always be met by a suitable choice of preheating and welding conditions. In practice, a reduction in the HAZ strength level can be achieved by applying a PWHT. The tempering effect of different temperature-time combinations can be described by the Hollomon-Jaffe parameter.P (7-10) PI = T(B* + lnt) where Tis in K (absolute temperature). In Fig. 7.22 the isothermal hardness data reported by Olsen et ale 36have been plotted against the empirical Hollomon-Jaffe parameter. In this particular case the best fit is obtained if the constant B* in equation (7-10) is equal to 16.5 (with t in seconds). It is evident from Fig. 7.22 that tempering at, say, 600C for 1 h is more than sufficient to reduce the HAZ peak hardness to values below 280 VPN. This implies that PWHT is an effective (but expensive) way of reducing the HAZ strength level. Deposition of temper weld beads has been suggested as an alternative means of reducing the hardness of the HAZ.36-38 This procedure is indicated schematically in Fig. 7.23, showing two temper beads (black) in the lower sketch. If the beads are properly positioned with respect to the fusion line, the outer ACI contour of the HAZ produced by the temper bead should just touch the fusion line of the last filler pass, as indicated in the upper sketch of Fig. 7.23. The material reaustenitised by the temper bead would then be weld metal, while the HAZ remaining from the last filler pass would be tempered below the transformation range.
I
c
o 591 K
400
A 782 K v 869 K
0668 K
"E
S ..c:
~ 300 ~c
5000
10000
15000
-
~ = T(16.5
Fig. 7.22. Hollomon-Jaffe
20000
+ log t)
PROPERTIES
OF WELDMENTS
501
Since the Hollomon-J affe parameter is an empirical criterion developed for isothermal tempering of medium and high carbon steels, it cannot readily be applied to pulsed tempering. A better approach would be to use the so-called Dom parameter-'? which in an integral form can be written as:39,40 (7-11)
where Qapp. is the apparent activation energy for the controlling diffusion reaction. The Dom parameter has proved useful to compare isothermal and pulsed tempering data on the assumption that the kinetics of softening, in the actual range of hardness, are controlled by diffusion of carbon in ferrite. Qualitatively, the aptness of equation (7-11) can be illustrated in a plot of measured hardness against the diffusional parameter P2 (see Fig. 7.24). It is evident from Fig. 7.24 that the isothermal data points can be represented by a smooth curve which coincides with the upper boundary of the scatter band obtained in pulsed tempering. The slightly higher hardness observed after isothermal tempering arises probably from a brief period of heating that makes the effective time somewhat less than lOs.
Case Study (7.1)
As an illustration of principles, Fig. 7.25 shows a case of identical welding parameters for the last filler pass and the temper bead, the latter one being positioned so as to give a peak temperature of 720C at the fusion line of the former one. The temperature field around the two
502
!
x as
E
0
100
0
A
[]
400
> ::r: 3;
en
Ch
c
80
350
>
z n,
en
en (1) c
:r:
~...
CD S
300
S .c
"E
"E
:J:
60
250
~ ~0
\ r;::-t
V ~
0.01
1/2 S ----(Qapp.
0.02
Fig. 7.24. Measured hardness ratio HVIHVmax. vs the Dom parameter P2 from Olsen et al. 36
= 83.14 kJ mol:').
Data
beads is clearly the same. In Fig. 7.25 an estimate has been based on the simplified Rykalin thick plate solution, which applies to a fast moving high power source on a semi-infinite body (see equation (1-73) in Chapter 1). At T-To ::::: 1500C, a fusion line radius of about 4.4 mm is obtained for a net heat input of 0.8 kJ mrrr '. The corresponding ACI radius is 6.S mm. The temperature-time pattern is shown in the lower left graphs of Fig. 7.25 for three different positions in the HAZ, i.e. y = 0 (former fusion line), y = 1 mm, and y = 2 mm (z = 0). The corresponding plots of dP2 Idt vs t are shown to the right. Taking the area P2 under each curve and reading the hardness ratio at {p; from Fig. 7.24, an expected hardness profile is obtained, as shown in the upper diagram of Fig. 7.25. The expected effect of tempering is seen to range from a hardness of about 65% (HV:::::265 VPN) at the fusion line to about 80% (HV::::: 340 VPN) close to the outer boundary of the HAZ (y = 2 mm). If the centre-line displacement had been different from the chosen optimum of 2.1 mm (e.g. say 3 mm), the predicted hardness curve would be shifted to about 75% and 90% of the peak hardness at y = 0 and y = 2 mm, respectively. On the other hand, if the centre-line distance had been shorter, say 1 mm, a narrow zone of the original HAZ would be re-austenitised and therefore about as hard as before deposition of the temper bead. The results from the above modelling exercise show that the HAZ hardness of weld toes and cap layers can be reduced by applying an appropriate temper bead technique. However, this requires an extremely good process control, since the temper beads must be positioned very precisely for a successful result. Consequently, the use of temper beads for improvement of the HAZ properties has not found a wide application in the industry. 36,41 7.2.2.3 HAZ toughness In spite of the recent developments in steel plate manufacturing technology, there is still concern about the HAZ toughness of low-carbon micro alloyed steels because of their tendency to
PROPERTIES OF WELDMENTS
503
~ > J: 3>
J:
o
0
t-=
800 600 400 200
0
0 0 0
"';"
Q)
y=O; 1
rfVP; =9.9
3
COo
....
20
y=1 mm;10
Yri=3.8
3
10
y=2 mm;10
Yri=1.S 8
t,s
10
20
30 t.s
Fig. 7.25. Application of Dom parameter to weld bead tempering (Case Study (7.1)).
form brittle microstructures within specific thermal regions of the weld.41,42 Moreover, improvement of the HAZ toughness through PWHT is sometimes found to be difficult in contrast to experience with more traditional C-Mn steels.41,43 Consequently, the increasing use of lowcarbon microalloyed steels in various welded structures has introduced new problems related to the HAZ brittle fracture resistance which formerly did not appear to be of particular concern." Fully transformed region Specific effects of peak temperature on HAZ toughness, as assessed on the basis of thermally cycled CVN specimens, are shown in Fig. 7.26. It is apparent from the graph that embrittlement in the fully transformed part of the HAZ is often locatedin the grain coarsened region adjacent
504
300 r---rr----r-r-r----r----r'-..,.---,~-~-___,
t
J Q)
SR
250
IR
GRR
GCR
~ 200 2>
c:
~ 150 (J) .a ~
<
100 50
O~--~----L-~~----L---~~~~
600 700 800 900
__~ __~
Peak temperature,
Fig. 7.26. Effects of peak temperature on the CVN energy absorption at --40C (SR: subcritical region, IR: intercritical region, GRR: grain refined region, GCR: grain coarsened region). Data from Akselsen et al.45
to the fusion boundary where the peak temperature of the thermal cycle has been above about 1200C. The problem can mainly be ascribed to the presence of low-toughness microstructures such as upper bainite and Widmanstatten ferrite which form typically at intermediate and slow cooling rates (see Fig. 6.55 in Chapter 6). In contrast, the grain refined region will almost always exhibit a satisfactory low-temperature toughness owing to the characteristic fine polygonal ferrite microstructure.i! An exception is low heat input welds produced from steels with a heavily banded pearlite/ferrite microstructure, where the risk of a toughness deterioration is imminent due to martensite formation along the prior base metal pearlite bands.45,46 In recent years a new class of low-carbon microalloyed steels has emerged which is characterised by an excellent low temperature HAZ toughness, even at high heat inputs (see Fig. 7.27). This particular grade is frequently referred to as Ti-O steels due to their content of indigenous titanium oxide inclusions (presumably Ti203). Although the mechanisms involved are not yet fully understood, it is reasonable to assume that the improved toughness at high heat inputs arises from a refinement of the HAZ microstructure, as discussed previously in Section 6.3.6 (Chapter 6). It is interesting to note that the major effect of the titanium oxide inclusions in this case appears not to be control of the austenite grain size (which in some cases can exceed 500 urn at the fusion boundary), but is rather to act as favourable nucleation sites for acicular ferrite intragranularly.fv" Similar phenomena are well known from transformation kinetics of low-alloy steel weld deposits, where non-metallic inclusions play an important role in the development of the acicular ferrite microstructure+f Intercritical region The microstructural evolution in the intercritical HAZ of low-carbon steels has previously been discussed in Section 6.3.8.2 (Chapter 6). In order to understand the origin of embrittlement in the intercritical region, consideration must be given to the stress fields and the transformation strains developed in the ferrite matrix
PROPERTIES
OF WELDMENTS
505
1 o
0
40
~t8/5=
161 S
o Ti-O
20
/ / /
Pi
~
::s
E S
iii
0-
Q)
c -20
/
c cu
-c
-------
1450
Fig. 7.27. Response of modern Ti-O steels and traditional Ti-N steels to CVN testing following weld thermal simulation. Data from Homma et al. 47
as a result of the martensite formation.t? It follows from Fig. 7.28 that the hard martensiteaustenite (M-A) islands will give rise to significant stress concentrations at the martensite/ ferrite interface owing to the pertinent difference in the yield strength (stiffness) between the two phases. At the same time, the volume expansion associated with the austenite to martensite transformation leads to significant elastic and plastic straining of the ferrite.P" At moderately high temperatures and deformation strains, many of the matrix dislocations will be mobile, which means that the ferrite will maintain its ductility, while the stiffer M-A islands are exposed to cracking and debonding. With increasing strain, the cracks can grow into voids and further develop into deep holes, until final rupture occurs by hole/void coalescence due to internal necking.f? However, when mechanical testing is performed at subzero temperatures under high strain rate conditions (> 102 s-l for CVN testing), the flow strength of the ferrite increases significantly because of the reduced mobility of the screw dislocations.t! In addition, strain partitioning between the M-A islands and the ferrite may also occur, which further enhances the stress concentrations at the M-Alferrite interface.V Accordingly, the local stress level at the interface will eventually exceed the cleavage strength of the ferrite, with consequent initiation of brittle fracture. This conclusion is consistent with observations made from tensile testing of dual-phase steels, showing that failure of dual-phase microstructures often is caused by fracture in the ferrite region. 52-54 Because the intercritical HAZ toughness is closely related to the volume fraction of the M-A constituent in the matrix, 45,51,55 embrittlement can normally be avoided by decreasing the cooling rate through the critical transformation temperature range to facilitate pearlite formation (see Fig. 7.29). An exception is boron-containing steels, where the HAZ harden ability is high enough to stabilise the M-A constituent, even at slow cooling rates (see CVN data for steel B in Fig. 7.29).
506
METALLURGICAL
MODELLING
OF WELDING
U)
(f)
VJ Q)
"C
.~
(ij
E
(5
"-
.....
_-
- -
-'tmax
Normalized distance Fig. 7.28. Stress distribution in matrix caused by stiff inclusion (0',: radial stress, O'e: tangential stress, 49 'tmax.: maximum shear stress). Data from Chen et al.
t
J
200
Open symbols: Tp ~ 770C Filled symbols: Tp ~ 840C
150
e>
Q) Q)
c:
"0
100
-e
Q)
<
0 en .c
50
10
---50
60
Fig. 7.29. Effect of cooling time t:.t6/4 on the intercritical HAZ toughness at -20C (thermally cycled specimens). Steel A: 11 ppm B, Steel B: 26 ppm B. Data from Ramberg et al. 55
Effect of PWHT Considering the intercritical HAZ, a significant improvement of the CVN toughness can be achieved by applying a PWHT, as shown by the data of Akselsen et al.51 This effect arises partly from a reduction of the stress concentrations at the M-Alferrite interface as a result of martensite tempering and partly from relaxation of transformation strains within the ferrite matrix. 51 Such recovery reactions will start to occur when the temperature is raised above about 100C.
PROPERTIES OF WELDMENTS
507
(a)
(b)
1.4
c5
a "' ;15 0
0 l0
E E
0.4
0.2
0
0.020
Fig. 7.30. Effects of PWHT on the grain coarsened HAZ toughness; (a) Example of intergranular fracture along prior austenite grain boundaries after PWHT (600C - 1 h), (b) Measured CTOD vs base plate phosphorus content for post weld heat treated specimens (600C - 4 h). Data quoted by Grong and Akselsen."!
In contrast to the behaviour described above for the intercritical HAZ, the reported effect of PWHT on the grain coarsened HAZ toughness is much more complicated and rather confusing. However, experience has shown that particularly niobium-vanadium containing steels are sensitive to PWHT due to the strong precipitation hardening potential of Nb(C,N) and V(C,N).43.56 In addition, a toughness deterioration may occur as a result of segregation of impurity elements such as phosphorus, tin, and antimony to prior austenite grain boundaries. This, in combination with a tempered martensitic microstructure, can lead to intergranular
508
fracture when testing is performed at subzero temperatures (see Fig. 7.30(a)). The detrimental effect of phosphorus on the HAZ toughness of low carbon microalloyed steels after PWHT is shown in Fig. 7.30(b).
Example (7.3)
= 500A,
= 30V, v = 6mm
S-I,
11 = 0.95, To
= 20C
Table 7.2 contains data from CVN testing of the base plate and thermally cycled specimens. The weld thermal simulation experiments were carried out at three different peak temperatures (Le. 1350C, 1000C, and 780C) under cooling conditions similar to those employed in the SA welding trial. Based on the data in Table 7.2 and the simplified Rykalin thick plate solution (equation (1-73) in Chapter 1), estimate the locations of the brittle zones (referred to the fusion boundary) within the HAZ of the SA procedure test weld considered above.
Solution
It is evident from the CVN data in Table 7.2 that the HAZ toughness would be expected to be low in positions of the weld where the peak temperature has been close to 780 and 1350C, conforming to the intercritical and grain coarsened region, respectively. Based on the simplified Rykalin thick plate solution, the following expression can be derived for an arbitrary isothermal zone width, referred to the fusion boundary (see equation (5-47) in Chapter 5):
~r:,
m -
~r * _
1 -To
1 ~Tm -To
Taking pc and Tm equal to 0.005 J mrrr ' C-l and 1520C, respectively for low-alloy steels (from Table 1.1 in Chapter 1), we obtain:
Table 7.2 Results from CVN testing of base metal, thermally cycled specimens, and procedure test weld (Example (7.3
Test results Base metal Thermally cycled specimens WeldHAZt Absorbed energy at -40C (J) 320,310, Tp = 780C Tp = 1000C Tp = 1350C 305; average: 312 40,36, 34; average: 37 225, 220,219; average: 221 50,46,40; average: 46
tOCR: grain coarsened region; ORR: grain refined region; IR: intercritical region.
PROPERTIES OF WELDMENTS
509
I1r*
m
0.95x500x30
(1t/2)exO.005x6
1/2 ]
_1
~760
1_
~1500
mm:::::3.5mm
I1r*
m
1I2[ _1
~1330
1_ mm ~1500
G.Smm
From this we see that the brittle zones are located 3.5 and 0.5 mm from the fusion boundary, respectively. A comparison with the procedure test results in Table 7.2 shows that the measured CVN toughness after welding at these locations is slightly higher than that inferred from the weld thermal simulation experiments. This observation is not surprising, considering the fact the CVN specimens extracted from the procedure test weld, in practice, include a wide spectrum of thermal regions which have undergone highly different temperature-time programmes, whereas the microstructure within the thermally cycled CVN specimens is more homogeneous due to a similar temperature-time pattern across the whole gauge length (see Fig. 7.31). Hence, weld thermal simulation cannot replace procedure testing carried out on real welds. Nevertheless, it is a useful method of evaluating the microstructural stability and mechanical response of materials to reheating, as experienced in welding. 7.2.3 Hydrogen cracking Hydrogen embrittlement as a problem is mainly associated with ferritic steels and the risk of crack initiation in the grain coarsened HAZ following welding.57,58 As shown in Fig. 7.32, these cracks are usually situated at weld toes, weld root, or in an underbead position. Occationally, hydrogen cracks can also develop in the weld metal. A characteristic feature of hydrogen-induced cracking is that the process is time-dependent, i.e. the crack may first appear after several minutes or hours from the time of arc extinction. Consequently, the phenomenon is also referred to delayed cracking or cold cracking in the scientific literature. 7.2.3.1 Mechanisms of hydrogen cracking Hydrogen embrittlement in steels in characterised by:59,60 (i) The crystal structure dependence Hydrogen embrittlement is mainly associated with materials which exhibit a bee or a bet crystal structure, i.e. ferritic and martensitic steels. Austenitic stainless steels and aluminium alloys with a fcc crystal structure are usually not sensitive to hydrogen. The microstructure dependence A martensitic steel is generally more prone to hydrogen cracking than a ferritic steel, but a martensitic microstructure is not a requirement for crack initiation.
(ii)
510 (a)
(b)
Notch location
Weld metal
CVNspecimen
Base metal
Fig. 7.31. Methods for evaluation ofHAZ toughness (schematic); (a) Weld thermal simulation, (b) Weld procedure testing. (a)
Fig. 7.32. Schematic diagrams showing hydrogen-induced cracks in different types of welds; (a) Fillet weld, (b) Butt weld. The diagrams are based on the ideas of Coe.57
PROPERTIES OF WELDMENTS
511
10 %/min Uncharged
0.8
0.8
~ "
,~
....
0.4
0.4 0 ..200
Charged
"-r~
0
,. .J'
-100
100
-100
100
0.8 0.4
Charged
",
Test temperature,
-c
Fig. 7.33. Variation of true fracture strain with nominal strain rate and test temperature for charged and uncharged specimens. Data from Brown and Baldwin.t?
(iii)
The strain rate dependence Hydrogen embrittlement is most prominent at low strain rates typical of tensile testing, as shown in Fig. 7.33. At high strain rates the hydrogen diffusion is not fast enough to keep pace with the fracture development. The temperature dependence Hydrogen cracking occurs usually within the temperature range from -150 to +200oe. This temperature dependence reflects the fact that both the hydrogen concentration and the stress intensity at the crack tip must exceed some critical value before crack propagation occurs. The time dependence Since hydrogen embrittlement is a diffusion-controlled process, the cracks will propagate in a stepwise manner to allow for supply of hydrogen from the surrounding matrix to the crack tip (see Fig. 7.34).
(iv)
(v)
Over the years a number of mechanisms have been proposed to explain the origin of hydrogen embrittlement. The three most important are:
512
METALLURGICAL
MODELLING
OF WELDING
350 300
C9 . 250 0
c m ..c
0
Q)
CD 0)
200
~ u...
'0
:J
'00
(j)
Q)
c m
150 100
50
Applied stress:
1240 MPa
a:
J r: ..
i
stress:
....-._----A-
..rJM' ..J
1100 MPa
o~~----~----~~~--~~--~~~--~~--~ o 48 16 32
64 80 Time, min
96
112
Fig. 7.34. Example of stepwise crack propagation in notched tensile specimens, as inferred from electrical resistivity measurements. Data from Steigerwald et al. 60
(a)
The hydrogen gas pressure model, originally proposed by Zapfee and Sims.?' which postulates that atomic hydrogen will diffuse to microvoids where it recombines to form molecular hydrogen. In ferritic steels the equilibrium H2(g) pressure within the microvoids is typically of the order of 106 to 107 atm, which is more than sufficient to bring about a local fracture development. The surface energy model (Petch'<). According to this model hydrogen will reduce the effective surface energy of the crack. Under such conditions the crack can propagate at a lower nominal stress in the presence of hydrogen, in agreement with the Griffith's theory (equation (7-5)). . The slip softening model of Beachem/'" which accounts for the experimental observation that hydrogen-charged specimens generally exhibit a lower flow stress than hydrogen-free specimens. This.suggests that hydrogen interfers with dislocations in a manner which facilitates different types of fracture, including microvoid coalescence (or dimpled rupture), quasicleavage fracture, and intergranular fracture.64
(b)
(c)
Currently, it cannot be stated with certainty which of these three mechanisms that are operative under the conditions existing in welding. However, this question is of minor importance in the present context, since we here are mainly concerned with the factors responsible for hydrogen cracking in steel weldments.
PROPERTIES OF WELDMENTS
513
7.2.3.2 Solubility of hydrogen in steel Since hydrogen is the smallest of all atoms, it is readily soluble in iron. In general, both octahedral and tetrahedral lattice sites are potential traps for interstitials, as indicated in Fig. 7.35. In the case of hydrogen it is believed that the dissolved atoms are mainly present in tetrahedral positions in the form of protons.P Because of the pertinent difference in the size of the fcc and the bee interstices (see Fig. 7.35), the hydrogen solubility in iron will change stepwise with temperature following the ()Fe ~ 'YFe and 'YFe ~ UFe transformations, as shown previously in Fig. 2.7(c) (Chapter 2).
(b)
e o
a 0)
Fig. 7.35. Schematic representation of octahedral and tetrahedral lattice sites in different crystal structures; (a) Face-centred cubic (fcc) structures, (b) Body-centred cubic (bee) structures.
514
In addition to the interstitial fraction, hydrogen may be present in the form of molecular (gaseous) hydrogen trapped in microvoids or plane lattice defects. This amount is.frequently referred to as residual hydrogen, and can in many cases overshadow the equilibrium hydrogen content. For example, at room temperature the maximum solubility of atomic hydrogen in the iron lattice is estimated to be 0.001 to 0.01 ppm, while the analytical hydrogen content of steels varies typically from 1 to 10 ppm. This supersaturation is formidable and provides the necessary driving force for trapping of gaseous hydrogen in the microstructure. 7.2.3.3 Diffusivity of hydrogen in steel Published data for the diffusivity of hydrogen in steels are summarised in Fig. 7.36. At high temperatures, the diffusivity of hydrogen in ferritic steels is in reasonable agreement with the reported value for lattice diffusion of hydrogen in bee iron. However, when the temperature drops below say 200C, both the scatter and the discrepancy become more apparent due to the phenomenon of hydrogen trapping. Inclusion of the trapping effect has led to the introduction of an apparent diffusion coefficient for hydrogen in ferritic steels, D~p., which according to Oriani66 is given by:
where Dl is the lattice diffusion coefficient for hydrogen in bee iron, K is the density of trap sites (i.e. number of trap sites per number of lattice sites), and EE is the binding energy between hydrogen and the trap site. A graphical representation of equation (7-12) is shown in Fig. 7.36. A closer inspection of the graph reveals that the predicted temperature dependence of the apparent diffusion coefficient is in fair agreement with the reported diffusivity data for hydrogen in steel. Moreover, it is interesting to note that the hydrogen diffusion coefficient in austenite is nearly two orders of magnitude lower than the corresponding value for the ferrite phase at a given temperature. This observation is not surprising, considering the pertinent difference in the packing density between the fcc iron lattice and the bee iron lattice (74% and 68%, respectively). Thus, for diffusion of hydrogen in austenite, we have:57 DI(m2s-1):::::2.5x10-6exp where T is the absolute temperature (in K). 7.2.3.4 Diffusion of hydrogen in welds The thermodynamics and kinetics of hydrogen absorption in the weld pool have previously been discussed in Section 2.8 (Chapter 2). Since hydrogen is a very mobile atom (and therefore is easily lost to the surroundings), the hydrogen concentration will vary both in the longitudinal direction and in the through thickness direction of the weld, as shown in Fig. 7.37. This process will continue even after the weld has cooled down to room temperature due to the characteristic high diffusivity of atomic hydrogen in ferrite (see Fig. 7.38). ( -~ 59300) (7-13)
515
0
C
0 LO
10
co
0
0
CD
0 .q-
0 0 (\J
io
C\J
t
"'C\I
(/)
10-7
10 1: Q)
Q)
-8
==
0
== is
"'C
c 0 .Ci) ::3
<.>
10
-9
E
~
10-10
Austenitic steels
10 -11
10 -12 0.8
1.2
1.6
2.0
1000/T,K1
2.4
----
2.8
3.2
3.6
Fig. 7.36. Summary of reported diffusion coefficients of hydrogen in iron and steel. Data compiled by Coe57 and Yurioka and Suzuki. 58
Several successful attempts have been made in the past to model hydrogen diffusion in welds by means of numerical methods.68-7o Unfortunately, none of these solutions are simple enough to get a good overall indication of the hydrogen redistribution during cooling and subsequent PWHT. As an illustration of principles, we shall therefore present a simplified analytical solution to the hydrogen diffusion problem in welding, based on an analogy between diffusion and heat conduction. Diffusion model The idealised model considers a butt weld of uniform hydrogen concentration in the longitudinal direction, as shown in Fig. 7.39. The width of the fusion zone is 2L, while the initial hydrogen concentration at the time of solidification (i.e. at t = 0) is C; The hydrogen concentration in the base metal outside the fusion zone is Co. If element losses to the surroundings are neglected, the problem can be treated as uniaxial diffusion in an isotropic solid analogous to that described in Section 1.7 (Chapter 1) for heat conduction in thermit welding. Thus, in the limiting case where the diffusion coefficient can be regarded as constant, the hydrogen concentration (C) as a function of time (t) and distance (y) is given by equation (1-22):
516
..':c
o
I
0') C\J
Position x, mm
~
o
-L
~ Diffusible Residual
Fig. 7.37. Measured longitudinal and lateral distributions of hydrogen in a single pass SMA weld quenched right after welding. Data from Christensen et al.67
Q-
C-Co -~-~-2
_l[eif(
~4DHt
y+L J-eif(
~4DHt
y-L
J]
(7-14)
where DH is the hydrogen diffusivity, and eif(u) is the Gaussian error function (defined previously in Appendix 1.3, Chapter 1). In practice, it is necessary to rewrite equation (7-14) in a differential form to allow for the variation in the hydrogen diffusion coefficient with temperature. After some manipulation, we obtain: (7-15)
(y+ L)
4~
1tDH t3/2
exp
[(y+
L)2] 4D
H t
PROPERTIES OF WELDMENTS
517
Ill/cm3
a)
a
b
600
(f)
~::i
400 200
b)
12 h
Fig. 7.38. Redistribution of hydrogen following welding (numerical calculations); (a) Right after welding, (b) After 12 h at room temperature (IJlI cm ? = 0.0115 ppm). Data from Christensen.f
518
This differential equation can be integrated numerically in temperature-time space when the weld thermal programme is known. The boundary conditions are as follows: y<L:
.Q = 1 when t = 0, .Q ~
0 when t ~
00
y>L:
.Q = 0 when t = 0, .Q ~
0 when t ~
00
Although the above model does not give a true physical picture of the hydrogen redistribution in butt welds, it may provide valuable quantitative information about the extent of hydrogen diffusion occurring during cooling from the solidification temperature under different welding conditions. Figure 7.40(a) shows a sketch of a 2mm thick single pass butt steel weld made by means of the GTA process. For the purpose of convenience we shall assume that the temperature field around the heat source is given by the simplified Rykalin thin plate solution (equation (1-100) in Chapter 1). Thus, at (T -To) = 1500C a total width of the fusion zone of about 4.3mm is obtained for a net heat input of 67.2 J mrrr". The corresponding distance from the weld centre-line to the 1350C HAZ isotherm is 2.5mm. Figure 7 .40(b) shows computed temperature and hydrogen concentration profiles at the centre of the weld (y = 0) and in the grain coarsened HAZ (y = 2.5mm) for a chosen ambient temperature of 20C (no preheating). As expected, the hydrogen concentration within the fusion zone itself (y < L) decreases in a monotonic manner as the weld cools down. In contrast, the hydrogen concentration outside the fusion zone (y > L) is seen to pass through a local maximum. In the absence of preheating this maximum is attained after very long times, which may initiate hydrogen cracking in the grain coarsened HAZ if the microstructure is martensitic. The picture is completely changed if the ambient temperature is raised to 1000 (moderate preheating). As shown in Fig. 7 .40( c), the main effect of preheating is to decrease the cooling rate in the low-temperature regime (i.e. below 50QoC) after the completion of the austenite to ferrite transformation. The HAZ microstructure is therefore not significantly altered, but instead more hydrogen is allowed to diffuse out of the weld region before the temperature drops below the critical value where hydrogen cracking may occur. This is seen as a shift in the peak hydrogen concentrations towards lower absolute values and shorter times in Fig. 7.40( c).
7.2.3.5 Factors affecting the HAZ cracking resistance Safety against hydrogen cracking is an important aspect of weldability. In spite of the knowledge accumulated and the improvements made over the past decades, the current trends towards stronger steels and heavier sections still require continuous attention to the risk of cracking. Cold cracking test methods In principle, a proper weldability criterion should enable the user to select combinations of steel, consumables and operational conditions that will ensure sufficient crack safety at a minimum of total cost. It should also enable him to examine the effects of these main variables separately and establish a quantitative grading system for safety which takes into account the consequences of failure. In recognition of this situation, a number of empirical cracking test methods has been de-
PROPERTIES OF WELDMENTS
519
j.- 2L = 4.3 mm -1
E E
L
(b)
C\I
2000
0
0
I 1500
e ::l
CD CD a. E
0.8
0.6
Y
~
I
I
Q
(j0
~ 1000
0.4
t-
500
0.2
0 5 10
16
(c)
10
10
1cr
Time,s
10
10
1.0
2000
o
0
1500
0.8
~ 1000
(1)
'::J E
cD
0.6
y
~
I
a
" ::::=...
0
a.
ICD
0.4
500
0.2
Time,
..
Fig. 7.40. Computed temperature and hydrogen concentration profiles during GTA butt welding of a 2mm thin steel sheet (Case Study (7.2)); (a) Sketch of weld, (b) Redistribution of hydrogen in the absence of preheating, (c) Redistribution of hydrogen after preheating to 100C.
520
veloped over the years to study the mechanisms of hydrogen cracking in weldments (e.g. see the review ofYurioka and Suzuki'"), Broadly speaking, the cold cracking tests fall into either one of the two categories, i.e. self-restrained tests or externally loaded tests. Examples of the former type are the Tekken (oblique Y-groove) cracking test, the CTS (controlled thermal severity) cracking test, and the cruciform cracking test. Well-known externally loaded tests are the implant cracking test, the TRC (tensile restraint cracking) test, and the RRC (rigid restraint cracking) test. The implant method The implant technique is a good example of a cracking test method which allows separate assessment of the various metallurgical and operational factors that contribute to hydrogen cracking in welds. As shown in Fig. 7.41 the Scandinavian version of the implant test employs a cylindrical test bar made from the steel to be examined which is notched (threaded) at some distance from one of its ends." The bar is inserted into a reamed hole in the backing plate of a similar grade of steel, so that the threaded end is flushed with the plate surface. In order to execute a test, the hole is sealed off with a single weld bead deposited on the plate under carefully controlled conditions. After a specified delay of 60 s per kJ mrrr! gross heat input, the implant assembly is quenched and put under a static tensile load at room temperature. The stress applied is subsequently reported on the nominal cross-section of the bar. After a specified loading time (if no rupture has occurred), the assembly backing plate with the test bar may be sectioned and examined with respect to microcracks in the HAZ. The implant test may be performed at different stress levels, cooling rates and hydrogen contents. The implant rupture strength, R1R, is obtained in a programme of stepwise loading, going down or up by a fixed amount depending on rupture or not. It is then defined as the nominal stress at which the statistically probability of rupture is 50% when the load is applied for a long period. Implant test results The critical stress for rupture in the Scandinavian version of the implant test may be written in a differential form as a sum of four contributions:
Backing plate
Tensile load Fig. 7.41. Schematic representation of the implant test method.
521 (7-16)
dR
MIR
CEw
a log H FM dlogHFM
dR1R
The CEw parameter in equation (7-16) refers to the so-called IIW carbon equivalent, originally developed for C-Mn steels: CE
w=v
Oh C
%Mn +-6-+
%Cr+%Mo+%V 5
%Ni+%Cu 15
(7-17)
Moreover, Llt333 is a hydrogen diffusional parameter which takes into account variations in the measured implant rupture strength after various thermal treatments (including preheating and PWHT). According to Christensen and Simonsen," the extent of hydrogen diffusion which occurs in the low-temperature regime can be reported in the form of an equivalent isothermal hold time at 60C (or 333K), defined as:
T
dt
(7-18)
where T; refers to the local HAZ temperature at the moment of quenching (usually taken as 100C). It follows from equation (7-16) that the two first members reflect the influence of microstructure upon the implant rupture strength, and is therefore related to the HAZ peak hardness. The two last members take into account the effect of analytical HFM and local hydrogen concentrations. As shown in to Fig. 7.42, the numerical values of the partial derivatives dR1R/ dCEw, dR1Rldl1t8/5, dRIRld log HFM, and aR1R / a~ ~t333 may vary within relatively wide limits, depending on the steel chemical composition and the operational conditions applied. Nevertheless, the concept is still useful for quantitative predictions of the HAZ cracking resistance, as illustrated below.
Example (7.4)
Experience has shown that conventional pipeline steels with carbon equivalent CEw up to 0.4% can be welded with basic electrodes (L\t8/5 ~ 8-9 s) without the use of external preheating, provided that the weld metal hydrogen content is kept sufficiently low (HFM ~ 4 ppm). Suppose that the same procedure shall be employed in hyperbaric welding of pipeline steels at a depth of 320 m (33 bar total pressure). Based on the implant test data in Fig. 7.42, estimate the minimum reduction in the steel carbon equivalent (L\CEw) which must be incorporated in the specifications to compensate for the increased hydrogen absorption observed at such depths (HFM ~ 10 ppm).
Solution
The concept of partial derivatives implies that we will have the same safety against hydrogen cracking if there is no net change in the implant rupture strength (i.e. MIR = 0). Since the weld
522
(a)
800
HSLA steels
as a..
600
~-0:
a:-
400
200
H
FM
=3-4 ppm
0 0.30
0.40
CEw,
% ------
0.50
(b)
as a..
~
a:-
a:
400 HFM:
200
20 ssm
3ppm 6 ppm
O~------~------~--------~----I o 8 4 12
Fig. 7.42. Examples of implant test results: (a)
R/R
vs CEw, (b)
RIR VS Ilt8/5.
cooling programme is similar in both cases, the variation in ~t8/5 Hence, equation (7-16) reduces to: !:ill
IR
dR = -1lL~CE dCE
W W
dR IR ~logH dlogH FM
FM
=0
PROPERTIES OF WELDMENTS
523
(c)
800
HSLA steel
t
co a.
600
4~/5:::: 4~/5::::
8.6 S 4.3
S
(d)
1 min
1/4 h
1/2 h
1h
600
co n,
3 ppm 400
HFM:
0'"
"
'"
"
" .JJ OT 2
".-
..-QT3
C:
./
".-
".-
".-
".-
OT3
Fig. 7.42. Examples of implant test results (continued); (c) RIR Christensen and Simonsen."
VS
In the present example the total change in the weld metal hydrogen content between 1 and 33 bar is equal to: ~ log HFM= log 10 -log 4 = 0.4
524
METALLURGICAL
MODELLING
OF WELDING HFM
Moreover, the numerical values of aR1RlaCEw and aR1Rla log 7.42(a) and (c), respectively. When 11t8/5 ~ 8.6 s, we obtain:
see;
and
aR1R ~ -1850
This gives:
I1CEw:::::
-0.05
The above calculations suggest that the CEw carbon equivalent of pipeline steels should not exceed 0.35% if hydrogen cracking is to be avoided under hyperbaric welding conditions. 7.2.4 H2S stress corrosion cracking Hydrogen sulphide (H2S) stress corrosion cracking is a well-known phenomenon taking place in steels in environments containing sour oil and gas. As shown in Fig. 7.43 this type of cracking arises from corrosion reactions with subsequent absorption of hydrogen in the metal. The embrittlement mechanisms are therefore similar to those reported for hydrogen cracking in steels, and can be evaluated from standard test methods."? 7.2.4.1 Threshold stress for cracking For a given combination of steel, microstructure and H2S concentration there exists a lower limit for the imposed stress where cracking no longer will occur. Experience has shown that the threshold stress, O'th, is related to the yield strength RpO.2 through the following equationr'''
o.J
(7-19)
Equation (7-19) predicts that the threshold stress (and thus the steel cracking resistance) passes through a local maximum as the yield strength increases. The locus of this peak stress is obtained by setting dO'th1d RpO.2 = 0, which gives RpO.2 ~ 600MPa and O'th (max) ~ 360MPa. In practice, a hardness criterion rather than a yield strength criterion is used for ranking of steels with regard to H2S stress corrosion cracking resistance. According to Dieter,19 the following relation exists between RpO.2 and HV: R~
0.2
(MPa) ~ 3. 33(HV)[O.l]m
(7-20)
where m is the strain hardening exponent in the Ludwik equation. The value of m may vary between wide limits, depending on the steel chemical composition and the heat treatment conditions applied, but for HSLA steels m ~ 0.15 is a reasonable compromise.'? In that case the observed maximum in the cracking resistance at RpO.2 ~ 600MPa
PROPERTIES
OF WELDMENTS
525
Stress
Anode:
Crack
Embrittled
zone
Stress
Fig. 7.43. Mechanisms of hydrogen absorption in cathodic stress corrosion cracking (schematic).
corresponds to a hardness of about 250VPN. This value should be compared with the maximum hardness level of 22HRC (Rockwell C) or 248VPN incorporated in many offshore specifications. 7.2.4.2 Prediction of HAZ cracking resistance Prediction of the H2S stress corrosion cracking resistance based on equation (7-19) requires quantitative information about the HAZ strength level. The peak strength for various combinations of steels and welding conditions can easily be read from diagrams of the type shown in Fig. 7.20 or calculated from diverse empirical formulae.26,74 Figure 7.44 shows examples of computed (Jth - dts/5 profiles for three different types of steels spanning a range in the CEIl carbon equivalent from 0.41 to 0.50%. It is evident from these plots that the cooling time required to obtain the maximum threshold stress depends on the steel chemical composition. In general, ultra-low-carbon steels will exhibit the highest resistance against stress corrosion cracking, since a low HAZ hardenability will eliminate problems with martensite formation in the grain coarsened region during the -y to ex transformation. On the other hand, these steels suffer from a severe HAZ softening during high heat input welding, with consequent reduction in the threshold stress. Under such conditions it may be safer to use ordinary low-carbon microalloyed steels or C-Mn steels with a higher HAZ hardenability to compensate for the observed strength loss. Equation (7-19) can also be employed for assessment of the relevance of current hardness requirements. Figure 7.45 shows the same types of plots as in the preceding figure (on a normalised scale) where typical ranges for the HAZ hardness are indicated. It is evident that the threshold stress is rather insensitive to small variations in the hardness level. Only in cases where the HAZ hardness exceed 300VPN or drops below say 220VPN a significant deterioration in the H2S stress corrosion cracking resistance is to be expected. This suggests that the maximum hardness requirement of 248VPN incorporated in many offshore specifications is too stringent, since it imposes severe restrictions on the use of steels in welded structures without improving the service performance to any great extent.
526
METALLURGICAL
MODELLING
OF WELDING
)
ctS 0..
400
:a: ur en
Ci)
"'0
/'
300
CD ...
-0
.c en
CD
200
... ..c
I100
----
o------~----~--~~--~----~----~----~ o 10 20 30
/}.t
Ultra-Iow ..carbon steel (CEn=0.41 Low-carbon steel (CEn=O.45 %) C..Mn steel (CEu=O.50o/0)
%)
s/5
,s
ath -tl.tS/5
t
CI) CJ)
1.0
1i) 0.8
"C
(5
rv340 HV
e 0.6 :5
"C
s: CJ)
.~
Q)
~ o
0.4
--Ultra-low-carbon Low-carbon steel (CE =O.41
0.2
%)
O----~~--------~---~~--~----L---~ o 10 30
Fig. 7.45. Effect of peak hardness on the HAZ stress corrosion cracking resistance.
PROPERTIES OF WELDMENTS
527
As shown in Table 7.3, the welding of stainless steels is encumberred by a number of different metallurgical problems, including solidification cracking, hydrogen cracking, precipitation reactions, and grain growth. Some of these problems will be discussed below in the light of information available in recent Iiterature,">" 7.3.1 HAZ corrosion resistance Austenitic stainless steels containing about 0.07 wt% carbon or more are often susceptible to intergranular corrosion in the weld HAZ due to chromium carbide precipitation (see Section 6.4 in Chapter 6). This phenomenon is frequently referred to as the 'weld decay' in the scientific literature. 76
Major Elements Cr: 17-25 wt% Ni: 8-20 wt% Mn: -2wt% C: 0.03-0.15 wt% Cr: 18-27 wt% Ni: 7-8 wt% Mo: 1-4wt% C: < 0.1 wt% Cr: 12-30 wt% Mn: 1-1.5 wt% C: < 0.15 wt% Cr: 11-18 wt% Mn: -1.0 wt% C: 0.15-1.2 wt%
Minor Elements
Welding Problems HAZ chromium carbide precipitation HAZ grain growth Solidification cracking
Ti,N
Shift in the HAZ austenite/ferrite balance Reduced HAZ toughness Precipitation of a-phase, Cr2N etc. HAZ grain growth Low HAZ toughness HAZ chromium carbide precipitation Hydrogen cracking High HAZ hardness and transformation stresses Low HAZ toughness
AI, V, Zr
Ni,Mo,W,V
High resistance against general corrosion against stress corrosion cracking (in the presence of chlorides)
Adequate low-temperature
toughness
528
Like weld decay, knife-line attack is also related to intergranular Cr23C6-precipitation, but it differs from the former type in that the corrosion attack occurs in a narrow region adjacent to the fusion boundary following multipass welding ofTiINb stabilised stainless steels. As shown in Fig. 7.46, the origin of knife-line attack can be attributed to dissolution of TiC or NbC during the initial weld thermal cycle in regions of the HAZ where the peak temperature has been above say 1250C. Since the cooling rate within the high temperature regime is very high, reprecipitation of TiC or NbC does not take place on cooling. This leaves carbon free to react with chromium during subsequent reheating to peak temperatures of about 700-800C following deposition of the second layer (filler pass). Knife-line corrosion attack is usually observed in weldments where the second pass (top) (a)
~ .a ~
Q)
a.
r-
E Q)
Time
(b)
Time Fig. 7.46. Mechanisms of knife-line corrosion attack in austenitic stainless steel weldments (schematic); (a) Dissolution of TiC or NbC during the initial weld thermal cycle, (b) Precipitation of Cr23C6 in the low peak temperature region of the weld HAZ following deposition of the second layer. (The corresponding C-curves for precipitation of TiClNbC and Cr23C6 in the grain growth zone are displaced far to the right in the diagram.)
PROPERTIES OF WELDMENTS
529
overlaps the lower part of the fusion boundary of the first (root) pass, as indicated in Fig. 7.47(a). In practice, the problem can be eliminated by simply reversing the welding sequence or by changing the welding procedure so that the second pass overlaps the middle rather than the lower part of the root pass. The latter point is illustrated in Fig. 7.47(b). 7.3.2 HAZ strength level In single phase materials such as austenitic stainless steels, the primary grain size will make a direct contribution to strength. In general, the grain size dependence of the yield stress (0'0) is well described by the Hall-Petch relation: 19 (7-21) where c, is the friction stress (representing the overall resistance of the crystal lattice to dislocation movement), k is the locking parameter (which measures the relative hardening contribution of the grain boundaries), and D is the average grain diameter. It follows from the analysis in Section 5.4.2.5 (Chapter 5) that the grain size across the HAZ of austenitic stainless steel weldments may vary by a factor of three to five, depending on the
(a)
(b)
Fig. 7.47. Effect of welding performance on the corrosion resistance of TilNb-stabilised austenitic stainless steels; (a) Low resistance against knife-line corrosion attack, (b) High resistance against knife-line corrosion attack. The diagrams are based on the ideas of Kou.76
530
applied heat input. This means that a permanent soft zone will form adjacent to the fusion boundary after welding, which may reduce the overall load-bearing capacity of the joint.
Example (7.5)
Consider plasma arc butt welding of a 5mm thick plate of type 316 austenitic stainless steel under the following conditions:
Provided that the conditions for one dimensional heat flow are met, estimate on the basis of the nomograms in Fig. 5.30(b) (Chapter 5) the variation in the austenite grain size across the HAZ after welding. Calculate then via the Hall-Petch relation (equation (7-21)) the expected reduction in the HAZ strength level due to this change in the microstructure. Input data: Base metal yield strength: 300 MPa Locking parameter in Hall-Petch relation: k = 227 MPa Jlm1l2
Solution
First we need to calculate the net heat input per mm" of the weld:
(q
o
/vd)=_4-kJmm-2
8x5
=0.lkJmm-2
Readings from the nomograms in Fig. 5.30(b) give the HAZ grain size profile shown in Fig. 7.48. In order to obtain the resulting HAZ strength distribution, it is necessary to fix the value of the friction stress, O'j, in the Hall-Petch relation. In the present example, we have:
0'.
1
0'
from which
0' o
= 246.5
+ 227(1))-112
It follows from the graphical representation of the Hall-Petch relation in Fig. 7.48 that the variation in the yield strength across the HAZ is rather small under the prevailing circumstances. In fact, the maximum HAZ strength reduction which may occur in such materials because of grain growth is about 18%, corresponding to high heat input welding conditions. This implies that welding does not impose severe restrictions on the design stress as long as the steel is used in the fully annealed condition. 7.3.3 HAZ toughness All body-centred cubic metals (including ferritic and martensitic stainless steels) show a marked temperature dependence of the fracture toughness, as indicated in Fig. 7.49. At high temperatures fracture occurs normally by ductile rupture, whereas cleavage is the dominating fracture mode at low temperatures. This type of behaviour makes both ferritic and martensitic stainless
PROPERTIES
OF WELDMENTS
531
E ::i.
'Ci) c:::
60 275
.~
Q)
:5
0>
g
~
~
1i)
40
250
(j)
"0
20
:::s
CD
>=
Fig. 7.48. Computed HAZ grain size and yield strength profiles in a Smm thick butt weld of type 316 austenitic stainless steel (Example (7 .S.
>. ~
c
Q) Q) Q)
"'0
-e
~
o C/)
Test temperature
Fig. 7.49. Effect of temperature on notch toughness (schematic).
steels unsuitable for many structural applications where the HAZ toughness is of particular concern. In contrast, medium- and high-strength fcc metals have usually such high toughness that brittle fracture is not a problem at low temperatures, unless there is some special reactive chemical environment. Austenitic stainless steels and aluminium alloys fall within this category. When it comes to duplex stainless steel weldments, the situation is more complex. Here the HAZ toughness is determined by the austenite/ferrite balance in the weld, which, in
532
turn, depends on the steel chemical composition and the operational conditions applied.78-8o As shown in Fig. 7.50, complete ferritisation is normally achieved in regions close to the fusion boundary during the initial heating leg of the thermal cycle. Provided that the cooling rate through the critical transformation temperature range is kept reasonably low, a significant proportion of the ferrite may retransform back to austenite on cooling (see Fig. 7.51). This will contribute to a high HAZ toughness, even at subzero temperatures, as indicated by the CVN transition curves in Fig. 7.52. In general, an austenite content of about 30 vol % is sufficient to avoid problems with the HAZ toughness in duplex stainless steel welds. 7.3.4 Solidification cracking When a high alloy steel solidifies, a variety of microstructures can developed, depending on the steel chemical composition and the cooling conditions applied. Referring to the Fe-Ni-Cr phase diagram in Fig. 7.50, compositions on the Ni-rich side of the peritectic/eutectic liquidus solidify as primary austenite, while those on the Cr-rich side solidify as primary delta ferrite. The former condition is known to promote solidification cracking in welds due to partitioning of low-melting-point segregates of sulphur and/or phosphorus to the 'YFe-'YFe grain boundaries.?6-78 Since the solubility of sulphur and phosphorus is higher in delta ferrite than in austenite (0.18 and 2.8 wt%, respectively vs 0.05 and 0.25 wt% in austenite), the cracking susceptibility can normally be reduced if delta ferrite is present in sufficient amounts." At the same time, the interfacial 'wetting' conditions are significantly improved owing to the formation of a low-
maited zone = = =
1400 1200
t
P
1000
ca
CD
~::l
Co
Fig. 7.50. Schematic diagram defining different thermal regions within the HAZ of a single pass duplex stainless steel weld.
PROPERTIES OF WELDMENTS
533
30-45s at 1300
-c
o ~
.a
~
PI
Q) a. E
10
100
1000
Time,s --Fig. 7.51. CCT-diagram for a duplex stainless steel. Data from Mundt and Hoffmeister,"?
energy interface between 'YPe and OPe, which prevents spreading of the liquid along the grain boundaries.F For these reasons, a minimum weld metal delta ferrite content of about 5 to 10 vol % is usually specified for austenitic stainless steels. The quantitative relationship between the delta ferrite content and the weld metal chemical composition in austenitic stainless steels has been determined first by Schaeffler=' and later by Delong et al.84,85 The constitution diagram of Delong is shown in Fig. 7.53. Here the alloying elements are grouped into ferrite formers (i.e. Cr, Mo, Si, and Nb) and austenite formers (i.e. Ni, C, N, and Mn) to determine the corresponding chromium and nickel equivalents for a given alloy. The Delong diagram differs from the Schaeffler diagram in that the important nitrogen contribution also is included in the former, thus allowing a more accurate prediction of the weld metal delta ferrite content.
Example (7.6)
Consider plasma arc butt welding of a 5mm thick plate of austenitic stainless steel under conditions similar to those employed in Example (7.5). Data for the base metal (BM) and the filler wire (FW) chemical compositions are given in Table 7.4. Use this information together with the Delong diagram in Fig. 7.53 to determine which of the two filler wires (lor II) that provides the highest resistance against weld metal solidification cracking under the prevailing circumstances. In the present example we shall assume that the mixing ratio B/(B+D) is 0.57 (the mixing ratio is defined in Section 1.10.8, Chapter 1).
Solution
In the absence of oxidation losses, the weld metal chemical composition is given by the 'rule of mixtures' :
534
conditions
Peak temperature:
1300C
.. .. 0
--. -.-
-A-'-
20 s 10 S 5s
300
>.
Q) Q) Q)
E>
c:
"C
200
1: 0
.0
100
en
.
A.
[]
.......
. .....
.-
........
.............
...
.",-
..
..,.,.-
_.-'
.,.",.
. ....- ..
_ .....
[]
.... 0
t:;J .
. . . . . ......[]
9.
.. 40
o
Data from
Fig. 7.52. CVN transition curves for a duplex stainless steel after weld thermal simulation. Videm.s! Table 7.4 Chemical composition of base plate and filler wires used in Example (7.6).
Material" Base plate Filler wire I Filler wire II wt%C 0.03 0.03 0.03 wt%Cr 17.0 18.0 18.0 wt%Ni 12.0 11.5 15.5 wt%Mo 2.0 2.7
-
wt%N
-
0.07 0.07
= --[%X]BM
B+D
+ --[%X]FW
B+D shown in Table 7.5. We can now
This leads to the weld metal chemical compositions calculate the Cr- and Ni-equivalents for both welds: [Cr] I [Cr]n 17.4 17.4 11.8 13.5
[Nih
[Ni]n
+ 2.3 + 1.5 X 0.9 = 21.1 + 1.1 + 1.5 X 0.9 = 19.9 + 30 x 0.03 + 30 X 0.03 + 0.5 X 1.5 = 14.4 + 30 X 0.03 + 30 X 0.03 + 0.5 X 1.5 = 16.1
PROPERTIES OF WELDMENTS
535
c
::2;
)(
?fl. ci
10
z
)(
o ('t) + (J + Z
CD
'* '*
o ('t)
x
19
18
Austenite
17 16
~ ~
E
0-
14 13 12 11 17 18
S
CD
Austenite-terrtte
~ o Z
19
20
21
22
23
24
%
25
26
27
---
Fig. 7.53. The Delong diagram'" showing the relationship between delta ferrite content and weld metal chemical composition for stainless steels.
Readings from Fig. 7.53 then give: Weld I : About 93 vol % austenite and 7 vol % delta ferrite Weld II: 100 % austenite From this we see that wire I provides the highest resistance against weld metal solidification cracking for the combination of steel and operational conditions considered above. It should be emphasised that the Delong diagram gives no information about the real solidification microstructure, since it is based on measurements of retained delta ferrite at room temperature. Also, the important effect of cooling rate on the weld metal transformation behaviour is neglected in the present analysis. Consequently, the use of such empirical diagrams for selection of steel and welding consumables is a keenly debated question in the scientific literature.
536
Mn, Cr,Fe
Fe, Ti
PROPERTIES
OF WELDMENTS
537
AI~Mg system
O~
~ 2
~ 3
L-
4
------
Fig. 7.54. Solidification cracking sensitivity of binary AI-Si and AI-Mg alloys. Data from Dudas and Collins.86
may arise, however, if the weld metal becomes heavily depleted with respect to Si or Mg due to dilution with the parent metal. Welds of a high B/(B+D) ratio fall within this category. In such cases it may be necessary to use over-alloyed filler wires to obtain crack-free welds.
Example (7.7)
Consider plasma arc butt welding of a 10mm thick aluminium plate of type AA 6082-T6 (containing 0.7 wt% Mg and 0.9 wt% Si) under the following conditions:
Experience shows that the bead reinforcement amounts to 10% of the groove cross section (details of the groove geometry are given in Fig. 7.56). Two different filler wires are available, wire I with 5 wt% Si and wire II with 5 wt% Mg. Use this information along with the diagrams in Fig. 7.54 to determine which of the two filler wires (lor II) that provides the highest resistance against weld metal solidification cracking under the prevailing circumstances. In these calculations we shall assume that the temperature field around the heat source is given by the simplified Rykalin thin plate solution (equation (1-100) in Chapter 1). Thermal data for AIMg-Si alloys are contained in Table 1.1.
Solution
First we need to estimate the mixing ratio B/(B+D). can be obtained from equation (1-100):
10 x
2(20xl0 ) 10 x O.004-V21te(632)
mm=38.3mm
538
Fig. 7.55. Examples of solidification cracking in aluminium (a) AI-l wt% Si, (b) Al-l wt% Mg. After Cross.F
weldments
PROPERTIES OF WELDMENTS
539
2mm Fig. 7.56. Groove geometry for a single pass AI-Mg-Si butt weld (Example (7.7)).
This gives:
B = [38.3 x 10 - (102 tan 30+2 x 10)]mm2
::::: 306mm2
and
D = 1.1(102 tan 30+2 x 10)mm2
:::: 86mm2
from which
B (B+D)
= 0.78
In the absence of oxidation losses, the weld metal chemical composition is given by the 'rule of mixtures' :
[%X]weld = --[%X]BM
B+D
+ --[%X]FW
B+D
If wire I is used, the weld metal Mg and Si concentrations become: [%Mg]I [%Si]I
Similarly, in the case of wire II, we get: [%Mg]n [%Si]n = 0.78(0.7) + 0.22(5) 0.78(0.9) + 0.22(0)
= 1.65 wt%
= 0.70
wt%
A comparison with Fig. 7.54 shows that wire I provides the highest safety against weld metal solidification cracking under the prevailing circumstances, while wire II is unacceptable. The lower cracking resistance of the Al-Mg wire compared with the Al-Si wire at high dilution ratios arises from the pertinent difference in the fraction of eutectic liquid which forms during weld metal solidification. For pure binary alloys the eutectic fraction feut. is given by equation (3-46) in Chapter 3:
f eut. = [ zsu: C
o
]1/(ko -1)
540
METALLURGICAL
MODELLING
OF WELDING
where Ceut. is the eutectic concentration, and k; is the equilibrium partitioning coefficient (given by the binary AI-Si and AI-Mg phase diagrams). If the contribution from the accompanying alloying element is neglected, the values of feut. become: Weld I (1.80 wt% Si):
f eu t .
Weld II (1.65 wt% Mg):
(~)1/(O.14-1) ~
1.80
f eu.t
(~)1/(0.43-1) ~
1.65
0.005,
i.e.0.5vol%
From this we see that the fraction of eutectic liquid in weld I is so abundant that it backfills and 'heals' all incipient cracks, while feut. in weld II is just large enough to form continuous films at the columnar grain boundaries which, in tum, promotes solidification cracking. 7.4.2 Hot cracking Hot cracking is a phenomenon occurring within the high peak temperature regions of the HAZ during welding of aluminium alloys. In spite of the difference in location, hot cracking like solidification cracking is intergranular and arises from the combined action of grain boundary liquation and stresses induced by solidification shrinkage and thermal contraction. The formation of liquid phases within a weld HAZ can readily be explained with the help of a simple binary phase diagram of the type shown in Fig. 7.57. If the composition of the binary alloy is
Liquid (L)
A Composition _
Fig. 7.57. Schematic binary phase diagram defining the equilibrium conditions for partial melting during reheating.
PROPERTIES OF WELDMENTS
541
higher than Cmax., and the alloy is heated to a temperature above Teut., partial melting will occur. Since the eutectic phase is usually located at the grain boundaries, these sites become immediately covered with liquid films if the wetting conditions are favourable. 7.4.2.1 Constitutional liquation in binary Al-Si alloys Experience shows that even binary alloys with a nominal composition lower than Ceut. can undergo incipient melting at the eutectic temperature. This will obviously be the case if the alloy contains segregations so that the concentration locally exceeds the critical composition. Another possibility is melting due to 'constitutionalliquation', a mechanism originally proposed by Pepe and Savage88,89 for the formation of hot cracking in 18-Ni maraging steel weldments at temperatures well below the bulk solidus of the alloy. The same theory has later been applied by Reiso et al. 90-93 and Lohne and Ryum?" to explain incipient melting during homogenisation heat treatment of aluminium alloys. In the following, the treatment of Lohne and Ryum'" is adopted to illustrate the principles of constitutional liquation in binary AI-Si alloys under conditions applicable to welding. The theory is later extended to ternary AI-MgSi alloys. Figure 7.58 shows a section of the Al-rich corner of the binary AI-Si phase diagram. Consider next a pure AI-l.2 wt% Si alloy that first is brought to equilibrium at a lower temperature to form 10J..lmlarge Si particles (Fig. 7.59(a)) and then is rapidly heated to a higher temperature slightly above Teut.. If local equilibrium is maintained at the particle/matrix interface during heating, the composition is given by the AI-Si solvus boundary in Fig. 7.58. When Ceut. is reached incipient melting will take place by formation of eutectic liquid (according to the reaction Al + Si ---7 liquid), as shown in Fig. 7 .59(b). The process will continue until the whole Si particle has dissolved (Fig. 7.59(c)). Depending on the density/location of the Si particles and the interfacial wetting conditions, grain boundary liquation may eventually occur, as indicated in Fig. 7 .59( d).
700~----~-------~----~----~----------~----~----~
Liquid
1 o
o
600
Liquld-solid
~ 500
:::l
Q)
~ ~
E 400
300
AI+Si
200------------~~--~----------------~----~----~
AI 2 3
Silicon content, wt% Fig. 7.58. Section of the AI-rich comer of the binary AI-Si phase diagram. ..
542
(a)
(b)
(c)
(d)
Fig. 7.59. Optical micrographs showing the microstructural evolution during homogenisation heat treatment of binary AI-Si alloys; (a) Isolated Si particle embedded in a matrix of AI, (b) Partially melted Si particle surrounded by AI-Si eutectic (25s at 582C), (c) Globular AI-Si eutectic structure formed after complete dissolution of the Si particle (60s at 582C), (d) Spreading of eutectic liquid along a grain boundary (25s at 582C). Courtesy of O. Lohne, Sintef - Division of Metallurgy, 7034 Trondheim, Norway.
7.4.2.2 Constitutional liquation in ternary Al-Mg-Si alloys Constitutional liquation in ternary AI-Mg-Si alloys has been investigated by Reiso et at. 96 Figure 7.60 shows a section of the AI-rich comer of the quasi-binary AI-Mg2Si phase diagram. When an alloy of composition C < Ceut. is heated up rapidly, as in the case of welding, the primary Mg2Si particles do not have enough time to dissolve. When the quasi-binary eutectic temperature is reached at about 593C, incipient melting will occur, according to the reaction:96 (7-22) Further heating to a higher temperature (T> 593C) provides additional time for dissolution of Mg2Si and formation of more liquid of variable composition. In practice, reaction (7-
PROPERTIES OF WELDMENTS
543
0.4
0.8
1.2 wt%
1.6
M92 Si concentration,
Fig. 7.60. Quasi-binary section of the Al-rich corner of the ternary AI-Mg-Si Reiso et al. 96 and Phillips."?
22) may also proceed below the quasi-binary eutectic temperature Teub since the diffusion rate of Mg is higher than that of Si.96 By considering the kinetics it can be shown that the interface concentration will gradually move towards the silicon side of the quasi-binary line in the phase diagram as the Mg2Si particles dissolve, thereby reducing the temperature at which liquation occurs (from 593 down to about 580C). Moreover, in alloys containing excess amounts of silicon, two other side reactions can take placer"
a - AI+ Si
and
-7
liquid
(7-23) (7-24)
The former is analogous to the melting reaction in binary AI-Si alloys (Fig. 7.58), while the latter corresponds to the eutectic reaction in the ternary AI-Mg-Si system (Teut. varies from 555 to 559C, depending on the source). Since this eutectic temperature represents the lowest temperature at which a melt can exist within the system, it means that local melting of second phase particles cannot take place below, say, 555 to 559C during welding of AI-Mg-Si alloys.
544
7.4.2.3 Factors affecting the hot cracking susceptibility In addition to the thermodynamic and kinetic effects mentioned above, there are several other factors, some interrelated, which play an important part in the formation of hot cracks in AI-Mg-Si weldments. These are:76 (i) The number density and size distribution of Mg2Si and Si particles in the base metal. The total grain boundary area per unit volume available for absorption of eutectic liquid (determined by the HAZ grain size). The interfacial wetting conditions. The local tensile stress level in the partially melted region.
(ii)
(iii) (iv)
It follows that the hot cracking susceptibility may be significantly altered by a change in one of these parameters, but, in practice, there is very little that can be done to prevent grain boundary liquation if the base material already contains second phase particles. In fact, the only useful way of eliminating the cracking problem is to reduce the tensile stress in the partially melted HAZ through proper selection of welding consumables. This is because the composition of the weld metal can be adjusted so that solidification is completed first in the partially melted region and then in the fusion zone, thus avoiding hot cracking in the former. As an illustration of principles, the diagrams of Gittos and Scott 98 (reproduced in Fig. 7.61) will be considered. These diagrams show the variation of the weld metal solidus temperature with base metal dilution for two commercial filler wires (wire I: AI-5 wt% Si, and wire II: AI5 wt% Mg). It is evident that the risk of hot cracking is highest when wire II is used, particularly at high B/(B+D) ratios, since solidification occurs first in the weld metal and then in the partially melted region. In contrast, wire I provides a good HAZ cracking resistance over the whole composition range because of the resulting lower solidus temperature. In the latter case the weld metal solidification and thermal contraction stresses are imposed on the HAZ at a stage where liquid no longer exists at the grain boundaries.
Example (7.8)
Consider plasma arc butt welding of a 10mm thick aluminium plate of AA6082- T6 (containing 0.7 wt% Mg and 0.9 wt% Si) under conditions similar to those employed in Example (7.7). Use the diagrams in Fig. 7.61 to determine which of the two filler wires (lor II) that provides the highest resistance against hot cracking in the partially melted region under the prevailing circumstances.
Solution
In the previous example the base metal dilution ratio, B/(B+ D), was found to be 0.78. Under such conditions wire I provides the highest resistance against hot cracking, since solidification first occurs in the partially melted region and then in the weld metal. This explains why AI-Si based filler wires are usually recommended for single pass butt welding of AI-Mg-Si extrusions if the strength level is not of particular concern.
PROPERTIES OF WELDMENTS
545
(a)
600
0 0
::::J .....,
Q)
1
PI
e
en
::::J
590
580
.....,
a. E Q)
570
:2
en
(5
560
550
20
40
60
8/(8+D), 0/0
(b)
80
100
600
1 o
0
590
cu ~
::s
Q)
580
0..
B :-Q
rn ::::J
570
(5
(J)
560
8/(8+D), %
..,
Fig. 7.61. Variation of weld metal solidus temperature with dilution for 6082 aluminium alloys (AI-0.7 wt% Mg-0.9 wt% Si); (a) Wire I (AI-5 wt% Si), (b) Wire II(AI-5 wt% Mg). The melting temperature for different base metal constituent phases are indicated by the horizontal broken lines in the diagrams. Data from Gittos and Scott" and Reiso et al. 96
546 (a)
Peak temperature
.~
~.
t
------"'--HAZ
(b)
t
I
..,."
I \
. ~Z9.o.d .
:\
I
.~weld ~
i
Time
II~t.
thermal cycle
Fig. 7.62. Schematic diagrams showing the sequence of reactions occurring in the HAZ of 6082- T6 aluminium welds; (a) Hardness distribution following ~"(Mg2Si) dissolution; (b) Precipitation Of~/(Mg2Si) at dispersoids during the weld cooling cycle.
PROPERTIES OF WELDMENTS
547
7.4.3 HAZ microstructure and strength evolution during fusion welding The age-hardenable AI-Mg-Si alloys have been widely studied, to the extent that most of the underlying physical processes are well established. They offer tensile strength values higher than 350 MPa in the artificially aged (T6) condition owing to the presence of very fine, needleshaped ~"(Mg2Si) precipitates along < 100> directions in the aluminium matrix'" Although AI-Mg-Si alloys are readily weldable, they suffer from severe softening in the heat affected zone (HAZ) because of reversion (dissolution) of the ~"(Mg2Si) precipitates during the weld thermal cycle.76,l00-103 This type of mechanical impairment represents a major problem in engineering design, since it reduces the load-bearing capacity of the joint. 104 7.4.3.1 Effects of reheating on weld properties Microstructural changes in the HAZ of AI-Mg-Si alloys have been examined by several investigators,100-103 and the main results are summarised in Fig. 7.62. It is evident from Fig. 7.62(a) that reversion of ~"(Mg2Si) precipitates will occur to an increasing extent in the peak temperature range from 250 to 500oe. This is associated with a continuous decrease in the HAZ hardness until the dissolution process is completed. During cooling of the weld, some solute recombines to form coarse, metastable ~' (Mg2Si) precipitates which do not contribute to strengthening (Fig. 7 .62(b. However, close to the fusion boundary a large fraction of alloying elements will remain in solid solution at the end of the thermal cycle, thereby giving conditions for extensive age-hardening at room temperature over a period of 5 to 7 days (Fig. 7.62(c. In general, enhanced HAZ strength recovery can be achieved by the use of artificial ageing in the temperature range from 150 to 180C,76 but this possibility will not be considered here.
(c)
'V
250C
t
Resulting hardness profile :
Fig. 7.62. Schematic diagrams showing the sequence of reactions occurring in the HAZ of 6082- T6 aluminium welds (continued); (c) Hardness distribution after prolonged room temperature ageing.
548
7.4.3.2 Strengthening mechanisms in Al-Mg-Si alloys Due to the lack of experimental evidence of coherency strains around ~"(Mg2Si) precipitates in AI-Mg-Si alloys,99 it has been suggested that the increased resistance to dislocation motion accompanying the presence of these structures arises from the high energy required to break Mg-Si bonds in the particles as dislocations shear through them. Assuming that this strengthening effect is associated with order hardening, the net precipitation strength increment, ~(jp, can be calculated from the equation originally derived by Kelly and Nicholson: 105
~(j
y. = ---1.. 1=
(7-25)
where ~i is the internal interface (or antiphase) boundary energy, b is the Burgers vector, lis the particle volume fraction, and C3 is a kinetic constant. By introducing the relative particle volume fraction, 1110, we obtain: (7-26) where /0 is the initial volume fraction of ~"(Mg2Si) precipitates in the alloy, and C4 is a new kinetic constant (equal to c310). It is evident from equation (7-26) that ~crp = ~crp (max) = C4 whenf / fo = 1. Hence, this equation can be rewritten as:
(X=
(j (j max
(jmin (j min
-~P-=lllo
~(j
~(j
p(max)
(7-27)
Here
(Jmin
denotes the intrinsic matrix strength after complete particle dissolution, while
<Tmax is the original base metal strength in the artificially aged (T6) condition.
Provided that a linear relationship exists between yield strength and hardness, equation (727) can be rewritten as: 102
(X=
HV-HVmin BV max - BV
=/1/0
min
(7-28)
Equation (7-28) provides a basis for assessing the reaction kinetics through simple hardness measurements. Typical values for (jmax, BVmax, (jmin, and HVmin are given in Table 7.7. 7.4.3.3 Constitutive equations The kinetics of ~"(Mg2Si) dissolution during reheating of 6082- T6 aluminium alloys have been considered in Section 4.4.2 (Chapter 4).
Table 7.7 Properties of some AI-Mg-Si
HV max [VPN] 110 115 137
[MPa] 78 93 83
<;
\}I
0.43 0.43
PROPERTIES OF WELDMENTS
549
Reversion model When the (3"(Mg2Si) particles dissolve, the volume fraction falls from its initial value fm according to equation (4-36). A combination of equations (4-36) and (7-28) leads to the following expression for the dimensionless strength parameter within the partly reverted region of theHAZ: (7-29)
where is the maximum hold time required for complete particle dissolution at a given temperature (defined by equation (4-31) in Chapter 4), and n is a time exponent (~0.5). The variation of n with! I fo is shown in Fig. 4.20. Natural ageing model In general, the fraction of hardening (3"(Mg2Si) precipitates which forms during natural ageing if I fo) depends on the amount of remnant solute present in the matrix material after cooling of the weld. By considering the kinetics of the C-curve for precipitation of essentially nonhardening (3'(Mg2Si) particles at dispersoids during the weld cooling cycle, Myhr and Grongl02 arrived at the following relationship for I fo :
t;
Here <I>is a material constant (defined in Table 7.7), and is the critical time required to precipitate a certain fraction of ~'(X = X) at an arbitrary temperature (T). The variation of with temperature is given by equation (6-55) in Chapter 6. The net precipitation increment following natural ageing (<X2) can therefore be written as:
t;
t;
(7-32) Coupling of models Based on equations (7-29) and (7-32) it is possible to calculate the HAZ strength distribution after welding and subsequent natural ageing when the weld thermal programme is known. Figure 7.63 shows a sketch of the superimposed hardness profiles, as evaluated from these equations. Since the resulting strength level in the partly reverted region depends on the interplay between two competing processes (i.e. dissolution and reprecipitation), it is convenient to define the 'boundary' between the two models on the basis of the intersection point in Fig. 7.63 where <Xl = <X2, i.e.: HV and
= HVmin
+(HVmax-
(7-33) (7-34)
HV = nv.; +(HVmax-
550
'V
250C
t
~ __
:: pomt
~----- HV
:
max
{lnt!'lrSection :
.
........
... !'!~tl!r.B:I.~g~il)g.lJl.qcj~! HV
:
min
Distance from fusion line ----.. Fig. 7.63. Coupling of reversion and natural ageing models.
It follows that this locus also defines the minimum HAZ strength level, which is an important parameter in engineering design. 7.4.3.4 Predictions of HAZ hardness and strength distribution The predictions are based on computer programmes which utilise the medium thick plate heat flow solution described in Section 1.10.4 (Chapter 1) and the kinetic models outlined above to calculate the HAZ hardness distribution for specific welding conditions (details are given in Ref. 106). Accuracy ofpredictions Examples of measured and predicted HAZ hardness profiles are shown in Figs. 7.64 and 7.65. When stringer bead welding is carried out on a plate of medium thickness, the hardness distribution in the transverse y direction will vary with distance from the plate surface due to a continuous change in the heat flow conditions. A comparison between observed and predicted hardness profiles in Fig. 7.64 shows that such effects are readily accounted for in the present model. In contrast, a full penetration butt weld will always reveal a similar HAZ hardness distribution in the transverse section of the weld, as shown in Fig. 7.65. This situation arises from the lack of a temperature gradient in the through-thickness z direction of the plate. Moreover, it is evident from Figs. 7.64 and 7.65 that the final dimensions of the HAZ are strongly influenced by variations in welding parameters and operational conditions. Hence, it is difficult to justify the use of a constant safety factor for the width of the HAZ as recommended in current design rules for welded AI-Mg-Si alloys.104 Aptness of models Based on the kinetic models described in the proceding sections, it is possible to construct two-dimensional (2-D) maps which show characteristic hardness and peak temperature contours in the HAZ of 6082- T6 aluminium weldments. Examples of such diagrams are given in
PROPERTIES
OF WELDMENTS
551
600
(a)
500
1
0
0 ~
400
300 ~ E
~ a ~
200 n, ~ 100
z 10 20 ~Ym,mm--600 100 30 40 0 50
(b)
z 0... >
1
en en
Q)
500
t
0
~ :::J ~
"E
ca :r:
c:
E .m ~co
Q) 0...
y 20
z
0
10
20 Ym,mm~
30
40
0 50
Fig. 7.64. Comparison between measured and predicted HAZ hardness profiles in a stringer bead GMA weld; (a) Upper plate surface, (b) Lower plate surface. The peak temperature distribution is indicated by the broken lines in the graphs. (Operational conditions: qo = 9.1 kW, v = 5.1 mm S-l, d = 15mm). Data from Myhr and Grong.102
Fig. 7.66. Included is also a 3-D plot of the HAZ hardness distribution in the transverse section of the weld. The results in Fig. 7.66 reveal a direct relationship between the HAZ isothermal contours on the one hand and the resulting HAZ hardness/strength distribution on the other. In this particular example the soft zone closely follows the contour of the 400C isotherm. This, in tum, implies that the minimum HAZ strength level is fairly constant and virtually independent of choice of welding parameters (i.e. close to 60 VPN for single pass welds).
552 100
METALLURGICAL
MODELLING
OF WELDING
600
t z
a...
500
80 400 300
1
o
0
>
en
Q)
u)
C
~ ~
:::::J
...... /Tp
Q)
J:
"E cu
".
0.
40
y
20 .z=1 mm * z=6.5 mm o z=12 mm 0 10
..x:
z
20 30 ~Ym,mm-..... 40 0 50
Fig. 7.65. Comparison between measured and predicted HAZ hardness profiles in a single pass plasma arc butt weld. The peak temperature distribution is indicated by the broken line in the graph. (Operational conditions: qo = 14.0 kW, v = 5.8 mm s',d = 13mm). Data from Myhr and Grong.102
In practice, the HAZ hardness can be converted into an equivalent yield or ultimate tensile strength through the following regression formulae: 102 R and Rm (MPa) (MPa) = 3.0HV -48.1 (7-35)
PO.2
= 2.6HV
+ 39.8
(7-36)
From equation (7-35) we see that a minimum HAZ hardness of about 60 VPN corresponds to a strength reduction factor of: ~= 3.0(60)-48.1 =0.47 3.0(110) - 48.1
This value is in good agreement with the recommended strength reduction factor of 0.49 incorporated in many welding specifications and standards.I04,I07 Process diagrams for single pass butt welds Because of the complex temperature-time pattern in aluminium welding, it is not possible to condense general information about the HAZ strength distribution into 2-D process diagrams. An exception is single pass butt welding of plates, where the medium thick plate solution described in Section 1.10.4 (Chapter 1) can largely be simplified if the net arc power is kept sufficiently high compared with the plate thickness (e.g. qo /d > 0.5 kW mrrr"). Under such conditions the mode of heat flow becomes essentially one-dimensional, and the temperature distribution is determined by the ratio qo/ vd, kJ mm ? (see equation (1-100) in Chapter 1).
PROPERTIES OF WELDMENTS
553 y. mm
20
30
10
.......-r-------'--r-----+-O
5 E
E
o
(a)
10 N
(b)
Fig. 7.66. Computed HAZ hardness and peak temperature contours in the transverse section of a stringer bead GMA weld; (a) 2-D graphical representation, (b) 3-D graphical representation. (Operational conditions as in Fig. 7.64). Data from Myhr and Grong.102
Figure 7.67 shows plots of the HAZ hardness/strength profiles for different values of q 0 /vd. It follows that a narrow width of the HAZ requires the use of a low energy input per mm? of the weld. In practice, this can be achieved by the choice of an efficient welding process (e.g. electron beam or laser welding) which allows deposition of a full penetration butt weld without employing a groove preparation (i.e. eliminates the need for filler metals). Multipass welding The present process model can also be extended to multipass welding if it is assumed that reversion of indigenous f3' (Mg2Si) precipitates occurs instantaneously on reheating above the phase boundary solvus temperature (here taken equal to 520C).
554
METALLURGICAL
MODELLING
OF WELDING
RpO.2 [MPa]
Rm [MPa]
100
0.040.06
II I /
0.08 0.12
-:
0.16
0.20
250
300
z n, >
(/)
80
qo/vd [kJ/mm2]
200
150
250
en
"E
(J)
200
:c
ctS
100 150
10mm
1-----1
Scale:
50 100
Fig. 7.67. Process diagram for single pass 6082- T6 butt welds. Data from Myhr and Grong.102
Figure 7.68 shows 2-D and 3-D plots of computed HAZ hardness and peak temperature contours for a simulated two-pass butt weld. This model system consists of two imaginary stringer beads which are placed symmetrically on each side of a l Smm thick plate. Comparable hardness data for a single pass weld of same thickness are contained in Fig. 7.66. It is evident from Figs. 7.66 and 7.68 that deposition of a second pass will neither increase the width nor reduce the strength of the HAZ to any great extent. The explanation lies in the fact that the temperature field around each heat source tends to overlap with increasing distance from the fusion boundary due to symmetry effects (see Fig. 7.68(a)), which prevents excessive strength loss during reheating of the weld. Similar observations have also been made from actual testing of multipass AI-Mg-Si weldments.P"
Example (7.9)
Consider plasma arc butt welding of a 10mm thick aluminium plate of type AA 6082- T6 under conditions similar to those employed in Example (7.7). Estimate on the basis of the process diagram in Fig. 7.67 and the simplified Rykalin thin plate solution (equation (1-100) in Chapter 1) both the minimum HAZ strength level, the total width of the reduced strength zone after welding, as well as the lower temperature limit for dissolution of the (3"(Mg2Si) precipitates during the weld thermal cycle. Thermal data for AI-Mg-Si alloys are given in Table 1.1 (Chapter 1).
Solution
First we need to calculate the net heat input per mm- (q 0/ vd). In the present example, we have:
s; /vd=~kJmm-2
IOxIO
=0.2kJmm-2
PROPERTIES OF WELDMENTS
555
y. mrn
10
20
30 0 5
E E 10 N 15
(a)
(b)
Fig. 7.68. Computed HAZ hardness and peak temperature contours for a simulated two-pass butt weld (the second pass is deposited immediately after cooling of the first pass); (a) 2-D graphical representation, (b) 3-D graphical representation. (Operational conditions: qo = 9.1 kW, v = 5.1 mm s', d = 15 mm). Data from Myhr and Grong.102
PO.2
556
Similarly, by considering the extension of the HAZ and the corresponding scale bar in Fig. 7.67, the total width of the reduced strength zone becomes: ~Ym ~50mm
length of the
The relationship between peak temperature T = Tp and distance Y = Ym from the heat source can now be obtained by differentiating equation (1-100) with respect to time. After some manipulation, we obtain:
~Y
m
s. / vd)
pc~21Ce
[Ill
(Tp-20)
(Tm-20)
= 50
By substituting qo / vd = 200 J mm ", pc = 0.0027 J mm-3. C-l, and Tm = 652C into the above equation, the following temperature for incipient dissolution of the J3"(Mg2Si) precipitates is obtained:
It is obvious from the above calculations that the degree of HAZ softening occurring during welding is substantial under the prevailing circumstances. This explains why, for instance, high heat input deposition is usually not recommended for AI-Mg-Si alloys. 7.4.4 HAZ microstructure and strength evolution during friction welding Friction welding is a solid state joining process that involves both heating and plastic deformation of the parent material under extreme thermal and strain rate conditions.l09-113 During the' welding operation the material is deformed at high temperatures, initially at low strain rates, but due to the axial displacement of the specimen, the material is subsequently brought into the fully plasticised region where the strain rate may exceed 103 S-1(e.g. see Fig. (6.79) in Chapter 6). Although the resulting microstructural changes are similar to those observed in many hot working processes, it is obvious that transient effects must playa more dominant role in friction welding because of the rapid temperature and strain rate fluctuations. This, in tum, increases the complexity of the analysis. 7.4.4.1 Heat generation in friction welding A major problem in modelling of heat flow phenomena in friction welding is to obtain an accurate description of the energy input at the faying interface. This is because the friction coefficient Jl is changing continuously during the welding cycle from Jl > 1 at the dry sliding start, towards zero when the temperature for asperity melting is reached at the interface.U'' However, for the ideal case considered in Fig. 7.69, the torque required to rotate two circular shafts relative to one other under the action of an axial load pressure is given by:1l4,1l5
M(R)
M(Nm)
R*
dM
f
0
JlP(r)2rrr2dr
=~
/L'lTP(R*)3
(7-37)
where M is the interfacial torque, R* is the surface radius, and P(r) is the pressure distribution across the interface (here assumed constant and equal to P).
PROPERTIES
OF WELDMENTS
557
Pressure
Pressure
If all the shearing work at the interface is assumed to be converted into frictional heat, the average heat input per unit area and time becomes: 114 q 1 .zs: (Wmm -2) = A A
fM(R)
0
wdM=
fR*
0
r:
II.PudA
=-
fR*
0
,-
II.PU
max.
R*
2Trrdr=
where qo is the net power (in W), P is the friction pressure (in N mm"), A is the cross section (in mm-), and Umax. is the maximum surface velocity at the outer edge (in m S-I). Equation (7-38) provides a basis for estimating the heat generation at the interface during continuous drive friction welding in the absence of asperity melting. 7.4.4.2 Response of AI-Mg-Si alloys and AI-SiC MMCs to friction welding In friction welding of aluminium alloys and AI-SiC metal matrix composites, the deformation is maintained throughout the steady state period.!':' The microstructural changes that take place at this stage of the process can be classified as dynamic changes, since they occur under the action of variable plastic straining. Similarly, those taking place after the forge operation (when the weld cools) are referred to as static changes due to the lack of external plastic deformation. The former includes the generation of recovered subgrains and reversion of constituent precipitates in the area of high temperature and high deformation, whereas the latter involves precipitation reactions, e.g. formation of hardening J3"(Mg2Si) particles in the heat affected zone following natural ageing. 109-113 In order to assemble an adequate model for the microstructural" evolution during friction welding, it is necessary to divide the HAZ into different reaction zones. Referring to Fig. 7.70, the following three main regions are of specific interest in this context: (i) The fully plasticised region, Zpl., where the material is able to accommodate the plastic strain by dynamic recovery (or recrystallisation) of the microstructure.
558
I I I I
: Contact section
:---I
Z=Zpd.
..1.I
I I I I
Z=Zud.~
z=o
Fig. 7.70. Schematic diagram showing the three main reaction zones within a friction welded component (Zpl.: fully plasticised region, Zpd.: partly deformed region, Zud.: undeformed region).
(ii)
The partly deformed region, Zpd., where the degree of plastic deformation is accommodated by an increase in the dislocation density of the matrix grains. In this region the temperature is sufficiently high to facilitate dissolution of the base metal hardening precipitates. The undeformed region, Zud., characterised by partial reversion of the base metal precipitates.
(iii)
Aspects of HAZ subgrain evolution during continuous drive friction welding have been described in Section 6.5.2 (Chapter 6). In the following, the structural and mechanical response of T6 heat treated AI-Mg-Si alloys and AI-SiC metal matrix composites to the imposed heating and plastic deformation will be considered more in detail. 7.4.4.3 Constitutive equations Since the HAZ strength level in both types of materials is mainly controlled by dissolution and precipitation reactions occurring within the aluminium matrix during the weld thermal cycle,109-113 we can use equations (7-29) and (7-32) with minor modifications to describe the reversion and the natural ageing characteristics of the weld components (see Table 7.7). In addition, it is necessary to allow for the plastic deformation introduced during friction welding, because dislocations will be generated in the matrix material to accommodate the strain. 113 In general, work hardening of metals and alloys is a very complex problem which has not yet been properly solved. However, with the precision aimed at here it is sufficient to assume that this strength contribution is given by: (7-39) where \}J is a constant which is characteristic of the material under consideration (given in Table 7.7). The form of equation (7-39) indicates that the hardness is essentially constant within the plasticised regions of the weld HAZ. 7.4.4.4 Coupling of models Based on equations (7-29), (7-32), and (7-39) it is possible to calculate the HAZ strength distribution after friction welding and subsequent natural ageing when the weld thermal programme is known. Figure 7.71(a) and (b) show schematic representations of the superim-
PROPERTIES OF WELDMENTS
559
posed hardness profiles, as evaluated from these models. The resulting HAZ hardness distribution is indicated by the solid curve in the graphs for short and long duration thermal cycles, respectively. Since the present treatment oversimplifies the problem by only considering the strongest contribution with no interaction, the justification of this assumption relies on a good correlation between theory and experiments. It follows from Fig. 7.71 that particle dissolution is the major softening mechanism in friction welding of AI-Mg-Si alloys and AI-SiC metal matrix composites. At the same time a substantial strength recovery will occur as a result of external plastic straining in combination with intrinsic precipitation of hardening (3"(Mg2Si) particles following prolonged room temperature ageing. Depending on the operational conditions applied, this may give rise to differences in the shape of the HAZ hardness profiles, as shown in Fig. 7.71. (a)
Reversion model
t
: -, r,
. .. - - ~ .... . . ,: - - - - - - - Work hardening model
"E
~
o Q) c:
base material
Z~
(b)
el
t
(fl (fl
c: -e J:
CD
Q)
- - - - - Work hardening
model
- - ~- - - - - ~
, ~ .,
, ,..
" .. .....
. . .
Fig. 7.71. Schematic representation of the HAZ hardness distribution after friction welding and subsequent natural ageing; (a) Short duration thermal cycle, (b) Long duration thermal cycle. The parameters Zpl., Zpd. and Zud. are defined in Fig. 7.70.
560
7.4.4.5 Prediction of the HAZ hardness distribution The predictions are based on computer programmes which utilise the heat and material flow models described in Ref. 113 in combination with the constitutive equations given above to calculate the HAZ hardness distribution for specific welding conditions. Accuracy of predictions Examples of measured and predicted hardness profiles are given in Figs. 7.72 and 7.73. A closer inspection of the graphs reveals a good agreement between theory and experiments in all three cases. It is interesting to note that there is no clear distinction in the shape of the HAZ hardness profiles between friction welded AI-Mg-Si alloys and AI-SiC metal matrix composites when comparison is made on the basis of a similar temperature-time pattern (see Fig. 7.73). However, large diameter weld components will normally reveal a different hardness distribution, as shown in Fig. 7.72, because longer welding times will increase the total heat input. Under such conditions the contribution from the plastic deformation becomes negligible, which means that the resulting HAZ hardness profile will closely resemble that observed during conventional gas metal arc (OMA) and plasma arc welding of AI-Mg-Si alloys (see Figs. 7.64 and 7.65). Process diagrams Based on the above process model, it is possible to construct a series of diagrams which summarise information about the effect of important welding variables in a systematic and illustrative manner. Examples of such diagrams for 6082- T6 aluminium alloys and T6 heat treated AI-SiC metal matrix composites are given in Fig. 7.74(a) and (b), respectively.
120
z a.. >
C/}
Q)
100
u)
"E as
c:
:c
12
16
.
Axial distance, mm
Fig. 7.72. Comparison between measured and predicted HAZ hardness profiles in a <l>26mmAI-Mg-Si weld component. (Assumed input data: <I> = 'P = 0.56, HVmax = 110, HVmin = 42). Operational conditions: qiA = 17W mrrr? and ts = 6s. Data from Midling and Grong.U''
PROPERTIES OF WELDMENTS
561
(a)
140
z a.
120
>
CI) Q)
100
u)
co J:
c -e
80 -
60
Unaffected base material
40
Z~1. Zpd.
10
12
14
Axial distance, mm ~
(b)
140
z a. >
CI)CI) Q)
120
100 .
AI-SiC- T6 60 . Predicted Measured
m J:
"E
80 .
L:pl.
40 0
Axial distance, mm
..
12
14
Fig. 7. 73. Comparison between measured and predicted HAZ hardness profiles; (a) <I>16mm Al-Mg-Si weld component. (Operational conditions: qJA = 25W mm ? and ts= 0.9s), (b) <I>16mm friction welded AI-SiC metal matrix composite. (Operational conditions: qJA = 25W mm ? and ts = 3.8s). Data from Midling and Grong.l13
562 (a)
140
AA 6082 - T6
z
(f)
g;
(])
!
en
c
120
100
,
I
,~
, ,,
..
'
80
60
12
16
20 -
24
28
Axial distance, mm
g;
~
!
en en
(])
100
80
60
AI-SiC-T6
i /A r---qo
1--qo/A
= 41.7 W/mm
,ts= 1.0 s
I ts= 3.8 sl
5
S
ts= 8.0
20
28
Axial distance, mm
Fig. 7.74. Process diagrams for friction welding; (a) 6082-T6 aluminium alloys. (Operational conditions: Umax. = 2.5m S-1 and Jl = 0.5), (b) AI-SiC- T6 metal matrix composites. (Operational conditions: Umax. = 2.5m S-1 and Jl = 0.5). Data from Midling and Grong.U''
PROPERTIES OF WELDMENTS
563
It is evident from these diagrams that the HAZ hardness distribution depends on the total heat input applied during friction welding. Although the controlling parameters qo / A and ts (welding time), in practice, are kept within relatively narrow limits, it is obvious that a small width of the HAZ requires the use of a high specific power (q 0/ A) in combination with a short duration heating cycle (ts < 2 s). This is also in agreement with general experience.l'P"!'?
Example (7.10)
Consider continuous drive friction welding of a T6 heat treated AI-SiC metal matrix composite under the following conditions: P
= 30 MPa,
Umax.
= 2.5 m s
', ts
= 3.8 s
Use the process diagram in Fig. 7.7 4(b) to estimate the minimum HAZ hardness level as well as the total width of the strength reduced zone after welding. In this example we shall assume that the friction coefficient Jl is equal to 0.5.
Solution
First we need to calculate the frictional heat per unit area of the weld. From equation (7-38), we have: qo / A
= tJlPumax. = to.5
x 30 x 2.5Wmm-2
= 25 W mm?
A comparison with Fig. 7.7 4(b) shows that a specific power of 25 W mnr? corresponds to a minimum HAZ hardness of about 90 VPN, i.e. a reduction of 45 VPN compared with the base material. At the same time the total width of the reduced strength zone is seen to be 12mm. It should be emphasised that the observed strength loss is not permanent, since the resulting HAZ strength level is mainly controlled by dissolution reactions taking place within the aluminium matrix during the weld thermal cycle. Consequently, a full HAZ strength recovery can be achieved by the use of an appropriate post weld heat treatment, as shown by the tensile test data in Table 7.8.
Table 7.8 Mechanical properties of friction welded AI-SiC metal matrix composites. Data from Midling and Grong109,113
HV [VPN] 135 RpO.2 [MPa] 315 Rm [MPa] 352 eB [%] 3.6
Material Al-SiC-T6 (base material) Al-SiC-T6 (as-welded condition) Al-SiC-T6 (PWHT conditionr]
90
207
268
3.0
135
313
348
3.1
564
References
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. R.W.K. Honeycombe: Steels - Microstructure and Properties, 1980, London, Edward Arnold (Publishers) Ltd. H. Suzuki: Weld. World, 1982,20, 121-148. 0. Grong and D.K. Matlock: Int. Met. Rev., 1986,31,27-48. D.J. Abson and R.J. Pargeter: Int. Met. Rev., 1986,31, 141-194. P.L. Harrison and R.A. Ferrar: Int. Met. Rev., 1989,34,35-51. G.J. Davis and J.G. Garland: Int. Met. Rev., 1975,20,83-106. S.A. David and J.M. Vitek: Int. Mater. Rev., 1989,34,213-245. T. Gladman and F.B. Pickering: In Yield, Flow and Fracture of Polycrystals (Ed. T.N. Baker), 1983, London, Applied Science Publishers, 141-198. A.O. Kluken, M.1. Onsoien, O.M. Akselsen and G. Rervik: Joining Sci., 1991,1,14-22. P. Deb, K.D. Challenger and A.E. Therrien: Me tall. Trans., 1987, 18A, 987-999. P.T. Odland, C.W. Ramsay, D.K. Matlock and D.L. Olson: Weld J., 1989,68, 158s-168s. D.J. Widgery: Weld J., 1975, 54, 57s-68s. L. Devillers, D. Kaplan, B. Marandet, A. Ribes and P.V. Riboud: Proc. Int. Con! on Effects of Residual, Impurity and Microalloying Elements on Weldability and Weld Properties, London, Nov. 1983, Paper 1, Publ. The Welding Institute. R.A. Farrar: Welding and Metal Fabr., 1976,44, 578-58l. 0. Grong and A.O. Kluken: In Ferrous Alloy Weldments, (Eds D.L. Olson and T.H. North), 1992, Zurich (Switzerland), Trans. Tech. Publications, 21-46. R.H. Van Stone, T.B. Cox, J.R. Low and J.A. Psioda: Int. Met. Rev., 1985,30, 157-179. A.W. Thompson: Acta Me tall. , 1983,31,1517-1523. I. Andersen: MSc Thesis, 1989, Division of Metallurgy, The Norwegian Institute of Technology, Trondheim, Norway. G.E. Dieter: Mechanical Metallurgy, 3rd edn, 1986, New York, McGraw-Hill Book Company. O.M. Akselsen and 0. Grong: Mater. Sci. Eng., 1992, A159, 187-192. D.E. McRobie and J.F. Knott: Mater. Sci. Technol., 1985,1,357-365. 0. Grong, A.O. Kluken and B. Bjernbakk: Joining and Materials, 1988, 1,164-169. A.O. Kluken and 0. Grong: Proc. 3rd Int. Conf. on Trends in Welding Research, Gatlinburg, TN, June, 1992, pp. 569-574. Publ. ASM International, Materials Park, Ohio (1993). J.1. Ustinovshchikov: Acta Me tall., 1983,31, 355-364. A.J. Ardell: Acta Me tall. , 1972,20,601-609. O.M. Akselsen, G. Rervik, M.1. Onseien and 0. Grong: Weld J., 1989,68, 356s-362s. H. Suzuki: Trans. Jap. Weld. Soc., 1981, 16, 25-32. N. Yurioka, S. Ohshita and H. Tamehiro: Proc. Int. Symposium on Pipeline Welding in the '80s, March, 1981, pp. 1-15, Publ. Australian Welding Research Association. C.L.M. Cottrell; Metal Constr., 1984,16,740-744. H.J.D. Cotton: Metal Constr., 1987,19, 217R-223R. T. Kasuya and N. Yurioka: Weld J., 1993,72, 263s-268s. A.O. Kluken, S. Ibarra, S. Liu and D.L. Olson: Proc. 11th Int. Con! on Offshore Mechanics and Arctic Engineering, 1992, Publ. ASME, Book No. H0744A-92. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1990,38,449-460. O.M. Akselsen and G. Rervik: Mater. Sci. Technol., 1990,6,383-389. J.H.Hollomon and L.D. Jaffe: Trans. AIME, 1945, 162, 223-249. K. Olsen, D.L. Olson and N. Christensen: Scand. J. Metall., 1982,11,163-168. P.J. Alberry: Weld J., 1989,68, 410s-417s. P. Ravi Vishnu and K.E. Easterling: In Mathematical Modelling of Weld Phenomena (Eds H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, pp. 241-299.
14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38.
PROPERTIES OF WELDMENTS
565
39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55.
65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81.
O.D. Sherby and J.E. Dorn: Trans. AIME, 1953, 197, 324-330. T. Reti, M. Gergely and P. Tardy: Mater. Sci. Techno!., 1987,3,365-371. 0. Grong and O.M. Akselsen: Metal Constr., 1986,18,557-562. C. Thaulow, AJ. Paauw, A. Gunleiksrud and OJ. Nress: Metal Constr., 1985,17, 94R-99R. O.M. Akselsen, 0. Grong and G. Rervik: Scand. J. Metall., 1990,19,258-264. C. Thaulow, A.J. Paauw and K. Guttormsen: Weld J., 1987,66, 266s-279s. O.M. Akselsen, J.K. Solberg and 0. Grong: Scand. J. Metall., 1988,17,194-200. O.M. Akselsen, 0. Grong and P.E. Kvaale: Me tall. Trans., 1986, 17A, 1529-1536. H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld J., 1987,66, 301s-309s. Y-T. Pan and J-L. Lee: Proc. 3rd. Int. Conf. on Trends in Welding Research, Gatlinburg (Tennessee), June, 1992, pp. 539-543. Publ. ASM International, Materials Park, Ohio (1993). J .H. Chen, Y. Kikuta, T. Araki, M. Yoneda and Y. Matsuda: Acta M etall., 1984, 32, 1779-1788. C.A.N. Lanzillotto and EB. Pickering: Met. Sci., 1982, 16, 371-382. O.M. Akselsen, 0. Grong and J.K. Solberg: Mater. Sci. Techno!., 1987,3,649-655. H.P. Shen, T.C. Lei and J.Z. Liu: Mater. Sci. Techno!., 1986, 2, 28-33. NJ. Kim and G. Thomas: Me ta ll. Trans., 1981, 12A, 483-489. A.F. Szewezyk and J. Garland: Me tall. Trans., 1982, 13A, 1821-1826. M. Ramberg, O.M. Akselsen and 0. Grong: Proc. 1st Int. Conf. on Trends in Welding Research, Gatlinburg, TN, May, 1986, pp. 679-685. Publ. ASM International, Metals Park, Ohio (1987). N.E. Hannertz: Schweissen & Schneiden, 1976,28,379-382. F.R. Coe: Welding Steels Without Hydrogen Cracking, 1973, Abington (Cambridge), The Welding Institute. N. Yurioka and H. Suzuki: Int. Mater. Rev., 1990,35,217-249. I.T. Brown and W.M. Baldwin: Trans. A/ME, 1954,200,298-303. E.A. Steigerwald, EW.'Shaller andA.R. Troiano: Trans. AIME, 1959,215,1048-1052. C.A. Zapffe and C.E. Sims: Trans. AIME, 1941, 145, 225-237. N.J. Petch: Phil. Mag., 1956, 1, 331-337. C.D. Beachem: Metall. Trans., 1972,3,437-451. A.W. Thompson and I.M. Bernstein: In Effects of Hydrogen on the Behaviour of Metals (Eds I.M. Bernstein and A.W. Thompson), 1980, Warrendale, PA, Metallurgical Society of AIME, pp.291-308. J.D. Fast: Gases in Metals, 1976, London, Macmillan Press Ltd. R.A.Oriani: Acta Metall., 1970,18, 147-157. N. Christensen, K. Gjermundsen and R. Rose: Brit. Weld J., 1958,5,272-281. N. Christensen: Svetsen, 1975,34,22-32. N. Yurioka and S. Ohshita: IIW Doc. IX-1161-80 (1980). C. Zhang and J.A. Goldak: IIW Doc. IX-1662-92 (1992). N. Christensen and T. Simonsen: Scand. J. Metall., 1981, 10, 120-126. NACE Standard MR-01-75 (1975). M.1. Onseien, O.M. Akselsen, 0. Grong and P.E. Kvaale: Weld J., 1990,69 (No.1), 45-51. H. Suzuki: Trans. Jap. Weld. Soc., 1981,16,25-32. D.L. Olson: Weld J., 1985,64, 281s-295s. S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons, Inc. R.D. Campbell: In Ferrous Alloy Weldments (Eds D.L. Olson and T.H. North), 1992, Zurich, Trans. Tech. Publications Ltd., pp. 167-216. I. Varol, J.C. Lippold and W.A. Baeslack: ibid., pp. 217-252. R. Mundt and H. Hoffmeister: Arch. Eisenhtittenwes., 1983,54,253-256; ibid. 333-336. S. Atamert and J.E. King: Mater. Sci. Technol., 1992,8,896-911. M. Videm: MSc. Thesis, 1985, Department of Metallurgy, The Norwegian Institute of Technology, Trondheim, Norway.
566
82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93.
94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109.
110.
J.A. Brooks, A.W. Thompson and J.C. Williams: Weld t.. 1984,63, 71s-83s. A.L. Schaeffler: Metal Prog., 1949,56, 680-680B. W.T. Delong, G. Ostrom and E. Szumachowski: Weld L, 1956,35, 526s-533s. C.J. Long and W.T. Delong: Weld 1973,52, 281s-297s. J.H. Dudas and F.R. Collins: Weld 1966,45, 241s-249s. C.E. Cross: Ph.D Thesis, 1986, Colorado School of Mines, Golden, Colorado, USA. J.1. Pepe and W.E Savage: Weld i.. 1967,46, 411s-422s. J.1. Pepe and W.F. Savage: Weld r; 1970,49, 545s-553s. O. Reiso: Proc. 3rd. Int. Con! on Aluminium Extrusion Technology, Atlanta, GA, 1984, vol. 1, pp.31-40. Publ. Aluminium Association (1984). O. Reiso: Proc. 4th Int. Aluminium Extrusion Technology Seminar, Chicago, IL, 1988, vol. 2, pp.287-295. Pub I. Aluminium Association (1988). O. Reiso, H.G. 0verlie and N. Ryum: Me ta ll. Trans., 1990, 21A, 1689-1695. H. Gjestland, A.L. Dons, O. Lohne and O. Reiso: In Aluminium Alloys - Their Physical and Mechanical Properties, 1986, Warley (UK), Engineering Materials Advisory Service Ltd., pp. 359-370. O. Lohne and N. Ryum: Proc. 4th Int. Aluminium Extrusion Technology Seminar, Chicago, IL, 1988, vol. 2, pp. 303-308. Publ. Aluminium Association (1988). Metals Handbook, 8th Edition (vol. 8). O. Reiso, N. Ryum and J. Strid: Me ta ll. Trans. A, 1993, 24A, 2629-2641. H.W.L. Phillips: Annotated Equilibrium Diagrams of Some Aluminium Alloy Systems, 1959, London, The Institute of Metals, pp. 65-71. N.F. Gittos and M.H. Scott: Weld 1981,60, 95s-103s. lE. Hatch (Ed.): Aluminium - Properties and Physical Metallurgy, 1984, Ohio (USA), American Society for Metals. T. Enjo and T. Kuroda: Trans. fWRI, 1982,11,61-66. S.D. Dumolt: Ph.D Thesis, 1983, Carnegie-Mellon University, USA. O.R. Myhr and 0. Grong: Acta Metall. Mater., 1991,39,2693-2702; ibid., 2703-2708. 0. Grong and O.R. Myhr: In Mathematical Modelling of Weld Phenomena, (Eds H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, pp. 300-311. F.M. Muzzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publishing Inc. A. Kelly and R.B. Nicholson: Progr. Mat. Sci., 1963,10,151-156. O.R. Myhr, Ph.D Thesis, 1990, Division of Metallurgy, The Norwegian Institute of Technology, Trondheim, Norway. European Recommendations for Aluminium Alloy Structures, 1978. G. Steidl and R. Mossinger: Aluminium, 1977,53, 199-203. O.T. Midling, 0. Grong and M. Camping: Proc. 12th Risii Int. Symp. on Materials Science: Metal Matrix Composites-Processing, Microstructure and Properties, Roskilde, Denmark, 1991, pp. 529-534. Publ. Riso National Laboratory (1991). O.T. Midling, 0. Grong and D.H. Bratland: Proc. 3rd Int. Con! on Aluminium Alloys - Their Physical and Mechanical Properties, Trondheim, Norway, 1992, pp. 99-105. Publ. The Norwegian Institute of Technology, Department of Metallurgy (1992). O.T. Midling and 0. Grong: Proc. 3rd Int. Con! on Trends in Welding Research, June, 1992, Gatlinburg, TN, pp. 1147-1151. Publ. ASM International (1993). O.T. Midling and 0. Grong: Proc. Int. Con! Advanced Composites '93, Wollongon, Australia, February 1993, pp. 1221-1226. Pub I. The Minerals, Metals & Materials Society (1993). O.T. Midling and 0. Grong: Acta Metall. Mater., 1994,42,1595-1609; ibid., 1611-1622. N.N. Rykalin, A.I. Pugin, V.A. Vasil'eva: Weld. Prod., 1959,6,42-52. B. Crossland: Cont. Phys., 1971, 12,559-574.
s.. s..
i:
PROPERTIES OF WELDMENTS
567
lattice parameter (nm) cross section (mrrr', m2) start temperature of ferrite to austenite transformation (OC, K) end temperature of ferrite to austenite transformation (OC, K) Burgers vector (nm, m) body-centred cubic body-centred tetragonal amount of fused parent metal (mm-) kinetic constant in Hollomon-Jaffe rameter base metal half crack length (urn, m) various kinetic and empirical constants pa-
dp dv D VX
particle diameter (urn, m) inclusion diameter (urn, m) amount of deposited metal (mm/) diffusivity (rnm? S-I, m2 S-I) average grain size (urn, m) lattice diffusion coefficient (mmm2 s-l) elongation (%) gross heat input (kJ mm+) binding energy between hydrogen and trap site (J mol:", kJ mol:') Young's modulus of matrix (MPa, GN m-2) Young's modulus of particle (MPa, GN m-2) Gaussian error function particle volume fraction volume fraction of hardening precipitates in natural ageing model initial particle volume fraction eutectic fraction flux -cored arc welding face-centred cubic filler wire grain coarsened region gas metal arc welding grain refined region gas tungsten arc welding
ACI
AC3
D DL
b bcc bet B B*
s',
eB E EB
BM
C
Em
e,
erf(u)
Cl"'C4
(ppm, ml per
f
(wt%) (ppm,
c;
c,
f
!o
leut.
initial element concentration ml per 100 g, III cm-3) maximum solid solubility (wt%)
c.:
Co
base metal or weld metal element concentration (wt%, ppm, ml per 100 g, III crrr ') Yurioka carbon equivalent IIW carbon equivalent crack tip opening displacement Charpy V-notch plate thickness (mm)
568
HFM
hydrogen content related to fused metal (ml per 100 g or g per ton) Rockwell C hardness Vickers hardness (kg mm ? or VPN) Vickers hardness in the artificially aged condition (kg mm ? or VPN) matrix Vickers hardness in the absence of hardening precipitates (kg mm ? or VPN) kinetic strength of thermal cycle with respect to ~'(Mg2Si) formation intercritical region locking parameter in Hall-Petch equation (MPa J.1m1l2) equilibrium partitioning coefficient density of trap site half length of fusion zone in hydrogen diffusion model (mm, m) longitudinal direction (notch perpendicular to plate rolling direction) strain hardening exponent in Ludwik equation interfacial torque (N m) martensite-austenite constituent
PWHT
post weld heat treatment net power (W) apparent activation energy for softening reaction (1 mol ", kJ mol'") radius vector (mm, m) isothermal zone width referred to fusion boundary (mm) universal gas constant (8.314 1 mol '
HRC HV
qo
Qapp.
v.:
HVmin
r L1r:
K:')
11
R* RIR Rm
friction surface radius (mm, m) implant rupture strength (MPa) ultimate tensile strength (MPa) 0.2% proof stress (MPa) scanning electron microscope shielded metal arc welding subcritical region time (s)
IR k
RpO.2
SEM SMAW SR
ko
K
L
L-T
* t1 * t2
maximum hold time required for complete particle dissolution at T (s) critical hold time required to precipitate a certain amount of rl'(Mg2Si) at T (s) welding time (s) cooling time from 600 to 400C (s) cooling time from 800 to 5000e (s) cooling time from 1200 to 800C (s) equivalent 333K (s) isothermal hold time at
M M-A MMC n
n3
ts
time exponent
At8/5
dimensionless operating parameter in heat flow model friction pressure (N mm ? , MPa) Hollomon-Jaffe parameter
L1t12l8 At333
p PI
P2
P(r)
t;
Teut.
PROPERTIES OF WELDMENTS
569
strength (MPa) yield stress (MPa) intergranular fracture strength (MPa) matrix fracture strength (MPa)
melting point (Oe, K) ambient temperature (OC, K) peak temperature (OC, K) T6 TEM artificially aged condition transmission electron microscope transverse direction (notch parallel with plate rolling direction) velocity (m
S-I)
0"
O"i
O"max
friction stress (MPa) yield strength in artificially aged condition (MPa) matrix yield strength in the absence of hardening precipitates (MPa) critical (MPa) stress for particle cracking
T-L
U
O"min
Umax.
surface velocity (m
S-I)
Ur UTS v
area under stress-strain curve (J rrr ') ultimate tensile strength welding speed (mm
s')
or
radial stress (MPa) threshold stress for H2S stress corrosion cracking (MPa) tangential stress (MPa)
Vickers pyramid number welding direction (mm, m) transverse direction (mm, m) y-coordinate at maximum width of isotherm (mm) isothermal zone width (mm, m) through-thickness direction (mm, m)
'tmax.
'Yeff.
effective surface energy (J m-2) internal interface (or antiphase) boundary energy (J m-2) surface energy of particle (J m-2) friction coefficient maximum shear stress (MPa) dimensionless to 500C cooling time from 800
L\Ym
partly deformed region in friction welding model (mm) fully plasticised region in friction welding model (mm)
Zud.
8800
dimensionless temperature conforming to 800C dimensionless temperature conforming to 500C arc efficiency factor dimensionless supersaturation
undeformed region in friction welding model (mm) stress concentration factor true fracture strain dimensionless plate thickness
<1>, 'I"
pc
0c-1)
570
dimensionless
strength parameter in
strength increment
dimensionless strength parameter reversion model dimensionless strength parameter natural ageing model dimensionless strength parameter work hardening model
!lap (max) maximum precipitation strength increment (MPa) in HAZ strength reduction factor in
w
8.1 Introduction
This chapter contains a collection of different exercise problems which the author has adopted in his welding metallurgy course for graduate (mature) students. They illustrate how the models described in the previous chapters can be used to solve practical problems of more interdisciplinary nature. Each of them contains a 'problem description' and some background information on materials and welding conditions. The exercises are designed to illuminate the microstructural connections throughout the weld thermal cycle and show how the properties achieved depend on the operating conditions applied. Solutions to the problems are also presented. These are not complete or exhaustive, but are just meant as an aid to the reader to develop the ideas further.
The materials to be welded are a C-Mn steel and a Nb-microalloyed low carbon steel with chemical compositions and properties as listed in Tables 8.1 and 8.2. Details of welding parameters and operational conditions are given in Table 8.3 and 8.4, respectively. Table 8.1 Exercise problem I: Base plate chemical compositions (in wt%).
Steel C-Mnt LC-Nbt C
0.20 0.08
Si
0.35 0.26
Mn
1.46 1.44
P
0.003 0.003
S
0.002 0.003
Nb
-
AI
0.037 0.025
0.020
rn. -0.008,
Rm (MPa)
525 525
El. (0/0)
33 32
CVN
-40
(J)
150 225
572
METALLURGICAL
MODELLING
OF WELDING
U(V)
21
v (mm
4
S-1)
"The arc efficiency factor may be taken equal to 0.85 (see Table 1.3). No preheating is applied (To
= 20C).
t.
15 NI per min
1.0mm
6.0 m per min
C: 0.1 wt%, Si: 1.0 wt%, Mn: 1.7 wt% C: 0.09 wt%, Si: 0.7 wt%, Mn: 1.2 wt% RpO.2: 460 MPa, Rm: 560 MPa, El.: 260/0, CVN
50 J
-40:
tData compiled from dedicated filler wire catalogues and welding manuals.
:j:
EXERCISE
PROBLEMS
WITH SOLUTIONS
573
\ \ I
\ I
I I
\ \ I
1\ I I
"
\ \
Analysis: The students should work in groups (3 to 4 persons) where each group select a specific combination of base material and welding conditions (e.g. deposition of a cap layer on the top of a thick multipass C-Mn steel weld). The problem here is to evaluate the response of the base material to heat released by the welding arc. The analysis should be quantitative in nature and based on sound physical principles. The following points shall be considered: (a) Select an appropriate heat flow model for the system under consideration. (b) Estimate the minimum bead length which is required to achieve pseudo-steady state (i.e. a temperature field that does not vary with position when observed from a point located in the heat source). (c) Estimate the value of the deposition coefficient k' (in gA -1 s -1), the weld cross section areas D and B (in mm-), and the mixing ratio D/(B + D) during welding. (d) Estimate the weld metal chemical composition. Calculate then the following quantities: - Total loss of Si and Mn in the arc column - Total oxygen pick-up in the weld pool - Residual oxygen level and total amount of oxygen rejected from the weld pool during deoxidation - Total amount of slag formed during welding (in g per 100 gram weld metal) (e) Carry out a total oxygen balance for the system, and estimate the resulting CO content in the welding exhaust gas. (f) Estimate the chemical composition, volume fraction, and mean size (diameter) of the
574
METALLURGICAL
MODELLING
OF WELDING
oxide inclusions which form in the cold part of the weld pool. Calculate then the following inclusion characteristics: Number of particles per unit volume Number of particles per unit area Total surface area of particles per unit volume Mean particle centre to centre volume spacing
(g) Estimate the weld metal solidification mode and the resulting columnar grain morphology. Indicate also the type of substructure which form at different positions from the weld centre line. (h) Evaluate the thermal stability of the base metal grain boundary pinning precipitates. At which temperature will these precipitates dissolve? (i) Calculate the austenite grain size profile across the HAZ. Estimate also the size of the columnar austenite grains in the weld metal. the austenite to ferrite transformation. (k) Estimate the maximum hardness in the HAZ after welding. Use this information to evaluate the risk of hydrogen cracking and H2S stress corrosion cracking during service. (1) Estimate the CVN toughness both in the weld metal and the HAZ after welding. (m) Based on the results obtained explain why the carbon content of modern structural steels has been gradually lowered to values below 0.1 wt% in step with the progress in steel manufacturing technology. Solution: In all cases we can use stringer bead deposition on thick plates as a model system. It follows from the analysis in Section 1.10.7 (Chapter 1) that the pertinent difference in the effective heat diffusion area between a bead-on-plate weld and a groove weld may conveniently be accounted for by introducing a correction factor f, which depends on the geometry of the groove (see Fig. 1.68). Thus, in the general case the net (effective) power of the heat source can be written as:
G) Estimate the primary reaction products which form in the weld metal and the HAZ after
f = 1, but
In the following, we shall only consider deposition of a cap layer on a thick plate where the analysis can readily be applied to other combinations of steels and welding conditions as well (e.g. f < 1). In the former case, we get: qo
Table 1.1 (Chapter 1) contains relevant input data for the steel thermal properties. (a) The problem of interest is whether we must use the general (but complex) Rosenthal
EXERCISE
PROBLEMS
WITH SOLUTIONS
575
plate solution (equation (1-45 or can adopt the simplified solution for a fast moving power source (equation (1-73. Fig 1.24 provides a basis for such an evaluation. The critical position will be the fusion line. If we neglect the latent heat of melting, the 0ln3 at the melting point becomes: Op n3
==
==
0.22
Readings from Fig. 1.24 suggest that the error introduced by neglecting the contribution from heat flow in the welding direction is sufficiently small that it can be disregarded in the calculations of the HAZ thermal programme. This means that equation (1-73) can be used in replacement of equation (1-45) if that is desirable. (b) The duration of the transient heating period depends on the actual point of observation (i.e. the distance from the heat source). If we, as an illustration of principles, would like to apply the pseudo-steady state solution down to a peak temperature of, say, 700C, the corresponding n/Op ratio at that temperature becomes: n3 Op
==
From Fig. 1.21 we see that this ratio corresponds to a dimensionless radius vector o-.'m of about 5. The duration of the transient heating period may now be read from Fig. 1.18. A crude extrapolation gives: T~10 from which
L ==
2
2a
T -
==
2X 5 10 X --mm 4
~ 25 mm
The minimum bead length is thus 25 mm, which is surprisingly short. (c) The value of the deposition coefficient may be estimated from the data in Table 8.4.
k' ==
7T')
X 12 X ~
(6000)
S-1 ==
0.005 gA -1
S-1
This value corresponds to a k 'Ip ratio of about 0.65 mm 3A -1 S -1, which is in excellent agreement with the data quoted in Table 1.7. The area D of deposited metal thus becomes (see equation (1-120: D -
0.005
4 X 7.8 X 10-3
X
150 mm?
==
24 mm?
The corresponding area of fused parent metal is most conveniently read from Fig. 1.21. Taking the n/Op ratio at the melting point equal to (1/0.22) ~ 4.5, we obtain:
METALLURGICAL
MODELLING
OF WELDING
(4
X
42
52)
mm-
25 mm?
D (B+D) (24
24
+ 25)
0.49
This value is somewhat lower than the expected mixing ratio, which for low heat input welding is close to 0.67. (d) The composition data in Table 8.4 refer to all weld metal deposit. Since the dilution with respect to the base material in this case is small, the weld metal composition would be expected to be close to that given in Table 8.4. An estimate of the total burn-off of alloying elements during welding can be obtained by considering the difference in chemical composition between the filler wire and the weld metal. In the present case we get: [a %C] [a %Si]
= (0.1-0.09) = 0.01
=
(1.0-0.7)
0.3
[a %Mn] = (1.7-1.2) = 0.5 Loss of silicon As shown in Section 2.10.1.3 (Chapter 2), the silicon loss can partly be ascribed to SiO(g) formation in the arc column (with consequent fume formation), and partly to reactions with oxygen in the weld pool during the deoxidation stage (with consequent silicate slag formation). The former loss can be estimated from the fume formation data presented in Table 2.6. Taking the fume formation rate (FFF) of silicon equal to 63 mg min -1, the total loss of silicon in the arc column amounts to: 63 X 100 (0.5)2 X 6000
[a %Si]SiO = 7r
x 7.8 = 0.17
Loss of manganese As shown in Section 2.10.1.4 (Chapter 2), manganese is partly lost iri the arc column due evaporation and partly in the weld pool due to deoxidation reactions. Taking the fume formation rate of manganese equal to 14 mg min-1 (from Table 2.6), the total loss of Mn in the arc column amounts to:
577
[a%Mn]vap
= 1TX
7.8
0.04
The corresponding oxidation loss of manganese in the weld pool is thus: [a %Mn
]MnO
Oxygen pick-up in the weld pool When the oxidation losses of silicon and manganese in the weld pool are known, it is possible to calculate the total oxygen pick-up in the hot spot of the pool immediately beneath the root of the arc, according to the procedure outlined in Section 2.10.1.5 (Chapter 2). However, first we need to estimate the residual weld metal oxygen content on the basis of the thermodynamic model presented in Fig. 2.56. In the present example, the numerical value of the deoxidation parameter is: ([0/0Si][0/0Mn])-0.25
=
([0.7][1.2])-0.25 = 1.05
Reading from Fig. 2.56 gives a residual oxygen content of about 0.07 wt%. The total oxygen pick-up in the weld pool is thus: 32 16 [%0] tot. = 0.07 + 28.1 X 0.13 + 54.9 X 0.46 = 0.35 Rejected oxygen from the weld pool The amount of rejected oxygen is equal to the difference between the total and the residual oxygen level: [%O],ej. = 0.35 - 0.07 = 0.28 From this we see that most of the oxygen which is picked up at elevated temperatures is rejected again during cooling in the weld pool due to deoxidation reactions and subsequent phase separation. Manganese silicate slag formation The weld pool deoxidation reactions give rise to the formation of a top bead slag, as shown in Section 2.10.1.5 (Chapter 2). In the present example the amount of slag per 100g weld metal is equal to: m
slag
= ( 28.1
60.1
X
0.13
+ 54.9
70.9
) 0.46 g = 0.87g
A comparison with Fig. 2.35 shows that the calculated weight of slag is in reasonable agreement with experimental observations. (e) The oxygen balance is carried out in accordance with the procedure outlined in Section 2.10.1.7 (Chapter 2). First we need to estimate the total mass of weld metal produced per unit time:
578 mw == 60vp(B
METALLURGICAL
MODELLING
OF WELDING
+ D) == 60 x 4 X 7.8 X 10-3
(25
0.01 100 X 12
92 mol CO min
-1
Oxidation of silicon: 2
X
92 mol CO min
-1
92 mol CO min
-1
== 8.4
92 mol CO min
-1
The total CO evolution is equal to the sum of these four contributions: (1.5
The resulting CO content in the welding exhaust gas is thus: 33.5 vol% CO
=( X
15
A comparison with the experimental data in Table 2.2 shows that the calculated CO content is of the expected order of magnitude. (f) The deoxidation model in Section 2.12.4.1 (Chapter 2) can be used to estimate the inclusion composition. From Fig. 2.68 we see that the inclusions are essentially pure manganese silicates with an overall composition close to MnSi03. When the inclusion composition is known, it is possible to convert the residual weld metal oxygen content into an equivalent inclusion volume fraction according to the procedure outlined in Section 2.12.l.Taking the stoichiometric conversion factor equal to 5.0 X 10-2 for manganese silicate slags, we obtain:
v, = 5 X 10-
0.07 == 3.5
10-3
Moreover, we can use equation (2-79) in Section 2.12.2.2 to calculate the mean diameter of the inclusions: 150 X 21
)113
,urn = 0.31,um
EXERCISE
PROBLEMS
WITH SOLUTIONS
579
The different inclusion characteristics may now be estimated from equations (2-80) to (2-83): Number of particles per mm': N;
6Vv 7T(d )3
v
= 2.2 X 108 mm ?
108 X 0.31
10-3 mm ?
= 6.8 X 104 mm ?
Y'3 = 0.554
(2.2 ~ 108
)113
10
/LID
= 0.92
/LID
A comparison with Table 2.11 shows that the calculated inclusion characteristics reasonable agreement with those reported for C-Mn steel weld metals.
are in
(g) The characteristic growth pattern of columnar grains in bead-on-plate welds is shown schematically in Fig. 3.33. The first phase to form will be delta ferrite which subsequently decomposes to austenite via a peritectic transformation (see Fig. 3.72). The important question is whether re-nucleation of the grains will occur during solidification. In practice, this depends on the interplay between a number of variables which cannot readily be accounted for in a simplifiedanalysis, including the weld pool geometry, the cooling rate and the nucleation potency of the non-metallic inclusions. Broadly speaking, the energy barrier associated with nucleation of delta ferrite at manganese silicates is rather high (e.g. see Fig. 3.30), which suggests that formation of new grains ahead of the advancing solid/liquid interface is not very likely under the prevailing circumstances. Hence, the columnar grain zone would be expected to extend entirely from the fusion line towards the centre of the weld, as frequently observed in this type of welds. Moreover, Fig. 3.43 provides a basis for estimating the substructure of the weld metal columnar grains. Close to weld centre-line the local crystal growth rate will approach the welding speed (i.e. RL ~ 4 mm s-I).At the same time a simple analytical solution exists for the thermal gradient in the weld pool (equation (3-28: 2 X 1T X 0.025(1520- 20)2 ---------C mm 2678
-1
132C mm
-1
From this we see that a cellular-dendritic type of substructure is likely to form within the central parts of the fusion zone, in agreement with general experience (see Fig. 3.36). (h) Fig. 5.25 shows the location of the cap layer. Since the base plate is a Nb-microalloyed steel, the important grain boundary pinning precipitates within the HAZ are either NbC, NbN or a mixture of these. In the former case the equilibrium dissolution temperature may be estimated from the solubility product of the pure binary compounds. From equation (4-4) and Table 4.1, we have:
580
METALLURGICAL
MODELLING
OF WELDING
Td(NbC) and
6770 2.26-log[0.020][0.08]
10230 Td(NbN)
This shows that NbC is thermodynamically more stable than NbN. In practice, the real dissolution temperature may be significantly higher than that predicted from equation (4-4) because of the kinetic superheating (see discussion in Section 4.4, Chapter 4). The grain growth diagram in Fig. 5.21(a) provides a basis for estimating the effect of heating rate (heat input) on the dissolution kinetics. Taking the ordinate q 0 Iv equal to 2678/4000 = 0.67 kJ mm -1, we obtain:
This corresponds to a kinetic superheating of about 200C in the case of NbC. In the HAZ on the weld metal side (see Fig. 5.25), oxide inclusions may act as effective grain boundary pinning precipitates. These will be thermodynamically stable up to the melting point of the steeL (i) The austenite grain size profile across the base plate HAZ can be read from Fig. 5.21(a). Taking the ordinate qjv.equal to 0.67 kJ mm ", we see that the maximum austenite grain size at the fusion boundary will exceed 100 J-tm because of dissolution of the base metal grain boundary pinning precipitates. In the HAZ on the weld metal side, the situation is different. Here the stable weld metal oxide inclusions will impede austenite grain growth to a much larger extent. The limiting austenite grain size may be calculated from equation (5-21). Taking the Zener coefficient equal to 0.5 for oxide inclusions in steel (Fig. 5.4), we obtain: dv D1im = 0.5 2V = 0.5 2
v
0.31 3 5 X 10-3
J-tm
= 22 J.tm
Because of the phenomenon of epitaxial grain growth (see Section 3.3, Chapter 3), the initial size of the weld metal delta ferrite/austenite columnar grains would be expected to be comparable to the size of the HAZ austenite grains adjacent to the fusion boundary. Since the latter varies along the periphery of the fusion boundary at the same time as competitive grain growth leads to a general coarsening of the solidification microstructure with increasing distance from the fusion boundary, an average columnar austenite grain size of about 50 J-tm seems reasonable under the prevailing circumstances.
G) As an illustration of principles, we shall assume that the CCT diagram in Fig. 6.27(a) provides an adequate description of the base plate transformation behaviour during welding. The cooling time from 800 to 500C can be calculated from equation (1-67):
EXERCISE
581
LlTS/5 = n3(-e
500
1 -
-e -)
2a
SOD
from which
LltS/5
=~
LlTs/5
=~
2X5
X
Readings from Fig. 6.27(a) give the following microstructures and grain refined region of the HAZ, respectively:
Grain coarsened region (Tp = 1350C): Microstructure : 1000/0lath martensite Transformation start temperature: ,....... 470C Grain refined region (Tp = 10000e): Microstructure : ferrite + pearlite Transformation start temperature: ,..... 600C It follows that the observed difference in the HAZ transformation behaviour can mainly be attributed to a corresponding difference in the prior austenite grain size, which according to Fig. 5.21(a) is about 50 J.Lm at Tp = 1350 C and below 10 J.Lm at Tp = 1000C. In addition, small islands of plate martensite will form within the intercritical (partly transformed) HAZ, where the peak temperature of the thermal cycle has been between ACt and AC3 (see discussion in Section 6.3.8.2, Chapter 6). Just above the ACt temperature the volume fraction of the M-A (martensite-austenite) constituent is approximately equal to the base plate pearlite content (Fig. 6.66), which in the present case is about 8 vol 0/0,as judged from the steel carbon content. Considering the weld metal, the situation is different. Here the oxide inclusions will strongly affect the microstructure evolution by promoting intragranular nucleation of acicular ferrite (see discussion in Section 6.3.5, Chapter 6). In practice, the role of inclusions in weld metal transformation kinetics is difficult to assess and hence, we will take a more simplistic (pragmatic) approach to this problem by just comparing the total surface area available for nucleation of ferrite at prior austenite grain boundaries and inclusions, respectively (Sv(GB) versus Sv(I). The following three regions of the weld are considered: As-deposited weld metal: 2 2
S (GB)
v
= -
dy
= --
0.05
mm " = 40 mm-t
2
= -
2
= --
dy
0.022
mm " = 91 mm-1
582
METALLURGICAL
MODELLING
OF WELDING
Reheated weld metal (farfrom fusion line): In this case an estimate will be made for d" = 10 S (GB) = v d
p,m. =
From the above calculations it is apparent that the conditions for acicular ferrite formation are particularly favourable within the as-deposited weld metal (Sv(I) > Sv(GB)), and somewhat less favourable within the high peak temperature region of the weld HAZ (Sv(GB) > Sv(I)). In contrast, acicular ferrite would not be expected to form within the low peak temperature region of the HAZ, since nucleation of ferrite at austenite grain boundaries in this case will completely override nucleation at inclusions (Sv(GB) Sv(I)). This is also in agreement with general experience (e.g. see photographs of typical microstructures in Fig. 6.19(c) and (d)). (k) The maximum hardness/strength level within the grain coarsened region of the HAZ can be estimated from the diagrams presented in Section 7.2.2 (Chapter 7) if the steel composition and welding parameters fall within the specified range. Alternatively, we can use Fig. 7.19, which applies to low carbon micro alloyed steels. Taking the cooling time from 800 to 500C, Il.tS/5 ' equal to 3.3 s, we obtain:
HV max = ---380 VPN
"
200 mm ?
and
RpO.2
In general, a hardness rather than a strength criterion is used as a basis for evaluation of the risk of hydrogen cracking and H2S stress corrosion cracking during service. In the former case an upper limit of about 300 to 325 VPN is incorporated in many welding specifications to avoid problems with hydrogen cracking, but this restriction can be relaxed if specific precautionary actions are taken during the welding operation to reduce the supply of hydrogen as shown in Section 7.2.3 (Chapter 7). Considering the H2S stress corrosion cracking resistance a maximum hardness level of 248 VPN is strictly enforced in many welding specifications, as discussed previously in Section 7.2.4 (Chapter 7). Hence, significant tempering of the martensite would be required if the weldment is going to be used in environments containing sour oil or gas. (1) In general, the toughness requirements vary with the type of application, but for offshore structures a minimum CVN toughness of 35J at -40C is frequently specified. From the CVN data in Tables 8.2 and 8.4 it apparent that both the base plate and the weld metal meet this requirement. Moreover, auto-tempered low carbon martensite and polygonal ferrite, which form within the grain coarsened and grain refined region of the HAZ, respectively are known to have an adequate cleavage resistance.This means that the intercritical HAZ is the most critical region of the joint when it comes to toughness due to the presence of high carbon plate martensite within the ferrite matrix (see Figs. 6.61 through 6.65 and discussion in Section 7.2.2.3, Chapter 7). In practice, the problem may be solved by applying an appropriate post weld heat treatment (PWHT). (m) Since the properties of martensite depend on the carbon content, C-Mn steel weldments will generally be more prone to hydrogen cracking, H2S stress corrosion cracking and brittle fraction initiation in the HAZ than low carbon microalloyed steel weldments. This explains
EXERCISE
PROBLEMS
WITH SOLUTIONS
583
why the base plate carbon content has been gradually lowered to values well below 0.1 wt% in step with the progress in steel plate manufacturing technology.
Mn 2.0
Cr 16
Ni 12
0.03
(mm s") 5
200
15
"The arc efficiency factor may be taken equal to 0.4. No preheating is applied (To = 20C).
C: 0.02 wt%, Si: 0.5 wt%, Mn: 1.6 wt% Cr:18.5 wt%, Ni: 12 wt%, Mo: 2.75 wt% 470 MPa, Rm: 650 MPa, E1.: 32%, CVN -60: 150 J
RpO.2:
tData compiled from dedicated filler wire catalogues and welding manuals.
:j:
Analysis: The problem here is to evaluate the response of the base material to welding under the conditions described above. The analysis should be quantitative in nature and based on sound physical principles. The following input data are recommended:
A = 0.025 W mm -1C-1 pc = 0.005 J mm ? C-1
584
METALLURGICAL
MODELLING
OF WELDING
Specific questions: (a) Select an appropriate heat flow model for the system under consideration. (b) Estimate the minimum bead length which is required to achieve pseudo-steady down to a peak temperature of 1000 "C, state
(c) Estimate the deposition rate (in gA-1s-l), the weld cross section areas D and B (in mm-), and the dilution ratio BI(B + D) during welding. (d) Estimate the weld metal chemical composition for the given combination plate, filler wire and dilution ratio. of base
(e) Sketch the contour of the weld pool and the resulting columnar grain morphology in the x-y plane after solidification. Estimate also the weld metal delta ferrite content. (f) Evaluate the risk of solidification cracking during welding. (g) Calculate the austenite grain size profile across the HAZ. Estimate also the size of the columnar grains in the weld metal. (h) Evaluate the risk of chromium carbide formation in the HAZ during welding. (i) Estimate on the basis of the Hall-Petch relation the maximum load bearing capacity of the joint during service. Solution: (a) The problem of interest is whether we must use the general (but complex) Rosenthal plate solution (equation (1-81 or can adopt the simplified solution for a fast moving power source (equation (1-100. Fig 1.43 provides a basis for such an evaluation. The critical position will be the fusion line. If we neglect the latent heat of melting, the fJpln3 at the melting point becomes:
-L = nj 200XO.4x15x5
fJ
47TX 52 X 9.4
=049 .
Similarly, the dimensionless plate thickness is equal to: 5X2 8=--=1 2X5 Readings from Fig. l.43 show that we are outside the validity range of the simplified I-D model close to the fusion line, but that this solution is a good approximation within the low peak temperature region of the HAZ. Here equation (1-100)may be used in replacement of equation (1-81). (b) The duration of the transient heating period depends on the actual point of observation (i.e. the distance from the heat source). If we would like to apply the pseudo-steady state solution down to a peak temperature of 1000"C, the corresponding 11:/ 8fJ ratio becomes: . p
EXERCISE
585
1 (1480) OpD = 0.49 (980) = 3.1 From Fig. 1.31 we see that this ratio corresponds to a dimensionless radius vector (J5m of about 5. The duration of the transient heating period may now be read from Fig. 1.28. A crude extrapolation gives: r=10 from which 2a L =r- =10x
2
2x5 -5
mm=20mm
-IS-I)
The total area of fused metal can be read from Fig. 1.31. At the melting point the niOpD ratio is close to 2, which gives:
+ D) =
----u- = 0.25
12-9
Note that in these calculations we have assumed that A2 is equal to the sum of (B+D) in order to achieve realistic numbers. (d) The weld metal composition can be calulated from a simple 'rule of mixtures': [%X]weld = B
+ D [%X]BM
D B
+ D [%X]FW
By using input data from Tables 8.5 and 8.7, we get: [%C] = 0.02
[%Mn]
= 0.25
2 + 0.75
1.6 = 1.7
= 0.25
=
X 0
X X
+ 0.75
X 0.5
X X X
= 0.38
=
0.25
16 + 0.75 12 + 0.75
18.5
17.9
= 0.25
=
12 = 12.0 2.75
=
0.25
0 + 0.75
2.1
(e) The bead morphology can be read from Fig. 1.29. Taking the Ooln3 ratio at the melting point equal to 0.5, it is easy to verify that the geometry of the weld pool in this case is tearshaped. The columnar grain structure is therefore similar to that shown in Fig. 3.11(b). When the composition is known the weld metal microstructure can be read from Fig. 7.53 by considering the resulting chromium and nickel equivalents: [Cr]eq [Ni]eq
=
17.9 + 2.1
+ 1.5
0.38
= X
20.6 1.7
= 12.0 + 30
0.025
+ 0.5
= 13.6
This gives a delta ferrite content of about 7 vol 0/0. (f) Normally, a minimum delta ferrite content of about 5 to 10 vol % is specified to avoid problems with solidification cracking in the weld metal (see discussion in Section 7.3.4, Chapter 7). This requirement is clearly met under the prevailing circumstances. (g) The HAZ austenite grain size in different positions from the fusion boundary can be read from Fig. 5.30(b). In the present example the net heat input per mm- of the weld is equal to: q/vd
= 200 X
? =
0.12 k.lmm ?
This corresponds to a maximum austenite grain size of about 60jLm close to the fusion boundary, which also is a reasonable estimate of the weld metal columnar grain size. (h) The most critical position is the low peak temperature region of the weld HAZ where Tp is between 800 and 1000 DC,as shown in Section 6.4.2 (Chapter 6). However, it is evident from Fig. 6.69 that the risk of chromium carbide formation in this case is negligible because of the low base plate carbon content. Hence, the corrosion resistance will not be significantly affected by the welding operation. (i) The minimum HAZ strength level may conveniently be calculated from equation (7-21), using input data from Example 7.5 (page 530):
RpO.2
(min)
246.5
+ 227/(60)112 MPa
276 MPa
This gives the following strength reduction factor for the joint:
f = 276/300 = 0.92
EXERCISE
PROBLEMS
WITH SOLUTIONS
587
(in wt%). Mn
0.52
Mg
0.64
Fe
0.19
v (mm
10
S-l)
tThe arc efficiency factor may be taken equal to 0.8. No preheating is applied (To = 20 e).
Table 8.10 Exercise problem III: Operational Shielding gas: Gas flow rate: Wire diameter: Wire feed rate: Wire composition: Argon
20 NI per min
1.6mm
5.5 m per min
Wire I: RpO.2 : 55 MPa, Rm: 165 MPa, El.: 180/0 Wire II: RpO.2: >130 MPa, Rm: >280 MPa, El.: >170/0, CVN+2o:>30 J
tData compiled from dedicated filler wire catalogues and welding manuals.
t Values refer to all weld metal deposit.
Analysis: The problem here is to evaluate the response of the base material to welding under the conditions described above. The analysis should be quantitative in nature and based on sound physical principles. The following input data are recommended: A = 0.248 W mm-1oC-l
pc
= 0.004 J mm-3C-l
Hm -H
0
=2.51 Lmm
>'
Specific questions:
588
METALLURGICAL
MODELLING
OF WELDING
1500
30
40
50
60
70
80
90
100
1414C 1300
1100
cD .a
Q)
900
0.
E 700
500
300
AI
10
20
30
40
50
60
70
80
90
100
Si
700
10
20
30
40
50
60
70
80
90
100
600
500
cD
Cti
:J
CD
400
0.
E 300
200
100
AI
10
20
30
40
50
60
magnesium
70
80
90
100
Mg
Weight percent
EXERCISE
PROBLEMS
WITH SOLUTIONS
589
(a) Select an appropriate heat flow model for the system under consideration. (b) Estimate the minimum bead length which is required to achieve pseudo-steady state down to a peak temperature of 200C. (c) Estimate the value of the deposition coefficient k' (in gA -is-i), the weld cross section areas Band D (in mm"), and the dilution ratio BI(B + D) during welding. (d) Estimate the content of Mg and Si in the weld metal. (e) Sketch the weld metal columnar grain structure and the segregation pattern during solidification. Indicate also the type of substructure which forms at different positions along the periphery of the fusion boundary. Relevant binary phase diagrams are given in Figs. 8.5 and 8.6. (f) Evaluate the risk of solidification cracking during welding. (g) Evaluate the risk of liquation cracking in the HAZ during welding. (h) Sketch the sequence of reactions occurring within the HAZ during welding. Then estimate the following quantities: - The temperature for incipient dissolution of {3" .
- The total width of the HAZ (referred to the fusion boundary). - The temperature for full dissolution of {3" .
- The total width of the fully reverted HAZ (referred to the fusion boundary). (i) Estimate for each combination of filler wire and parent material an overall strength reduction factor which determines the load bearing capacity of the joint.
G) Imagine now that the same extrusion instead is used in the fully annealed (0- temper) condition with a Vickers hardness and tensile yield strength of 50 VPN and 100 MPa, respectively. To what extent will the temper condition affect the microstructure and strength evolution during welding?
Solution: (a) The problem of interest is whether we must use the general (but complex) Rosenthal thin plate solution (equation (1-81 or can adopt the simplified solution for a fast moving high power source equation (1-100. Fig 1.43 provides a basis for such an evaluation. The most critical position will be the fusion line. If we neglect the latent heat of melting, the 8p In3 ratio at the melting point becomes:
47T X
622
2.51
0.40
590
METALLURGICAL
MODELLING
OF WELDING
Readings from Fig. 1.43 show that we are outside the validity range of the simplified I-D solution close to the fusion line, but that equation (1-100) may be used (with some reservations) within the low peak temperature region of the HAZ. (b) The duration of the transient heating period depends on the actual point of observation (i.e. the distance from the heat source). If we would like to apply the pseudo-steady state solution down to a peak temperature of 200C, the corresponding njoOp ratio becomes:
n3
0/3
= 0.37
= 3.25
From Fig. 1.31 we see that this ratio corresponds to a dimensionless radius vector (ISm of about 5. The duration of the transient heating period may now be read from Fig. 1.28. A crude extrapolation gives: T~10 from which
L =
2
2a
T -
2 X 62 = lOX --mm=s l Zd mm 10
It follows that the minimum bead length required to achieve pseudo-steady state during aluminium welding is much longer than in steel welding due to the pertinent differences in the heat flow conditions (e.g. see Example 1.5, Chapter 1). (c) The value of the deposition coefficient may be estimated from the data in Table 8.10:
k' = (4) X 1.62 X ~
(5500)
X
2.7
10 -3
1 200 gA -IS
-1
= 0.0025 gA -1S-1
This value corresponds to a k '/p ratio of about 0.92 mm? A -IS -1, which is in excellent agreement with the data quoted in Table 1.7. The area D of deposited metal thus becomes (see equation (1-120)): 0.0025
D = 10 X 2.7 X 10 -3 X 200 mm? ~ 20 mm?
The total area of fused metal can be read from Fig. 1.31. At the melting point the ratio is close to 0.93, which gives:
A2
11:/ f) . p
= 0.98
= 0.9 X
0.4
= 0.36
+ D) = A2
4a2
EXERCISE
PROBLEMS
WITH SOLUTIONS
591
B (B+D)
55 -20 55
0.64
Note that in these calculations we have assumed that A2 is equal to the sum of (B order to achieve realistic numbers. (d) The weld metal composition can be calulated from a simple 'rule of mixtures':
+ D) in
B
[%X]weld
= B + D [%X]BM
+ B +D
[%X]FW
By using input data from Tables 8.8 and 8.10, we get: Wire I: [%Mg]
[%Si]
=
0.64
X
0.64 + 0.36
0 = 0.41
= 0.64
0.98
+ 0.36
5 = 2.43
0.64
X
0.64 + 0.36
X
5 = 2.21
= 0.64
0.98 + 0.36
0 = 0.63
(e) The bead morphology can be read from Fig. 1.29. Taking the (Joln3 ratio at the melting point equal to 1, it is easy to verify that the shape of weld pool in this case is elliptical. The columnar grain structure is therefore similar to that shown in Fig. 3.11(a). Moreover, Fig. 3.43 provides a basis for estimating the substructure of the weld metal columnar grains. Close to weld centre-line the local crystal growth rate will approach the welding speed (i.e. RL = 10 mm S-1). At the same time a simple analytical solution exists for the thermal gradient in the weld pool (equation (3-29:
X 0.248 X 0.004 X 10
G L -
21T"
(200
0.8
28/5)2
-1
= 20 Cmm -,1
From this we see that a cellular-dendritic type of substructure is likely to form within the central parts of the fusion zone, in agreement with general experience. If we only consider the contribution from the major alloying element in each case, the Scheil equation (equation (3-46 may be used for an analysis of the segregation pattern during solidification. By using input data from the binary phase diagrams in Figs. 8.5 and 8.6, we get:
592 Wire I: ko
= 0.16 =
Cs (min) Ceut.
=
ko Co = 0.16
2.43 wt%
0.39 wt%
12.6 wt%
t.: = ( 2.43
Wire II: ko
==
12.6
)11(0.16-1)
==
0.44
Cs (min)
= ko Co
==
0.44
2.21 wt%
= 0.97 wtO/o
Ceut. = 36 wt 0/0
t.: = ( 2.21
36
)11(0.44-1)
From this we see that the amount of eutectic liquid which forms during solidification is sensitive to variations in the filler wire chemical composition (i.e. the Si or Mg content). (f) Fig. 7-54 provides a basis for evaluation of the hot cracking susceptibility. Wire I In this case the fraction of eutectic liquid is so abundant that it backfills and 'heals' all incipient cracks. Hence, the hot cracking susceptibility is low. Wire II When the AI-Mg filler wire is used the fraction of eutectic liquid is just large enough to form continuous films at the columnar grain boundaries. Hence, the hot cracking susceptibility is high. (g) Liquation cracking arises from melting of specific phases present within the base material (e.g. Mg2Si and Si), as discussed in Section 7. 4.2.1 (Chapter 7). Fig. 7.61 provides a basis for evaluating the HAZ cracking susceptibility: Wire I In this case the risk of liquation cracking is small because the solidus temperature of the weld metal is lower than the actual melting temperatures of the base metal constituent phases. Wire II Due to the high B/(B + D) ratio involved, the solidus temperature of the weld metal will exceed the actual melting temperatures of the base metal constituent phases. This may lead to liquation cracking in parts of the HAZ where the peak temperature is greater than, say, 555 to 559C. (h) The sequence of reactions occurring within the HAZ during welding of AA 6082-T6 aluminium alloys is shown in Fig. 7.62. In the present case we can use Fig. 4.24 for a quantitative
EXERCISE
PROBLEMS
WITH SOLUTIONS
593
analysis of the (3" dissolution kinetics. Taking the net heat input per mm- qjvd equal to 0.08 klmm ", we obtain: Total width of HAZ LlYm(J) Temperature for incipient dissolution of f3" First we need to estimate the corresponding
l/lm
20 mm
from which
Tp(J) = 20
+ 0.31(652-20) C
216C
Temperature for complete dissolution of f3" First we need to estimate the corresponding
l/lm l/lm
= 0.4 + 3 X 10/(2
62) = 0.64
from which
Tp(2) = 20
+ 0.84(652-20) C
550C
A comparison with the phase diagram in Fig. 4.8 shows that the calculated temperature for incipient dissolution of {3"is in good agreement with that obtained from the solubility product. (i) The yield strength in HAZ and the weld metal can be obtained from Fig. 7.67 and Table 8.10, respectively:
594
METALLURGICAL
~ELLING
OF WELDING
RpO.2
::::::
. pO 2
::::::
280 MPa
= 55/280 = 0.20
RpO.2
RpO.2
= 280 MPa
U) When the material is present in the O-temper condition, it will contain an appreciable amount of the equilibrium /3-Mg2Si phase. This will tend to accelerate the problem with liquation cracking within the HAZ during welding.
In addition, it is evident from Figs. 4.4 and 4.8 that the equilibrium /3-Mg2Si phase is thermodynamically much more stable than the metastable /3" phase. In practice, this means that only a narrow solutionised zone will form adjacent to the fusion boundary. However, within this zone significant strength recovery may occur after welding due to natural ageing effects (see Fig. 4.5), which may result in a HAZ hardness and tensile yield strength level of about 80 VPN and 190 MPa, respectively. Hence, for the O-tempered material, we get: Wire I HAZ: RpO.2 (max) :::::: 190 MPa, Weld metal:
RpO.2::::::
RpO.2
::::::
100 MPa
f = 55/100 = 0.55
Wire II HAZ: RpO.2 (max) :::::: 190 MPa, Weld metal: > 130 MPa, Base metal:
= 100 MPa
RpO.2
RpO.2
f = 100/100 = 1
595
Subject Index
A absorption of elements see hydrogen, nitrogen, oxygen acicular ferrite in low-alloy steels
558-59
during friction welding 555-63 hardness and strength distribution 550-5, 560-63 strengthening mechanisms in alloys 547-8 amplitude of weaving - definition
428-43
crystallography of 428-9 nature of 430-2 nucleation and growth of 432-43 texture components of 429-30 acicular ferrite in wrought steels 444 aluminium as alloying element in steel effect on inclusion composition
example (1.16) - SA welding of steel 97-8 example (1.17) - SMAW welding of steel 98-9 example (1.18) - Jackson equation
99-100
Bessel functions modified 46, 47,
49
boron in steel effect on transformation behaviour
80
arc atmosphere composition 132 see also shielding gases arc efficiency factors 26-7 definition 26 selected values 27 arc welding 24-5 definition of processes 24-5 austenite grain size in low-alloy steels 409-
413-17
segregation of 294 weld properties 493-4, 505-6 bowing of crystal 240-2 Bramfitt's planar lattice disregistry model 244 see also solidification of welds
202,206-12
effect on solidification microstructure 246-7, 272, 293 effect on weld properties 481-3,
486-88
solubility product of precipitates
303
aluminium weldments 458-67,536-
13
primary precipitation in fusion welds 292-5 austenite formation in low-alloy steels 449-50 conditions for 450-3 austenitic stainless steels 453-8 see also stainless steel weldments characteristics of 527 chromium carbide formation 456 grain growth diagrams for steel welding 375-9 weld decay area 456-8 Avrami equation in solid state transformations see also solid state transformation and transportation kinetics additivity in 404-5,475-6 exponents in 401
63
age-hardenable alloys 458-67 quench sensitivity 459-64 precipitation conditions during cooling 459-61 strength recovery during natural ageing 461-4 subgrain evolution in friction welding 464-7 characteristics 536 constitutionalliquation in AI-Mg-Si alloys 542-3 in AI-Si alloys 541-2 example (7.9) - minimum HAZ strength level 554-6 example (7.8) - weld metal hot cracking 544-6 example (7.7) - weld metal solidification cracking 537-40 example (7.10) - minimum HAZ hardness level 562-563 HAZ microstructure evolution
C carbon equivalents 496,521 carbon as alloying element in steel austenitic stainless steels 453-58 weld deposits 424 carbon-manganese steel weld metals, grain growth in 370-2 casting, structural zones 221-2 cell/dendrite alignment angle 249-50 see also solidification of welds cellular substructure 251-2 see also solidification of welds chemical reaction model - overall
116-17
chromium carbide formation in austenitic stainless steels 456 chromium-molybdenum steel welds, grain growth in 372-5 columnar grains 228-30 see also solidification of welds columnar to equiaxial transition 268-
72
B Bain orientation region 436-37 bainite in low-alloy steels 444-7 lower 447 upper 444-7 bead morphology 96-100 bead penetration 99-100 deposit and fused parent metal 968 see also solidification of welds competitive grain growth 234-9 see also solidification of welds concentration displacements during welding see oxygen, absorption of cooling condition during solidification 221-2 cooling rate, C.R. thick plate welding 37
547-63
constitutive equations 548-50,
558-59
during friction welding 555-63 during fusion welding 547-56 hot cracking 540-6 factors affecting 544-6 solidification cracking 536-40
596
cooling rate, C.R. (cont'd) thin plate welding 53-4 cooling time, I1t8/5 thick plate welding 36 thin plate welding 53 cooling time, t 100 103
METALLURGICAL MODELLING OF WELDING implications of model 86 model limitations 86-7 2-D heat flow model 80-2 see also heat flow models Dorn parameter 501 duplex stainless steels 531-2 HAZ toughness 532 HAZ transformation behaviour 532 Sievert's law 124 gas porosity in fusion welds 279-85 growth and detachment of gas bubbles 281-3 nucleation of gas bubbles 279-81 separation of gas bubbles 283-5 Gaussian error functions, definition 112 Gaussian heat distribution 112-15 see also distributed heat sources Gibbs-Thomson law 309 Gladman equation 344-45 grain boundary ferrite 408-27 crystallography of 408 growth of 422-7 nucleation of 408-22 grain detachment 250 see also solidification of welds grain growth 337-86 computer simulation 380-3 diagrams construction of 360-4 axes and features of 363-4 calibration procedures 361-3 heat flow models 360-1 for steel welding 360-79 case studies 364-79 C-Mn steel weld metals 370-2 Cr-Mo low alloy steels 372-5 niobium-microalloyed steels 367-70 titanium-microalloyed steels 364-7 type 316 austenitic stainless steels 375-9 driving pressure for 339-40 example (5.3) - austenite grain size in niobium-microalloyed steels 358-60 example (5.2) - austenite grain size in Ti microalloyed steels 354-5 example (5.1) -limiting austenite grain size in steel weld metals 344-5 grain boundary mobility 337-43 drag from impurities 340-1,3423 drag froma random particle distribution 341-3 driving pressure for growth 339-40 grain structures, characteristics 337-9 growth mechanisms 345-60 nomenclature 384-6 normal grain growth 343-60 size, limiting 343-5 Griffith's equation 486, 494 gross heat input - definition 37
D Delong diagram 535 delta ferrite, primary precipitation of 290-1,292-5 dendrite arm spacing 261-8 primary 261-4 secondary 264-8 dendrite fragmentation 250 see also solidification of welds dendrite substructure 252-4 see also solidification of welds dendrite tip radius 260 see also solidification of welds deoxidation reactions in weld pools 180-92 example (2.9) - homogeneous nucleation of MnSi03 182-4 growth and separation of oxide inclusions 184-90 buoyancy (Stokes flotation) 1856 fluid flow pattern 186-8 separation model 188-90 nucleation model 182 nucleation of inclusions 182-4, 219-20 overall deoxidation model 201 deposit - amount of weld metal 96 deposition rate 96 dissociation of gases in arc column 117-20 distributed heat sources 77-91, 11215 general solution 77-80 simplified solution (Gaussian heat distribution) 112-15 simplified solution (planar heat distribution) 80-91 case study (1.2) - surfacing with strip electrodes 87 case study (1.3) - GTA welding with a weaving technique 87-91 dimensionless operating parameter 82 dimensionless time 82 dimensionless y- andz-coordinates 82 example (1.13) - effect of weaving on temperature distribution 83-6
E energy barrier to solidification 2256 see also solidification of welds enthalpy of reaction 302-3 definition of 302 values 303 entropy of reaction 302-3 definition 302 values 303 epitaxial solidification 222-7 equiaxed dentritic growth 268-72 equilibrium dissolution temperature of precipitates 303-4 see also solidification of welds error functions see Gaussian error functions
F
fluid flow pattern in weld pools 1868,228-9 flux basicity index 171 friction welding 18-24 see also aluminium weldments dimensionless temperature 20 dimensionless time 20 dimensionless x-coordinate 21 example ( 1.4) - peak temperature distribution 23-4 heat flow model 18-22 temperature-time pattern 23 Fritz equation 281-2 fume formation, rate of iron 157 manganese 156-60 silicon 152-57 fused parent metal - amount of 98
G gas absorption, kinetics of 120-22 rate of element absorption 121-2 thin film model 120 gas desorption, kinetics of 123-4 rate of element desorption 123-4
SUBJECT INDEX growth rate of crystals 230-9 local 234-9 nominal 230-4 see also solidification of welds I implant testing 520-1 see also hydrogen cracking inclusions in welds - origin 192-
597
heat flow model 7 temperature-time pattern 8 low-alloy steel weldments 477-526 acicular ferrite in 428-43 crystallography of 428-9 nature of 430-2 nucleation and growth in 432-43 texture components of 429-30 austenite formation in 449-50 conditions for 450-3 bainite in 444-7 lower 447 upper 444-7 case study (7.1) - weld bead tempering 501-2 example (7.1) - low-temperature toughness of welds 488-91 example (7.2) peak HAZ strength level 496-9 example (7.3) -location of brittle zones 508-9 HAZ mechanical properties 494-
212
H Hall-Petch relation 529 heat flow models distributed heat sources constituent elements and phases in inclusions 202-4 example (2.10) - computation of inclusion volume fraction
77-91,
194-5
example (2.12) - computation of total number of constituent phases in inclusions 211-12 prediction of inclusion composition
112-15
grain growth diagrams 360-1 instantaneous heat sources 5-6 local preheating 100-3 medium thick plate solution 59-76 thermal conditions during interrupted welding 91-5 thermal conditions during root pass welding 95-6 thick plate solutions 26-45 thin plate solutions 45-59 heat input see heat flow models Hellman and Hillert equation 344-45 Hollomon-Jaffe parameter 500 hydrogen, absorption of 128-41 content in welds 132 covered electrodes 134-8 combined partial pressure of 134 example (2.1) - hydrogen absorption in GTAW 133-4 example (2.2) - hydrogen absorption in SMA W 136 in gas-shielded welding 131-4 hydrogen determination 128-30 implications of Sievert's law 140 reaction model 130-1 sources of hydrogen 128 in submerged arc welding 138-40 effect of water content in flux 138 example (2.3) - hydrogen absorption in SAW 139-40 hydrogen cracking in low-alloy steel weld metals 509-24 diffusion in welds 514-18 diffusivity in steel 514 HAZ cracking resistance 518-24 mechanisms of 509-12 solubility in steel 513-14 hydrogen in multi-run weldments
204-10
size distribution of inclusions 195-
202
coarsening mechanism 196-201 effect of heat input 196 example (2.11) - computation of number density and size distribution of inclusions 201-2 volume fraction 193-5 stoichiometric conversion factors
509
hardness and strength level 495-
194-5
instantaneous heat sources 5-6 line source 5--6 plane source 5 point source 5-6 interface stability 254-9 interfacial energies 242-43 see also solidification of welds interrupted welding, thermal conditions 91-5 example (1.14) - repair welding of steel casting 93-5 heat flow models 93
9
tempering 500-2 toughness 502-9 hydrogen cracking 509-24 diffusion in welds 514-18 diffusivity in steel 514 example (7.4) - hydrogen cracking under hyperbaric welding conditions 521-4 HAZ cracking resistance 518-24 implant testing 520-1 mechanisms of 509-12 solubility in steel 513-14 hydrogen sulphide corrosion cracking 524-6 prediction of 525-6 threshold stress for 524-5 martensite in 447-53 austenite formation, kinetics of
448
449-50
L latent heat of melting 3 lattice disregistry see Bramfitt's planar lattice disregistry model local fusion in arc strikes 7-10 dimensionless operating parameter
7
conditions
for
450-3
plate (twinned) 447-53 mechanical properties 477-94 ductile to brittle transition 486-
140-1
hydrogen in non-ferrous weldments
91
reheating 491-4 resistance to cleavage fracture
141
hydrogen sulphide corrosion cracking in low-alloy steel weld metals 524-6 prediction of 525-6 threshold stress for 524-5 hyperbaric welding 176-80, 521-4
dimensionless radius vector 7 dimensionless temperature 7 dimensionless time 7 example (1.1) - weld crater formation and cooling conditions 9
485-6
resistance to ductile fracture 4805 strength level 478-80 transformation behaviour 290-5,
406-53
598
METALLURGICAL MODELLING OF WELDING melting efficiency factor 89 mixing ratio 98-9 moving heat sources 24 see also heat flow models Ostwald ripening see particle coarsening oxygen, absorption of 148-80 classification of shielding gases 166 overall oxygen balance 166-9 content in welds 148 covered electrodes 173-80 absorption of carbon and oxygen
low-alloy steel weldments (cont'd) solidification primary precipitation of austenite 292-5 primary precipitation of delta ferrite 290-1, 292-5 solid state acicular ferrite 428-44 bainite 444-7 grain boundary ferrite 408-27 martensite 447-53 microstructure classification
N
net arc power, definition 26 niobium-microalloyed steels, grain growth in 367-70 nitrogen, absorption of 141-7 content in welds 143 covered electrodes 143-6 gas-shielded welding 142 nominal composition 147 sources of 142 submerged arc welding 146-7 example (2.4) - nitrogen content in weld metal deposit 146-7 non -isothermal transformations additivity principle 403 and Avrami equation 404-5 isokinetic reactions 404 non-additive reactions 405-6 non-steady heat conduction biaxial conduction 2 triaxial conduction 2 uniaxial conduction 2 nucleation, energy barrier to 225-6 nucleation, homogeneous 182-4,
176-7
loss of silicon and manganese
177-8
the product [%C] [%0] 179-80 reaction model 174-6 effects of welding parameters 169-
406
nomenclature for 406-8 Widmanstatten ferrite 427-8 Ludwik equation 524
70
amperage 169-70 voltage 170 welding speed 170 example (2.8) - oxygen consumption and total CO evolution during GMAW 166-9 gas arc metal welding 148-70 manganese evaporation 156-60 example (2.6) - fume formation rate of manganese 157-60 sampling of elevated concentrations 149 carbon oxidation 149-52 silicon oxidation 152-6 example (2.5) - fume formation rate of silicon 156 SiO formation 154-5 total oxygen absorption 162-4,
303
martensite in low-alloy steels 447-53 austenite formation, kinetics of
449-50
lath 447 M-A formation, conditions for
450-3
plate (twinned) 447-53 martensitic stainless steels; characteristics of 527 mass transfer in weld pool, overall kinetic model of 124-7 medium thick plate solution 59-76 see also heat flow models dimensionless maps for heat flow analysis 61-72 case study (1.1) - temperature distribution in steel and aluminium weldments 69-72 construction of maps 61 cooling conditions close to weld centre line 63 example (1.12) - aluminium welding 68-9 isothermal contours 65 limitation of maps 65-8 peak temperature distribution 613 retention times at elevated temperatures 63-5 experimental verification 72-5 peak temperature and isothermal contours 75 weld cooling programme 72-4 weld thermal cycles 72 general heat flow model 59-61 practical implications 75-6
219-20
see also deoxidation reactions in weld pools nucleation, potency of particles 242-
173
transient oxygen concentrations
6
see also solidification of welds nucleation, rate of heterogeneous during solidification 248-9,
160-4
example (2.7) - slag formation in GMAW 164-5 submerged arc welding 170-4 concentration displacements
272
nucleation in solid state transformation kinetics 389-95 in C-curve modelling 390-95 nucleation of gas bubbles in fusion welds 279-81 nucleation of grain boundary ferrite in low-alloy steels 408-22 austenite grain size 409-13 boron alloying 413-17 factors affecting ferrite grain size
172
flux basicity index 171 total oxygen absorption 173 transient oxygen absorption
172-4
oxygen, retained in weld metal 1902 implications of model 192 thermodynamic model of 190-2
420-2
solidification-induced segregation
417-19
p particle coarsening 314-16 applications to continuous heating and cooling 314-16 example (4.4) - coarsening of titanium nitride in steel 315-
o
operating parameter, dimensionless point and line heat source models
31
weaving model 82
16
kinetics 314
SUBJECT INDEX particle dissolution 316-33 analytical solution 316-25 case study (4.1) - solute distribution across HAZ 330-2 example (4.5) - isothermal dissolution of NbC in steel 320-2 example (4.6) - dissolution of NbC within fully transformed HAZ 323-5 numerical solution 325-33 application to continuous heating and cooling 329-30 process diagrams for aluminium butt welds 332-3 Peclet number for weld pools 18687 peritectic solidification in welds 2905 see also low alloy steel weldments primary precipitation of 'Yp -phase 290 transformation behaviour 290-5 precipitate growth mechanisms liquid state 196-201 solid state 395-400 diffusion-controlled 397-400 interface-controlled 396-7 precipitate stability 301-36 see also particle coarsening and particle dissolution example (4.1) - equilibrium dissolution temperature of nitride precipitates 304 example (4.2) - equilibrium volume fraction of Mg.Si 307-8 example (4.3) metastable W'(Mg2Si) solvus 312-14 nomenclature 334-6 solubility product 301-14 equilibrium dissolution temperature 303-4 stable and metastable solvus boundaries 304-14 thermodynamic background 301-3 preheating, local 100-3 heat flow model 100 dimensionless half width of preheated zone 101 dimensionless temperature 101 dimensionless time 101 example (1.19) - cooling conditions during steel welding 102-3 time constant 101 pseudo-equilibrium, concept of 1223 pseudo-steady state temperature distribution, definition 24 R reversion see particle dissolution example (1.15) - cooling conditions during root pass welding 95-6 heat flow model 95 Reynold number definition 187 of gas bubbles 284 of particles 187 root pass welding, thermal conditions in 95-6 Rosenthal equations see thick and thin plate solutions
599
example (3.13) - application of Scheil equation 276-7 interface stability criterion 254-9 example (3.6)-critical temperature gradient for planar solidification front in AI-Si welds 256-8 example (3.7) - substructure characteristics of AI-Mg welds 258-9 primary dendrite arm spacing 2614
S Scheil equation 272-76 modified 276 original 272-5 separation of gas bubbles in fusion welds 283-5 shielding gases see oxygen, hydrogen and nitrogen, absorption of CO-evolution 166-69 Sievert's law 124, 140 silicon in aluminium solubility product of precipitates 303 solid state transformations in welds 387-476 AI-Mg-Si alloys 458-67 austenitic stainless steels 453-8 Avrami equation in, additivity in 404-5,475-6 high strength low-alloy steels 40653 kinetics see transformation kinetics nomenclature 471-4 solid state transformation kinetics 387-406 see also transformation kinetics driving force for 387-8 non-isothermal transformations 402-6 nucleation in solids 389-95 overall 400-2 precipitates, growth of 395-400 solidification cracking in weldments aluminium 536-40 stainless steel 532-5 solidification microstructures 251-68 columnar to equiaxed transition 268-72 dendrite tip radius 260 equiaxed dendritic growth 268-72 example (3. J 2) equiaxed dendritic growth in AI-Si welds 270-2
example (3.8) - effect of heat input on primary dendrite arm spacing in welds 262-3 example (3.9) - variation ofprimary dendrite arm spacing across fusion zone 263-4 secondary dendrite arm spacing 264-8 example (3. J 0) - secondary dendrite arm spacing in thick plate GTA AI-Si welds 2667 example (3. J 1) - secondary dendrite arm spacing in thin plate GTA AI-Si butt welds 267-8 local solidification time 265-8 substructure characteristics 251-4 cellular 251-2 dendritic 252-154 solidification of welds 221-300 columnar grain structures and morphology 228-30 epitaxial solidification 222-7 energy barrier to solidification 2256 implications of 226-7 growth rate of columnar grains 230-9 example (3.1) - nominal crystal growth rate in thin sheet welding of austenitic stainless steels 234 example (3.2) - local dendrite growth rate in single crystal welds 237-9 local crystal growth rate 234-9 nominal crystal growth rate 2303 renucleation of crystals 242-50 critical cell-dendrite alignment angle 249-50 dendrite fragmentation 250 example (3.4) - nucleation potency of TiN with respect to delta ferrite 245-6
600
METALLURGICAL MODELLING OF WELDING characteristics of 527 chromium carbide formation 456 grain growth diagrams for steel welding 375-9 example (7.5) - variation in HAZ austenite grain size and strength level 530 example (7.6) - weld metal solidification cracking 533-5 HAZ corrosion resistance 527-9 HAZ strength level 529-30 HAZ toughness 530-2 solidification cracking 532-5 weld decay area 456--8 duplex HAZ toughness 532 HAZ transformation behaviour 532 stereometric relationships (number of particles per unit volume, number of particles per unit area, total surface area per unit volume, and mean particle volume spacing) 201-2 Stokes law 185-6, 187, 284-5 substructure of welds 251-4 see also solidification of welds dimensionless y-coordinate 31 dimensionless z-coordinate 31 distribution of temperatures 3140 example (1.5) - duration of transient heating period in aluminium welding 30-1 example (1.6) - thermal contours 37-9 example (1.7) - weld geometry 39-40 isothermal zone widths 32-4 length of isothermal enclosures 34-5 simplified solution 41-5 example (1.8) - retention time in steel welding 44-5 temperature-time pattern 41-4 2-D heat flow model 41-2 volume of isothermal enclosures 35-6 transient heating period 29-30 thin plate solutions 45-59 see also heat flow models example (1.9) - duration of transient heating period in aluminium welding 48-9 pseudo-steady state temperature distribution 49-56 cooling conditions close to weld centre line 53-4 example ( 1.1 0) - weld geometry and cooling rate 54-6 isothermal zone widths 49-51 length of isothermal enclosures 51-3 simplified solution 56--9 example (1.11) - retention time in steel welding 59 I-D heat flow model 56-7 temperature-time pattern 57-8 transient heating period 29-30 titanium as alloying element in steel effect on inclusion composition 203-4,208-12 effect on solidification microstructure 244-46, 272-74 effect on grain growth 354-55, 364-67 effect on transformation behaviour 435-41,444 effect on weld properties 488-94 solubility product of precipitates 303 titanium-microalloyed steels, grain growth in 364-7 see also low alloy steel weldments transformation kinetics 387-406 Avrami equation 400-5, 475-6
renucleation of crystals (cont'd) example (3.5) - nucleation potency ofy-Al203 with respect to delta ferrite 246--7 grain detachment 250 nucleation potency of second phase particles 242-4 rate of heterogeneous nucleation 247-9 reorientation of columnar grains 239-50 bowing of crystal 240-2 example (3.3) - bowing by dendritic branching 240-2 structural zones 221-7 solubility of gases in liquids and solids 125-6 hydrogen in Al 125 hydrogen in Cu 125 hydrogen in Fe 125,513-4 hydrogen in Ni 125 nitrogen in Fe 126 see also gas absorption and gas desorption solubility product 301-14 equilibrium dissolution temperature 303-4 stable and metastable solvus boundaries 304-14 thermodynamic background 3013 solute redistribution in welds 272-89 example (3.14) - formation of hydrogen bubbles in weld pools 282-3 example (3.15) - separation of hydrogen bubbles in weld pools 284-5 example (3.16) - solute redistribution during cooling in austenite regime 287-9 gas porosity 279-85 homogenisation of microsegregations 286-9 macrosegregation 277-9 microsegregation 272-7 spot welding 10-14 dimensionless operating parameter 11 dimensionless radius vector 11 dimensionless time 11 example (1.2) - cooling conditions 12-14 heat flow model 11 refined model for 110-11 temperature-time pattern 12 stainless steel weldments 527-35 see also austenitic stainless steels austenitic
T texture in welds solidification 221-24, 290~93 solid state 429-30 thermal properties of metal and alloys 3-4 conductivity 3-4 diffusivity 3 heat content at melting point 3 latent heat of melting 3 melting point 3 volume heat capacity 3 thermit welding 14-18 dimensionless temperature 16 dimensionless time 16 dimensionless x-coordinate 16 example ( 1.3) - cooling conditions 16-18 heat flow model 14-15 temperature-time pattern 17 thick plate solutions 26-45 see also heat flow models pseudo-steady state temperature distribution 31-40 cooling conditions close to weld centre line 36-7 dimensionless operating parameter 31 dimensionless x-coordinate 31
SUBJECT INDEX transformation kinetics (cont' d) additivity in 404-5,475-6 exponents in 401 driving force for 387-8 example (6.1) - C-curve analysis 394-6 example (6.2) - conditions for ferrite formation within HAZ 410-12 example (6.3) - volume fraction of grain boundary ferrite in HAZ412-13 example (6.4) - ferrite/martensite formation in HAZ 416-17 example (6.5) - displacement of ferrite C-curve due to segregation 418-19 example (6.6) - variation in ferrite grain size across HAZ 421-2 example (6.7) - volume fraction of allotriomorphic ferrite in weld deposit 425-7 example (6.8) volume of acicular ferrite plate 440-1 example (6.9) - conditions for acicular ferrite formation 442-3 example (6.10) - conditions for chromium carbide formation 456-8 example (6.11) - conditions for W(Mg2Si) precipitation 460-1 example (6.12) - ageing characteristics of aluminium weldments 463-4 non-isothermal transformations 402-6 nucleation in solids 389-95 overall 400-2 precipitates, growth of 395-400 type 316 austentitic stainless steels, grain growth in 375-9 see also stainless steel weldments see also hydrogen absorption weld pool shape and geometry 22830 elliptical weld pool 229-30 tear-shaped weld pool 229-30 see also solidification of welds welding processes, definitions 24-5 see also arc welding processes wetting conditions 222-7, 242-9 interfacial energies 242-3, 247-9 wetting angle 225 see also solidification of welds Widmanstatten ferrite in low-alloy steels 427-8
601
Z Zener drag, definition of 341-42 in grain growth 341-42 Zener equation 342, 344 Zener-Hollomon parameter 465-67 zinc in aluminium solubility product of precipitates 303
V volume of weld metal 36 volume fraction of inclusions 193-5 volume heat capacity 3
W Wagner-Lifshitz equation 314,351 water content 137-8 in electrode coating 137 in welding flux 138
196-97,
602
Author Index
A
Aaron, H.B. 320, 326, 398-9 Aaronson, H.I. 408, 429 Abson, D.J. 407, 428, 440, 477-8, 485,493,504 Adams, C.M. 26 Adrian, H. 301, 303 Agren, J. 320-1, 326 Akselsen, a.M. 97, 345, 347, 349, 367,406,414-15,419,444,446, 448-54,481,484,488-90,4956,499,502-7,525 Alberry,~J.374,500,502 Alcock, C.B. 159 Ali, A. 427-8 American Society for Testing Materials (ASTM) 364 Andersen, I. .483 Anderson, M.P. 380 Anderson, P.D. 3 Ankem, S. 343, 351 Apold,A.174-5 Araki, T. 505-6 Ardell, A.J. 494 Ashby, M.F. 26, 201, 314, 318, 329, 360, 363-4, 375, 377-8, 459, 461,464 Asthana, R. 326 Atlas of isothermal transformation and cooling transformation diagrams 403 Avrami, M. 403, 422
Berge, J.O. 229 Bernstein, I.M. 512 Betzold, J. 413 Bhadeshia, H.K.D.H. 147,206,292, 408-9,413,422-9,431,433-4, 436,441,443-4 Bhatti, A.R. 208-10 Biloni, H. 229 Bjornbakk, B. 486, 491 Blander, M. 171, 173 Bolling, G.F. 290 Bonnet, C. 435, 440 Bradstreet, B.J. 186 Bramfitt, B.L. 244 Bratland, D.H. 459-62, 556-8, 562 British Iron and Steels Research Association 3 Brod~H.D.272,276 Brooks, J.A. 277-8, 533 Brown, A.M. 345 Brown, I.T. 509, 511 Brown, L.C. 314 Burck, P. 289 Burgardt, P. 229
Cochrane,R.C.292-3,407,428 Coe, F.R. 128-9,509-10,515 Coleman, M.C. 259,263-4 Collins, F.R. 537 Corbett, J.M. 203,428 Corderoy, D.J.H. 151, 155, 160-1 Cotton, HJ.U. 496 Cottrell, C.L.M. 496 Crafts, W. 190 Craig, I. 171 Cross,C.E. 251,259, 292-3,412,538 Crossland, B. 556 D Dallam, C.M. 441 D'annessa, A.T. 280 Das. G.452 Dauby, P. 180 David, S.A. 96, 99,105,210,222,228, 236, 239-41, 250, 260, 272, 278, 290,478 Davis, G.1. 221, 240, 247, 250, 278, 279,292,478 Davis, V. de L. 162 DeArdo, A.I 290 Deb, P. 434, 480 DebRoy, T. 210 Delong, W.T. 533 Demarest, V.A. 449-50 Devillers, L. 435, 440, 480, 482 Devletian, 1.H. 279,285,413 Dieter, G.E. 482, 486, 524-5, 529 Distin, P.A. 157 Doherty, R.D. 301, 309, 396 Dolby, R.E. 407,444 Dons, A.L. 438,459,541 Dorn, IE. 501-2 Dowling, 1M. 203,428 Dube, C.A. 408 Dudas, IH. 537 Dumolt, S.D. 547 E Eagar, T.W. 26, 96, 99, 105, 171-2, 174,228 Easterling, K.E. 26, 226, 247, 301, 303,309,314,318,345.360, 363-4, 367, 375, 377-8, 380, 389, 392, 403, 408, 427, 429, 435-6,441-4,448,500,502
B Babu, S.S. 210, 408, 443-4 Bach, H. 138 Bain, E.C. 408, 427, 436 Bakes, R.G. 15 Baldwin, W.M. 509,511 Balliger, N.K. 452 Bannister, S.R. 441, 443 Barbaro, FJ. 435-6, 441-4 Barin, I. 154 Barrie, G.S. 441,443 Barritte, G.S. 434-6, 441 Baskes, M.I. 277-8 Beachem, C.D. 512 Beaven, P.A. 440 Bell, H.B. 171,204-5 Bentley, K.P. 15
C Cahn, J.W. 337, 340-1, 345 Cai, X.-L. 450 Cameron, T.B. 413 Camping, M. 556-8, 562 Capes, J.F. 251, 292-3, 412 Carslaw, H.S. 2, 4 Chai, C.S. 171 Challenger, K.D. 434, 480 Chan, J.W. 403 Charpentier, F.P. 435, 440 Chen, J .H. 505-6 Chew, B. 132, 135 Chipman, J. 414 Choi, H.S. 450 Christensen, N. 24, 26-7,31-2, 80, 88,90-1,97,100,116,125,132, 143, 148-50, 153, 155, 158, 162, 170-1, 173-4, 176-9, 181-2, 186,189,193,207,345,347,349, 367,500,502,515-17,520-2 Christian, J.W. 329, 400-1, 403-4, 429,431 Cisse, J. 290 Claes, J. 180
AUTHOR INDEX Ebeling, R. 201 Edmonds, D.V. 409 Edwards, G.R. 227,259,422-3,425, 428,441 Eickhorn, F. 187-8 Elliott, J.F. 151, 162, 174-5, 179, 182, 184, 191 Engel, A. 187-8 Enjo, T. 547 Es-Souni, M. 440 European Recommendations for Aluminium Alloy Structures 552 Evans, G.M. 137-8, 192-3, 203, 420-1,435,440 Grewal, G. 343, 351 Griffiths, E. 3,4 Grong, 0, 26, 61, 73-5, 77, 80, 88, 90-2, 116, 149-50, 153, 155, 158, 161, 163-6, 170, 174, 176-9, 181-2, 185-6, 189, 192-204, 206-7, 209-11, 227, 247-8, 250-4, 256, 290, 292-4, 314, 327-30,345-7,349,355,360,364, 367-8, 371-2, 406, 412-15, 419, 422-3, 425, 428, 430-2, 435-6, 438,440-1,444,446-54,45~:, 464,465-6, 477-8, 480-1, 484tQ, 488-91, 493, 496, 502-7, 547-9, 551-8,560-3 Gunleiksrud, A. 503 Guo, Z.H. 405, 420-1
603
I Ibarra, S. 497 Indacochea, J .E. 171, 173 International Institute of Welding 129, 152 lon, J .C. 314, 318, 360, 363-4, 368 Ivanchev, I. 204-5
J
Jackobs, EA. 418 Jackson,C.E.89,99,100 Jaeger, J .C. 2, 4 Jaffe, L.D. 500 Janaf, ? 154 Jelmorini, G. 156 Jonas, J.J. 464 Jones, W.K.C. 374 Jordan,M.F. 143, 145,259,263-4 Joshi, Y. 96, 99, 105 Just, E. 413
F Fainstein, D. 320,398-9 Farrar, R.A. 435, 441, 443, 480 Fast, J.D. 513 Ferrar, R.A. 428, 435, 444, 478,485, 504 Fine, M.E. 389, 403 Fischer, W.A. 162 Fisher, D.J. 221, 234, 242,251,259, 261,265-6,270,274 Fleck, N.A. 422-3, 428, 441 Flemings, M.C. 221, 234, 242, 265, 272,275-6 Fortes, M.A. 374-5, 380-1 Fountain, R.W. 414 Fradkov, V.E. 380 Franklin, A.G. 195, 208 Fredriksson, H. 290 Frost, H.J. 380 Fruehan, RJ. 156 Fujibayashi, K. 146
G Garcia, C.1. 290 Garland, J. 505 Garland,J.G. 221, 240,247,250,278, 279,292-3,478 Garret-Reed, AJ. 450 Gergely, M. 501-2 Giovanola, B. 260 Gittos, N.E 544-5 Gjermundsen, K. 162, 516 Gjestland, H. 541 Gladman, T. 343-5,452, 479 Gleiser1M.151, 162, 174-5, 179, 191 Goldak, J .A. 515 Goolsby, R.D. 306 Greenwood, I.A. 15 Grest, G.S. 380 Gretoft, B. 147, 422-8 Grevillius, N.F. 182,185, 188
H Hftbrekke, T. 229 Halmoy, E. 151 Hannertz, N .E. 507 Harris, D.R. 414 Harrison, P.L. 428, 435, 444, 478, 485,504 Hatch, J.E. 3,458, 547 Hawkins, D.N. 208-10,435,440 Hazzledine, P.M. 342-3 Heckel, R.W. 326 Hehemann, R.F. 429, 452 Heile, R.E 154, 156-7, 169 Heintze, G.N. 244,247 Heiple, C.R. 229 Hellman, P. 339, 343-5 Hemmer, H. 371-2 Hilbert, M. 339, 343-5 Hill, D.C. 154, 156-7, 169 Hillert, M. 290 Hilty, D.C. 190 Hjelen, I. 195, 292,430-3, 438 Hocking, L.M. 196 Hollomon, J.H. 465, 500 Hollrigl-Rosta, E 413 Homma, H. 203, 444-5, 504-5 Hondros, E.D. 414 Honeycombe, R.W.K. 406, 408, 420, 429,431,444,447-8,453,486 Horii, Y. 187-8 Houghton, D.C. 303, 323 Howden,D.G. 141 Howell, P.R. 434-6, 441 Hu,H.337-8, 342-3, 345,430 Hultgren, R. 3 Hunderi,O.337,341-2,380
K Kaplan, D. 435, 440, 480, 482 Kasuya, T. 496 Kato, M. 233 Kawasaki, K. 380 Keene,B.J.96,99,105,228 Kelly, A. 548 Kelly, K.K. 3 Kern,A.380 Kerr,'H.W. 203, 247, 273, 290,428 Kiessling, R. 202-4 Kihara, H. 131, 133, 134 Kikuchi, T. 142-4 Kikuta, Y. 505-6 Kim, B.C. 444 Kim, I.S. 450 Kim, N.J. 444, 451, 505 Kim, Y.G. 451 Kinsman, K.R. 429,452 Kirkwood, P.R. 292-3 Kluken, A.O. 182, 186, 194-204, 206,209-11,247-8,250-4,256, 290,292-4,371-2,430-3,4356,438,440,446-7,479-80,484, 486,491,493,497 Knacke,O.154 Knagenhjem, H.O. 229 Knott, J.F. 486 Kobayashi, T. 142-4 Kotler, G.R. 320, 326, 398-9 Kou,S. 27, 75-6,96,99,105,228, 250,264-5,272,377,453,455, 458 Krauklis, P. 435-6, 441-4 Kraus, H.G. 228
604
METALLURGICAL
MODELLING
OF WELDING
Krauss, G. 418 Kubaschewski, 0. 159 Kuroda, T. 547 Kurz, W. 221, 234, Z42, 251, 259, 260-1,265-6,270,274 Kuwabara, M. 146 Kuwana, T. 142-4 Kvaale, P.E. 414-15, 419, 504
L Lancaster, J.E 118, 120, 162, 187 Lanzillotto, C.A.N. 452, 505 Le, Y. 27,75-6,265 Lee, D.Y. 444 Lee, J.-L. 504 Lei, T.C. 505 Li, W.B. 345, 367 Lifshitz, J .M. 314, 351 Lindborg, U. 170, 182-3, 185 Liu, J .Z. 505 Liu,S.251,292-3,412,422-3,42), 435-6,441,497 Liu, Y. 339 Loberg, B. 303 Lohne, 0. 380,459,541 Long, C.J. 533, 535 Lucke,K.337,340-1,345 Lutony, MJ. 380
Mills, K.C. 96, 99,105,228 Milner, D.R. 141 Miranda, R.M. 374-5 Mitra, U.171-2 Mizuno, M. 284 Moisio, T. 290 Mondolfo, L.P. 242-4 Mori, N. 187-8 Morigaki, O. 146 Morral, IE. 413 Mossinger, R. 554 Mundra, K. 210 Munitz,A.267 Murray, IL. 203 Muzzolani, EM. 547, 550, 552 Myers, P.S. 26 Myhr, O.R. 26, 61, 73-5, 77, 314, 327-30, 360, 458-62, 464, 496, 547-9,550-5 N Naess, OJ. 503 Nagai, T. 380 Nakagawa, H. 131, 133, 134 Nakata, K. 271 Nes,E.337,341-2,380 Nicholson, R.B. 548 Niles, R.W. 89 Nilles, P. 180 Nordgren, A. 303 North, T.H. 171 Nowicki, A. 171 Nylund, H.K. 438
p
Paauw, AJ. 446, 503 Pabi, S.K. 326 Pakrasi, S. 413 Pan, Y.-T. 504 Pande, C.S. 339 Pardo, E. 247, 273 Pargeter, R.J. 428, 440, 477-8, 485, 493,504 Patterson, B.R. 339 Pepe, I.J. 541 Petch, NJ. 512 Petty-Galis, I.L. 306 Phillip, R.H. 444 Phillips, H.W.L. 543 Pickering, EB. 301, 303, 452, 479, 505 Pitsch, W. 436-7 Plockinger, E. 186 Porter, D.A. 247,309,389,392,403, 408,413,427,429,435,448 Pottore, N.S. 290 Priestner, R. 451 Pugin, A.1. 556-7
M
Maitrepierre, Ph. 414 Marandet, B. 435, 440, 480, 482 Marder, A.R. 451 Marthinsen, K. 380 Martins, G.P. 182, 185-6, 192-3 Martukanitz, R.P. 459 Matlock, D.K. 193, 195, 201, 413, 422-3, 428, 432, 435, 440-1, 477-8,480,485,488,491,493, 504 Matsuda, E 131, 133, 134,233,271 Matsuda, S. 203, 208, 319, 444-5, 504-5 Matsuda, Y. 505-6 Matsunawa, A. 96, 99, 105 Mazzolani, EM. 458 McKnowlson, P. 15 McMahon, C.l. 418 McPherson, R. 244, 247 McQueen, H.J. 464 McRobie, D.E. 486 Mehl, R.E 408, 436-7 Metals Handbook 3,545 Midling, O.T. 465-6, 556-8, 560-3 Miller, R.L. 449-50 Mills, A.R. 435, 440
o
O'Brien, J.E. 143, 145 Odland, P.T. 480 Ohkita, S. 203, 444-5, 504-5 Ohno, S. 143 Ohshita, S. 103, 104,496,515 Ohta, S. 380 Okumura, M. 103, 104 Okumura, N. 208, 319 Olsen, K. 500, 502 Olson,D.L.171, 173-4, 176-9, 1812, 185-6, 192-3, 422-3, 428, 436,441,480,497,500,502 Onsoien, M.1. 479, 448-9, 495-6, 525 Oreper, G.M. 96, 99, 105,228 Oriani, R.A. 514 Orr, R.L. 3 Ostrom, G. 533 0verlie, H.G. 541 Owen, W.S. 450 Ozturk, B. 156
R Ramachandran, S. 190-1, 204 Ramakrishna, V. 151, 162, 174-5, 179,191 Ramberg, M. 450-1, 505-6 Ramsay, C. W. 480 Rappaz,M. 236, 239, 241,260 Rath, B.B. 337-8, 342-3, 345 Ravi Vishnu, P. 500, 502 Reif, W. 380 Reiso, O. 541-3 Reti, T. 501-2 Ribes, A. 435, 440, 480, 482 Riboud, P.V. 480, 482 Ribound, P.V. (Riboud ?) 435, 440 Ricks, R.A. 434-5, 436, 441 Ringer, S.P. 345, 367 Rollett, A.D. 380 Roper, I.R. 229 Rorvik, G. 247-8,250-4,256,2924,430,446-9,495-6,499,503, 507,525 Rose, R. 516 Rosenthal, D. 26, 28, 31, 33, 38,41, 48, 51, 56, 59-61, 76, 98, 133, 360 Roux, R. 140 Rykalin, N.N. 18,21,26,41,45,56, 93,95,556-7 Ryum, N. 326, 337, 341-2, 345,347, 349,367,380,382,390,396,403, 541-3
AUTHOR INDEX
605
W Wagner, C. 201,314,351 Wahlster, M. 186 WaIJhr'R.A. 190-1, 204 Walton, D.T. 380 Wa~H.96,99, 105,228 Weatherly, G.C. 303, 323 Welding Handbook 24 Welz, W. 21, 25 Whelan, M.1. 319,356 Whiteman, J.A. 208-10,435,440 Widgery, D J. 480-1 Willgoss, R.A. 132 Williams, J .C. 533 Williams, T.M. 414 Wolstenholme, D.A. 174 Woods, W.E. 279, 285 Womer, C.H. 342-3 Wriedt, H.A. 203
S
Saetre, T.O. 380,382 Saggese, M.E. 208-10,435,440 Sagmo,G.97 Saito, S. 103, 104 Sakaguchi, A. 284 Savage, W.E 541 Schaeffler, A.L. 533 Scheil,E.403 Schriever, U. 380 Schumacher, J.E 162 Schwan, M. 21, 25 Scott, M.H: 544-5 Seah, M.P. 414 Sea~E.B.96,99, 105 Senda, T.233 Shackleton, D.N. 166-7 Shaller, EW. 509, 512 Shen, H.P. 505 Shen, X.P. 451 Sherby, O.D. 501-2 Shercliff, H.R. 314, 329,~59-62, 464 Shinozaki, K. 131, 133, 134 Siewert, T.A. 182, 185-6, 192-3, 227,425,428 Sigworth, G .K. 162 Simonsen, T. 520-2 Sims, C.E. 512 Ska1and, T. 346 Skjolberg, E.M. 140-1 Slyozov, V.V. 314, 351 Smith, A.A. 166-7, 169, 170 Soares, A. 380-1 Solberg, J .K. 446, 450-4, 504-6 Sommerville, I.D. 204-5 Speich, G.R. 449-50 Srolovitz, DJ. 380 Staley, J.T. 394-5,459 Steidl, G. 554 Steigerwald, E.A. 509,512 Stjemdahl, J. 290 Stoneham, A.M. 414 Strangwood, M. 428-9, 431, 444 Strid,J.303,542-3 Sttiwe, H. 337,340-1,345 Sugden,A.A.B.292,431 Suutala, N. 290 Suzuki, H. 406,444,477,496,509, 515,520 Suzuki,S. 303, 323 Svensson, L.E. 147, 206,413,4228,431,433-4,441,444,536 Szeke1y,J. 96,99,105,120, 162, 18~, 187,228,281,284 Szewezyk, A.E 505 Szumachowski, E. 533
T
Takalo, T. 290 Tamehiro, H. 496 Tamura, I. 405,420-1 Tanigaki, T. 146 Tanzilli, R.A. 326 Tardy, P. 501-2 Tensi, H.M. 21, 25 Thau1ow, C. 503 Themelis, N.I 120, 162, 183, 187,281, 284 Therrien, A.E. 434, 480 Thewlis, G. 203, 435, 440-1 Thivellier, D. 414 Thomas, G. 505 Thompson,A.W. 480, 512 Thompson, C.V. 380 Tichelaar, G.W. 156 Tjotta, S. 459, 460 Tomii, Y. 284 Torsell, K.182-3, 185 Tricot, R. 414 Trivedi, R. 260, 400, 427 Troiano, A.R. 509, 512 Tsai, N.S. 26 Tsukamoto, K. 271 Tunda1, V.H. 326 Turkdogan, E.T. 126, 182, 184-6, 191-2,195-6,207,214 Turpin, M.L. 182, 184
y
Yamamoto, K. 203, 444-5, 504-5 Yang, J.R. 428-9 Yi, J.1. 450 Yoneda,M.505-6 Yurioka, N. 103, 104,496,509,515, 520
U Uda, M. 143 Udler, D.G. 380 tne, R.L. 96, 99, 105 Umemoto, M. 405,420-1 Underwood,E.E.201,338 Unstinovshchikov, J.1. 494
Z Zacharia,T.96,99,105,228 Zapffe, C.A. 512 Zener, C. 341-2, 344, 465 Zhang, C. 515 Zhang, Z. 441,443
V Van Den Heuvel, G.1.P.M. 156 Van Stone, R.H. 480 VanderVoort,G.F.394,403 Vandermeer, R.A. 341 Vasil' eva, V.A. 556-7 Verhoeven, J.D. 286,429,431,433, 448-9 Villafuerte, IC. 247, 273 Vitek, 1M. 96, 99, 105,210,222,228, 240,250,272,278,478