Sunteți pe pagina 1din 104

Accepted Manuscript

Title: An Overview of Direct Laser Deposition for Additive


Manufacturing; Part II: Mechanical Behavior, Process
Parameter Optimization and Control

Author: Nima Shamsaei Aref Yadollahi Linkan Bian Scott M.


Thompson

PII: S2214-8604(15)00032-9
DOI: http://dx.doi.org/doi:10.1016/j.addma.2015.07.002
Reference: ADDMA 43

To appear in:

Received date: 23-12-2014


Revised date: 7-5-2015
Accepted date: 20-7-2015

Please cite this article as: Shamsaei N, Yadollahi A, Bian L, Thompson SM,
An Overview of Direct Laser Deposition for Additive Manufacturing; Part II:
Mechanical Behavior, Process Parameter Optimization and Control, Addit Manuf
(2015), http://dx.doi.org/10.1016/j.addma.2015.07.002

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
*Manuscript

1
2
3
4
5 An Overview of Direct Laser Deposition for Additive Manufacturing;
6
7 Part II: Mechanical Behavior, Process Parameter Optimization and
8
9 Control
10
11
12

t
13

ip
14
15
16

cr
17 Nima Shamsaei1,2†, Aref Yadollahi1, Linkan Bian3, Scott M. Thompson1,2
18
19 1
Department of Mechanical Engineering, Mississippi State University, MS 39762

us
20 2
21 Center for Advanced Vehicular Systems (CAVS), Mississippi State University, MS
22 39762
23 3
Department of Industrial and Systems Engineering, Mississippi State University, MS
24

an
25
39762
26
27

28 Corresponding author:
M
29 Department of Mechanical Engineering, Box 9552, Mississippi State University
30
31 Mississippi State, MS 39762, USA
32 Email: shamsaei@me.msstate.edu
33 Phone: (662) 325 2364
d

34
35
te

36
37
38
p

39
40
41
ce

42
43
44
45
Ac

46
47
48 Submitted to the Journal of:
49 Additive Manufacturing
50
51
52
53
54 Revised Paper Submission: May 2015
55
56 Original Submission: December 2014
57
58
59
60
61
62
63
64 Page 1 of 103
65
1
2
3
4 Abstract
5
6
7 The mechanical behavior, and thus the ‘trustworthiness’/durability, of engineering
8
9 components fabricated via laser-based additive manufacturing (LBAM) is still not well
10
11
12 understood. This is adversely affecting the continual adaption of LBAM for part

t
13

ip
14 fabrication/repair within the global industry at-large. Hence, it is important to determine
15
16
the mechanical properties of parts fabricated via LBAM as to predict their performance

cr
17
18
19 while in service. This article is part of two-part series that provides an overview of Direct

us
20
21 Laser Deposition (DLD) for additive manufacturing (AM) of functional parts. The first part
22
23
24 (Part I) provides a general overview of the thermo-fluid physics inherent to the DLD

an
25
26 process. The objective of this current article (Part II) is to provide an overview of the
27
28
mechanical characteristics and behavior of metallic parts fabricated via DLD, while also
M
29
30
31 discussing methods to optimize and control the DLD process. Topics to be discussed
32
33
d

34 include part microstructure, tensile properties, fatigue behavior and residual stress –
35
te

36 specifically with their relation to DLD and post-DLD process parameters (e.g. heat
37
38
treatment, machining). Methods for controlling/optimizing the DLD process for targeted
p

39
40
41 part design will be discussed – with an emphasis on monitored part temperature and/or
ce

42
43 melt pool morphology. Some future challenges for advancing the knowledge in AM-part
44
45
Ac

46 adaption are discussed. Despite various research efforts into DLD characteristics and
47
48 process optimization, it is clear that there are still many areas that require further
49
50
51 investigation.
52
53
54 KEYWORDS: Direct Laser Deposition (DLD); Additive Manufacturing (AM);
55
56 Microstructure; Fatigue Behavior; Process Control
57
58
59
60
61
62
63
64 Page 2 of 103
65
1
2
3
4 NOMENCLATURE
5
6 ∆𝐾 Stress intensity factor ranges, MPa√m
7 a Crack length, µm
8
9 da/dN Crack extension per cycle, mm/cycle
10 C Specific heat, J/kg∙K
11 G Temperature gradient at solid-liquid interface, K/m
12 GS Smallest grain size, µm

t
13
h Height of substrate, mm

ip
14
15 JIc J-integral fracture toughness, KJ/m2
16 k Number of functional dimensions

cr
17 K Stress intensity factor, MPa√m
18 KI Stress intensity factor in Mode I, MPa√m
19
KIC Plane-strain fracture toughness, MPa√m

us
20
21 Kmax Maximum stress intensity, MPa√m
22 Kt System gain
23 l Melt pool length, mm
24

an
L Latent heat of fusion, J/kg
25
26 M Powder feed rate, g/min
27 MS Characteristic length scale for microstructural interaction
28 𝑁sp Non-dimensional specific energy index
M
29
N Physical variables
30
31 Nf Number of cycles to failure
32 NInc Number of cycles to incubate a crack
33 NLC Number of cycles required for long crack propagation
d

34 NMSC Number of cycles required for propagation of a microstructurally small


35
crack
te

36
37 NPSC Number of cycles required for propagation of a physically small crack
38 NTotal Total fatigue life
p

39 P Dimensionless parameters
40 R Solidification rate
41
ce

42 Rs Stress ratio or strain ratio


43 Ra Roughness, µm
44 Q Laser beam radius, m
45 rb Laser power, W
Ac

46
47
S Nominal stress, MPa
48 s Transformed coordinate
49 Ta Ambient temperature, K or ºC
50 T Temperature, K or ºC
51 TL Liquidus temperature, ᵒC
52
53 V Traverse speed, mm/s
54 α Fitting parameter for Eq. (3)
55 β Fitting parameter for Eq. (3)
56 γ Fitting parameter for Eq. (3)
57
58
Λ Sphericity coefficient
59 𝜀 Local strain
60 τ Time constant, s
61
62
63
64 Page 3 of 103
65
1
2
3
4
5
6 GLOSSARY
7 3D Three Dimensional
8
9 AC Air Cooling
10 AM Additive Manufacturing
11 AI Artificial Intelligence
12 BCC Body Centered Cubic

t
13 CAD Computer Aided Design

ip
14
15 CT Compact Tension
16 CTOD Crack Tip Opening Displacement

cr
17 DCS Dendrite Cell Size
18 DLD Direct Laser Deposition
19
DOE Design of Experiment

us
20
21 FC Furnace Cooling
22 FCGR Fatigue Crack Growth Rates
23 FFNN Feed Forward Neural Network
24

an
HAZ Heat-Affected Zone
25
26 HCF High Cycle Fatigue
27 HCP Hexagonal Close Packed
28 HIP Hot Isostatic Pressing
M
29 LBAM Laser-Based Additive Manufacturing
30
31
MPLF Metal Powder Laser Forming
32 NND Nearest Neighbor Distance
33 PID Proportional, Integral, and Derivative
d

34 PSC Physically Small Crack


35 SLM Selective Laser Melting
te

36
37
38
p

39
40
41
ce

42
43
44
45
Ac

46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 4 of 103
65
1
2
3
4 1. Introduction
5
6
7
8 Direct Laser Deposition (DLD) is a type of laser-based additive manufacturing (LBAM)
9
10
11
process for fabricating metallic, functional parts. In contrast to Selective Laser Melting
12

t
13 (SLM), which utilizes a bed of powdered metal that is ‘selectively’ melted via a laser, DLD

ip
14
15 is accomplished by simultaneously delivering metallic powder (or wire) and focused laser
16

cr
17
18 energy. In DLD, a relatively high powered laser (e.g. Nd:YAG or CO2) is utilized to create
19

us
20 a melt pool (or molten pool) atop the surface of a substrate within an inert atmosphere (e.g.
21
22
23 argon); simultaneously, powder is injected/blown through the laser beam and into the melt
24

an
25 pool. Using a sliced 3D CAD (computer aided drawing) file, parts are then built layer by
26
27
28
layer, with each layer assembled track by track via a user-defined tool path. The DLD
M
29
30 process can involve multiple nozzles or a single nozzle for the blown powder deposition;
31
32 for example, a common technology for accomplishing multi-nozzle, blown-powder DLD
33
d

34
35 is Laser Engineered Net Shaping (LENS) [1]. Table 1 demonstrates that, in addition to
te

36
37 LENS, there are many other aliases and commercialized technologies related to DLD [2–
38
p

39
40 23]. Further details on the development and current variations on DLD are provided in
41
ce

42 Part I.
43
44
45 Direct Laser Deposition is unique in that it can be used for the additive
Ac

46
47 manufacturing (AM) of functionally-graded and pure-metal parts, as well as for laser
48
49
cladding/repair. It also has the potential to produce many materials to exact dimensions for
50
51
52 aerospace, medical, or military industries. Some common materials that have been
53
54 investigated for DLD include: titanium alloys [24–34], steels [6,18,35–41], nickel base
55
56
57 superalloys [40,42–47], cobalt base alloys [48–50], aluminum, and copper alloys [51].
58
59 Grylls [52] has specifically listed the following alloys from these material systems as
60
61
62
63
64 Page 5 of 103
65
1
2
3
4 compatible with the DLD process: titanium alloys (e.g. Ti-22Al-23Nb, Ti-48-2-2, TiC),
5
6
7 steels (e.g. 10V, 15-5 PH, 410, 416, AISI 309, Aermet 100, A2, MM 10, CPM S7), nickel
8
9 base superalloys (e.g. CMSX-3, Haynes 188, Haynes 230, IN600, IN690, IN713,
10
11
12 MarM247, Rene 142, Rene N5), aluminum alloys (e.g. CP Al, 6061, 2024) and copper

t
13

ip
14 alloys (e.g. Cu-10%Sn and GRCop-84). DLD is also capable of utilizing different
15
16
feedstock (preform) materials simultaneously to produce various alloys and functionally-

cr
17
18
19 graded materials (FGMs) [14,53–55]. Nozzle feed rates can be regulated to generate

us
20
21 materials with distinct microstructural properties and chemical compositions [56].
22
23
24 Direct Laser Deposition also provides high deposition rates and a relatively wide

an
25
26 process window to fabricate larger items relative to other metal-based AM methods [57].
27
28
Moreover, DLD provides for a relatively small heat affected zone (HAZ) on the part during
M
29
30
31 processing, excellent density and metallurgical bonding, minimal effect on the component
32
33
d

34 (e.g. distortion, micro-cracking) and precise deposition – making it a great tool for
35
te

36 repairing high-value components [56]. Repair of damaged turbine blades [58–60] made
37
38
from nickel base superalloys and engine cylinder heads and blocks [61] are two examples
p

39
40
41 of such repairs. Some current disadvantages of utilizing DLD for AM or repair is the
ce

42
43 relatively low powder efficiency and rough, post-DLD surface finishes [57]. As the field
44
45
Ac

46 of LBAM is evolving very quickly, there have been several in-depth literature review
47
48 papers and book chapters on this subject [57,62–66].
49
50
51 The temperature of parts during DLD will have a direct impact on their
52
53 microstructural homogeneity and post-fabricated mechanical characteristics. As discussed
54
55
56
in Part I, this has motivated a large number of studies into DLD process modeling and
57
58 monitoring. The aim of this article, Part II, is to provide an overview of the microstructural
59
60
61
62
63
64 Page 6 of 103
65
1
2
3
4 features and post-manufactured characteristics of DLD parts, as well as process parameter
5
6
7 optimization and the monitoring/control of the DLD process. First, the effect of thermal
8
9 history, process parameters, and part geometry on solidification parameters and
10
11
12 microstructure evolution during the DLD process are described. Next, the post-

t
13

ip
14 manufacturing characteristics, in particular, tensile and fatigue behavior of DLD parts are
15
16
presented. Since less attention has been given to fatigue properties of DLD parts, a special

cr
17
18
19 effort is made to focus on the literature in this area. Some process control parameters and

us
20
21 design patterns on the fatigue behavior of DLD parts are discussed. The capability of
22
23
24 various data-driven methods that aim at optimizing DLD processes is then reviewed while

an
25
26 also focusing on thermal monitoring/control. Finally, some conclusions and ongoing
27
28
research areas in the post-manufacturing characterizations and process parameter
M
29
30
31 optimization are provided.
32
33
d

34 Although the breadth of this article (Part II) is relatively broad, it is by no means
35
te

36 comprehensive, as discussion on effects of powder size and morphology, multi-material


37
38
processing, etc. are not included. The goal of this paper is not to encompass all the existing
p

39
40
41 studies related to DLD but to provide an overview for researchers to understand important
ce

42
43 and imminent research problems/opportunities in this area. The reader is encouraged to
44
45
Ac

46 reference Part I for a more in-depth review and description of the DLD process and
47
48 accompanying thermo-fluidic transport phenomena.
49
50
51
52 2. Microstructure and Process Parameters
53
54
55 2.1. Microstructure
56
57
58
59
60
61
62
63
64 Page 7 of 103
65
1
2
3
4 The microstructural characteristics (e.g. morphology and grain size) of DLD parts
5
6
7 are strongly sensitive to their thermal history during the build, which may include high
8
9 heating/cooling rates, significant temperature gradients, bulk temperature rises and more.
10
11
12 Since many process variables/parameters impact the thermal history, predicting the

t
13

ip
14 microstructural features of DLD parts, and the degree of their dependence on process
15
16
parameters, is still a major challenge. However, overcoming this challenge is vital for

cr
17
18
19 establishing the effective control mechanisms for fabricating DLD parts with superior

us
20
21 mechanical properties. Various authors have investigated the effects of certain parameters
22
23
24 on the microstructural characteristics and material properties of DLD parts with specific

an
25
26 shapes [13,24,67–70]. However, it is still unclear how to apply these findings to fabricate
27
28
complex parts with various shapes since their microstructures will have a unique
M
29
30
31 dependence on thermal history.
32
33
d

34 The solidified microstructure depends on: local solidification rates within the melt
35
te

36 pool, the ratio of cooling rate to thermal gradient, R, and the temperature gradient at the
37
38
solid-liquid interface, 𝐺. Two critical solidification parameters are the ratio, 𝐺/𝑅, which
p

39
40
41 affects the solid-liquid interface shape, and the cooling rate, 𝐺 ∗ 𝑅 , which affects
ce

42
43
microstructure dimensions [57,71]. Different G and R values may result in three major
44
45
Ac

46 structure morphologies within DLD parts: columnar (elongated grain morphology),


47
48 columnar-plus-equiaxed, and equiaxed (isotropic grain morphology). It has been found that
49
50
51 a higher solidification rate promotes the transition from columnar to equiaxed grain
52
53 morphologies [68] and that increasing the cooling rate, 𝐺 ∗ 𝑅 , leads to a finer
54
55
56 microstructure. The tendency to form a columnar structure increases by increasing the ratio
57
58 𝐺/𝑅, while decreasing 𝐺/𝑅 is favorable for equiaxed structures [57]. For thin wall parts
59
60
61
62
63
64 Page 8 of 103
65
1
2
3
4 fabricated via LENS, the ranges for G and 𝐺 ∗ 𝑅 are approximately 100-200 K/mm and
5
6
7 200-6,000 K/s, respectively [72,73]. Generally, cooling rates of 103 to 104 K/s are
8
9 reported for obtaining desired microstructural and mechanical properties via DLD
10
11
12 [69,74,75]. The optimal G and R values also depend on part geometry,

t
13

ip
14 environmental/machine conditions, material properties and other process parameters [57].
15
16

cr
17
18
19 2.1.1. Micro-hardness

us
20
21
22
Distinct microstructure regions with fairly different micro-hardness have been
23
24 reported for DLD AISI P20 tool steel [76], AISI H13 tool steel [6,13,77,78], AISI 420 tool

an
25
26 steel [77], AISI 4140 steel [37], AISI 410 stainless steel (SS) [79], AISI 316L stainless
27
28
steel [75], Inconel 625 [45], and Ti-6Al-4V [81] components. It has been reported that for
M
29
30
31 steels, the micro-hardness values of subsequent deposited layers decreases from the first
32
33
d

34 deposited layer and then increases towards top layers [37,75,81]. This inhomogeneity can
35
te

36 be attributed to the time-variable cooling rate of the melt pool and a relatively slower
37
38
velocity of solidification in the middle region. The middle region is also exposed to cyclic
p

39
40
41 reheating (related to subsequent layer deposition) and experiences a substantial heat
ce

42
43
44
affected zone (HAZ) for a longer period of time. As a result, higher micro-hardness
45
Ac

46 generally has been measured/observed at the top and bottom of DLD parts which undergo
47
48 higher cooling rates during the DLD process as compared to the middle region. For
49
50
51 instance, in laser-deposited Ti-6Al-4V parts, the sizes of the α and β laths (e.g. a thin and
52
53 narrow strip of crystalline structure) vary with location [57,80]. The microstructure in both
54
55
56 the top and bottom of the deposited Ti-6Al-4V part consists of the Widmanstätten structure;
57
58 however, the morphology is different in each region. The top region primarily displays
59
60
61
62
63
64 Page 9 of 103
65
1
2
3
4 colonies of parallel, very fine lamellae, and larger laths, as presented in Fig. 1(a), whereas
5
6
7 the bottom region has thicker lamellae, as shown in Fig. 1(b) [80]. These differences in Ti-
8
9 6Al-4V microstructure can be attributed to the temperature gradients, cooling rates and
10
11
12 repeated heat treatments during subsequent laser-deposition steps.

t
13

ip
14
15
16
2.1.2. Residual stress

cr
17
18
19 Residual stress is defined as the “stress in a body which is at rest and in equilibrium

us
20
21 and at uniform temperature in the absence of external and mass forces [82].” The dynamic
22
23
24 temperature distribution and heating/cooling rates along the part during DLD – which

an
25
26 consists of high thermal gradients and repetitious/rapid local heat transfer rates - are known
27
28
to cause residual stresses in DLD parts [83,84]. The thermal history during the DLD
M
29
30
31 process can result in the establishment and evolution of an anisotropic microstructure and
32
33
d

34 the presence of residual stresses throughout the part. Other material properties such as
35
te

36 tensile and fatigue resistance are directly related to the microstructural features of the part
37
38
and can be adversely affected in the presence of tensile residual stresses. Material
p

39
40
41 properties (e.g. the thermal conductivity, CTE, elastic modulus, and yield stress), phase
ce

42
43 transformation, part geometry, process parameters during fabrication, and scanning pattern
44
45
Ac

46 all influence the magnitude and pattern of the residual stress within the fabricated part
47
48 [83,85]. The presence of residual stresses can reduce the strength or life of mechanical parts
49
50
51 and can also result in dimensional inaccuracies due to warping [85,86]. The magnitude of
52
53 local residual stresses can be a large percentage of the nominal yield strength of the
54
55
56
material; for example, a maximum of 75% nominal yield strength [83].
57
58
59
60
61
62
63
64 Page 10 of 103
65
1
2
3
4 The material stress–strain relationship and strain mismatch during the cooling
5
6
7 phase affects the magnitude of resultant residual stresses during the DLD process.
8
9 Materials with higher Young’s modulus (modulus of elasticity) and yield stress usually
10
11
12 foster the formation of residual stresses with higher magnitude. For instance, residual

t
13

ip
14 stresses within DLD Inconel 718 specimens (yield strength of 1,100 MPa) have been
15
16
measured to be approximately 1.5 times larger than DLD AISI 316 SS specimens (yield

cr
17
18
19 strength of 450 MPa) with similar elastic modulus and CTE at room temperature [83]. The

us
20
21 yield stress–temperature curve of a material system is another important factor with regard
22
23
24 to residual stress formation. The strain mismatch will be low at high temperatures for some

an
25
26 materials (e.g. AISI 316 SS) since the yield stress decreases rapidly with increasing
27
28
temperature. Further, these thermally-produced volume changes can be accommodated by
M
29
30
31 plastic flow [83]. Therefore, materials with high yield strength at elevated temperatures,
32
33
d

34 such as superalloys (e.g. Inconel 718) produce greater residual stresses [83].
35
te

36 According to Rangaswamy et al. [83], the highest residual stress value, at least for
37
38
DLD AISI 316 SS, is compressive and occurs along the build direction (normal to the build
p

39
40
41 plate). Generally, by increasing the build height, the level of residual stresses increases
ce

42
43 [57]. Studies on various metals, such as H13 tool steel [87], Ti-6Al-4V [88], Inconel 718
44
45
Ac

46 and AISI 316 SS [83] demonstrate that the residual stresses within the building plane are
47
48 typically aligned with the laser scanning direction; being compressive at the center and
49
50
51 tensile at the edges of the part. Furthermore, residual stresses are lower at the starting
52
53 location of laser scanning and reach their maximum value at the end of the laser scanning
54
55
56
path [88]. Compressive residual stresses are larger near the substrate and transition to
57
58
59
60
61
62
63
64 Page 11 of 103
65
1
2
3
4 tensile stresses with lower magnitudes towards the top of parts as the number of layers
5
6
7 increases [88].
8
9 The presence of residual stress and distortion during DLD can be reduced by
10
11
12 weakening local thermal gradients (and thus increasing part isothermality). This can be

t
13

ip
14 accomplished by optimizing and controlling the process parameters (e.g., laser power and
15
16
beam traverse speed), pre-heating the substrate and previously-deposited layers,

cr
17
18
19 maintaining optimal melt pool morphology/size and selecting a proper laser scanning

us
20
21 pattern [83,85,86]. Manipulating the process parameters to maintain constant melt pool
22
23
24 size and temperature throughout the part, especially at the corners, leads to reduction in

an
25
26 residual stresses [83,86].
27
28
M
29
30
31 2.2. Effects of Process Parameters
32
33
d

34 There are many DLD process parameters that affect the thermal history (and thus
35
te

36 microstructure, residual stress, etc.) of the part, such as: powder feed rate, laser power,
37
38
laser/substrate relative traverse speed and laser scanning strategy. As discussed in Part I,
p

39
40
41 these parameters affect the melt pool shape and incident energy, and consequently, the
ce

42
43 cooling rate and local thermal gradients. Localized heat transfer, especially in the HAZ,
44
45
Ac

46 can result in layer re-melting and numerous reheating cycles; both which can change the
47
48 microstructure by tempering and aging the part.
49
50
51
52 2.2.1. Laser parameters
53
54 The combination of higher traverse velocity and lower laser power results in lower
55
56
57 incident energy at the top of the part, typically resulting in finer microstructures due to
58
59 higher cooling rates. In contrast, lower cooling rates and coarser microstructures can be
60
61
62
63
64 Page 12 of 103
65
1
2
3
4 garnered by decreasing traverse speed and increasing laser power [57,68,69]. Although
5
6
7 material-type undoubtedly influences grain morphologies of DLD parts, lower incident
8
9 energy, which can be due to laser attenuation and/or radiation effects (see Part I), tends to
10
11
12 result in finer equiaxed structures while higher incident energy generally results in

t
13

ip
14 columnar grains and coarser microstructures [57]. For laser-deposited, thin walls of Ti-
15
16
6Al-4V, grain size increases more by increasing the incident energy (e.g. decreasing laser

cr
17
18
19 traverse speed and/or increasing laser power) [68,70]. Increasing powder feed rate results

us
20
21 in a coarser microstructure (e.g. an increase in length and width of both α and β laths) for
22
23
24 Ti-6Al-4V; while the feed rate has fewer effects at higher laser powers. Faster traverse

an
25
26 velocities result in slight decreases in the size of α and β laths and more porosity. Similar
27
28
effects of process parameters on the microstructural properties have been reported for burn-
M
29
30
31 resistant Ti alloy (Ti-25V-15Cr-2Al-0.2C), Waspalloy [89], and Inconel 625 [45]. For
32
33
d

34 burn-resistant Ti alloy with mainly equiaxed grains, less dependency of microstructure on


35
te

36 process parameters has been observed. Increasing the feed rates during manufacturing of
37
38
Waspalloy also results in higher density of precipitates due to the larger interdendritic area
p

39
40
41 resulting from finer grain sizes [45,57,89].
ce

42
43 During the DLD process, the majority of sensible incident energy is transferred via
44
45
Ac

46 conduction through the deposited structure [32]. Heat is quickly conducted away by the
47
48 substrate at the bottom of the sample, whereas convection and radiation become more
49
50
51 prevalent at the top [57]. Since relatively high cooling rates occur near the substrate or
52
53 through previously-deposited layers, highly directional columnar structures are typically
54
55
56
formed. These columnar structures can extend across the deposited layers indicating
57
58 epitaxial growth of dendrites from the substrate or previously deposited layers [57]. The
59
60
61
62
63
64 Page 13 of 103
65
1
2
3
4 slower cooling rates in the middle of the DLD part favor the formation of secondary
5
6
7 dendrites since the thermal cycling of the previously-solidified material generally results
8
9 in slower cooling rates and a higher 𝐺/𝑅 [57]. These thermal cycles can also activate a
10
11
12 variety of metallurgical phenomena responsible for progressively modifying

t
13

ip
14 microstructural properties such as grain size [56]. Variation in the thermal history
15
16
experienced at different locations along a DLD part causes heterogeneous behavior in the

cr
17
18
19 microstructure which affects other mechanical properties such as tensile strength and

us
20
21
22
fatigue resistance [90]. The macrostructure of laser-deposited Ti-6Al-4V parts has
23
24 columnar prior-beta grains elongated along the solidification direction as presented in Fig.

an
25
26 2(a). The microstructure includes very fine Widmanstätten platelet alpha (basket weave or
27
28
colony structures of hexagonal-close-packed (HCP) α phase) lamellae in β matrix (body-
M
29
30
31 centered-cubic (BCC)), as shown in Fig. 2(b), which is due to rapid cooling rates during
32
33
d

34 the solidification stage [32,51].


35
te

36 Powder feed rate has an immediate impact on the distribution of powder density in
37
38
melt pool (deposited mass flow rate) [91] and thus layer height and microstructures during
p

39
40
41 DLD. Liu and Dupont demonstrated a linear increase in the layer height as the powder feed
ce

42
43 rate increases, which in turn results in a coarser microstructure [92]. For a fixed powder
44
45
Ac

46 feed rate, the amount of powder that is injected into the melt pool varies for different laser
47
48 scanning directions because of the distance between the powder stream and laser spot.
49
50
51 Depending on scanning direction, the powder injection point may be ahead or behind the
52
53 laser spot, as shown in Fig. 3. This can result in a highly asymmetric melt pool – thus
54
55
56 impacting the boundary and solidification heat transfer and thus clad height. As shown in
57
58 Fig. 3, the relative locations between the powder injection point and laser beam, and thus
59
60
61
62
63
64 Page 14 of 103
65
1
2
3
4 the quantity of powder melted, affects the amount of powder injected into the melt pool
5
6
7 [91]. The shadow area shown in Fig. 3 represents the overlap between the powder stream
8
9 and laser irradiation area. It may be seen that the overall shape of the melt pool is elongated
10
11
12 and consists of no rotational symmetry; however, there is a degree of symmetry in the YZ

t
13

ip
14 plane. When the powder injection point (A) is ahead of the laser spot (O), the amount of
15
16
powder injected into the melt pool is less than the scenario in which the powder injection

cr
17
18
19 point is slightly behind the laser spot.

us
20
21 To maintain a constant deposited mass flow rate, the powder feed rate and laser
22
23
24 scanning velocity can be adjusted based on scan direction and the distance between laser

an
25
26 beam and powder nozzle (other methods described in Section 4). However, there are
27
28
currently no quantitative studies that suggest the optimal control parameters, such as levels
M
29
30
31 of powder feed rate and laser scanning velocity due to complexity associated with the
32
33
d

34 underlying process mechanics. Adding to this complexity is the dynamics of the DLD
35
te

36 process, as the relative positions of the laser beam and powder injection point change as
37
38
parts become taller due to deposition. Therefore, any process optimization and control
p

39
40
41 scheme should be designed in a time-varying fashion that accounts for the height of
ce

42
43 deposited layers. The flow rate of shielding gas during the DLD process also affects the
44
45
Ac

46 amount of powder injected into the melt pool. Experimental studies demonstrate that as the
47
48 gas flow rate increases, the powder density and the layer height first increase and then
49
50
51 decrease since the shielding gas improves the amount of powder injected into melt pool,
52
53 while the particle velocity can impact its potential to reflect off the melt pool surface [6].
54
55
56
57
58
59
60
61
62
63
64 Page 15 of 103
65
1
2
3
4 2.2.2. Deposition Patterns
5
6
7 In general, there are four common deposition patterns (raster, bi-directional, offset,
8
9 and fractal patterns) for use in AM [93]. In addition to these deposition patterns, additional
10
11
12 ‘offset patterns’ can be employed by indicating direction, such as: offset-out and offset-in.

t
13

ip
14 These offset patterns depend on the starting point of deposition, as depicted in Fig. 4.
15
16
Different deposition patterns significantly affect the geometric and mechanical properties

cr
17
18
19 of the fabricated parts [85,86]. Choosing proper scanning patterns reduces the production

us
20
21 of residual stresses and any thermal distortion. Dai and Shaw indicate that the offset-out
22
23
24 pattern can reduce the out-of-plane distortion to one-third of the one induced by bi-

an
25
26 directional scanning pattern, as presented in Fig. 4 [85].
27
28
The raster pattern is the most commonly used deposition pattern mainly due to its
M
29
30
31 ease of implementation. As shown in Fig. 4, the laser scanning path of the raster pattern
32
33
d

34 does not depend on the shape of the fabricated part, and thus can be implemented to
35
te

36 fabricate a part of any shape [93]. Nickel et al. [94] found that a deposition pattern with
37
38
lines oriented 90º from the substrate’s longer axis reduces part deflection.
p

39
40
41 Additional research is needed to determine customized deposition paths for
ce

42
43 individual parts fabricated using the offset or fractal patterns. This is challenging for a part
44
45
Ac

46 with complex geometry. Nevertheless, the fractal and offset patterns have advantages as
47
48 they can provide for geometric accuracy and less energy consumption. Nickel et al. [94]
49
50
51 studied the effects of the raster, offset, and fractal deposition patterns on the geometric
52
53 accuracy of the fabricated part. They found that the fractal and offset-out deposition
54
55
56
patterns generate the smallest substrate deformations. Square corners can lead to the offset
57
58
59
60
61
62
63
64 Page 16 of 103
65
1
2
3
4 and fractal deposition patterns causing interior defects; therefore, adopting fractal or offset
5
6
7 patterns with arc paths can produce parts with higher quality.
8
9
10
11
12 2.2.3. Layer Slicing Strategy

t
13

ip
14 In conventional DLD processes, the part CAD is sliced into parallel layers and
15
16
produced in a layer-wise manner. Parts are built to completion layer by layer, from bottom

cr
17
18
19 to top. Due to the finite thickness of a deposition layer, surfaces whose normal vectors do

us
20
21 not make an angle of either 0º or 90º with the build direction can only be deposited
22
23
24 approximately. The designed part shape (e.g., curved surface) is typically approximated by

an
25
26 a series of parallel layers, as shown in Fig. 5. Also, the material shrinkage of a single layer
27
28
during DLD creates slanted walls around the layer itself while causing previously-
M
29
30
31 deposited layers to shrink (via “dragging”), as well. These combined effects of geometric
32
33
d

34 approximation and layer dragging generates a surface with step-like features. This
35
te

36 culmination is known as the “staircase effect” [95]. The magnitude of the staircase effect
37
38
is dependent on the layer thickness and the angle made by the surface normal with the build
p

39
40
41 direction. The staircase effect results in poor surface quality and requires additional
ce

42
43 machining or other post-manufacturing processing to form the desired shape, and in turn
44
45
Ac

46 increases the overall process time.


47
48 Research on slicing procedure such as controlling layer thickness [96] or the
49
50
51 volumetric difference between layers [97], attempts to reduce staircase effects. However,
52
53 these approaches are designed for fixed direction deposition processes and do not
54
55
56
completely eliminate the staircase effect. Multi-axis processing is another alternative to
57
58 mitigate the staircase effect, and various methods have been presented recently. Instead of
59
60
61
62
63
64 Page 17 of 103
65
1
2
3
4 adopting the traditional parallel slicing approach, multi-axis processing rotates the slicing
5
6
7 direction 90º when an overhang structure occurs [98,99]. An issue related to multi-axis
8
9 processing is that collision may occur when rotating the part orientation by 90º. For
10
11
12 example, for the part shown in Figs. 6 (a) and (c), collision between the powder nozzle and

t
13

ip
14 the deposited part occurs at the deposition of part (4), when the deposited part is rotated
15
16
90º

cr
17
18
19 To address the aforementioned challenges in layer slicing strategies, Ruan et al.

us
20
21 [100] presented a method of non-parallel layer slicing for more precise deposition along
22
23
24 the part geometry. In this method, the slicing direction is rotated only as needed (instead

an
25
26 of turning the slicing direction 90º), resulting in layers with non-uniform thickness for
27
28
better fitting shapes. In other words, the thickness varies at different locations. Although
M
29
30
31 adopting non-parallel layers helps in eliminating the staircase effect and in decreasing
32
33
d

34 manufacturing times, it imposes unprecedented challenges on implementation because


35
te

36 process control parameters must be delicately controlled so that the targeted layer thickness
37
38
can be achieved. This is further compounded by the fact that no analytical models are
p

39
40
41 available to characterize the relation between process parameters and layer thickness, as
ce

42
43 well as how changes in the former affect the latter. Current non-parallel layer slicing
44
45
Ac

46 strategies rely on empirical models to predict process parameters in advance in order to


47
48 achieve varying layer height. Linear empirical models have been used to correlate layer
49
50
51 height and laser scanning speed while fixing the laser power and the powder feed rate
52
53 [100]. Such empirical models are usually developed under specific experimental conditions
54
55
56
(e.g., single track deposition near the substrate); thus, the resulting prediction of the layer
57
58 height may not be accurate when the experimental conditions vary.
59
60
61
62
63
64 Page 18 of 103
65
1
2
3
4
5
6
7 3. Mechanical Characteristics
8
9
10 Post-manufacture characterization of parts is an inevitable and important procedure
11
12

t
13 for many design and manufacturing processes as it allows one to evaluate whether a

ip
14
15 material or part is suitable for its intended usage. Different mechanical tests, including:
16

cr
17
18 tensile, hardness, fracture toughness, impact resistance, and fatigue, are commonly used to
19

us
20 evaluate the mechanical performance of a part. Many failures have been attributed to cyclic
21
22
23
fracture or excessive deformation. Therefore, tensile and fatigue tests of a material are
24

an
25 widely used to determine these vital properties. Tensile tests provide vital information
26
27 about the strength and elongation-to-failure of a part under static loading, and fatigue tests
28
M
29
30 indicate the mechanical behavior of parts under cyclic loading.
31
32
33
d
3.1 Tensile Properties
34
35
te

36 The tension (or tensile) test is a very common mechanical test for measuring
37
38 specifications related to material strength. Tensile properties and hardness of DLD parts
p

39
40
41 are usually equal-to-or-greater than those of wrought and cast materials by virtue of high
ce

42
43 cooling rates. The high cooling rates during DLD foster the development of fine
44
45
Ac

46 microstructures distributed throughout the parts’ volume [56,101].


47
48 Tensile properties of various laser-produced and wrought materials are listed in
49
50
51 Table 2 [57,63,102–105]. As shown in Table 2, the ultimate tensile and yield strength of
52
53 DLD parts are higher than those fabricated from wrought materials in most cases. However,
54
55
56
the elongation-to-failure is typically lower for DLD parts and this can be attributed to
57
58 micro-porosity and oxide inclusions within the parts – a result of a non-optimized DLD
59
60 process. According to Table 2, it is difficult to draw any general conclusion regarding the
61
62
63
64 Page 19 of 103
65
1
2
3
4 effect of DLD on work hardening (strain hardening or cold working which is the process
5
6
7 of strengthening of a metal through plastic deformation). Although the higher cooling rates
8
9 experienced during DLD affects the grain size, as well as dendritic arm spacing, the results
10
11
12 indicate that work hardening is a more material-dependent phenomenon.

t
13

ip
14 The part’s building orientation (or tool path orientation relative to substrate) during
15
16
DLD affects the resultant tensile properties of the part. As shown in Fig. 6, specimens

cr
17
18
19 deposited in the direction along (e.g. parallel to) the length of the tensile samples (e.g. X-

us
20
21
direction) typically exhibit higher tensile strength than the layers deposited perpendicular
22
23
24 to the length of the tensile samples (e.g. Y- or Z-direction). A summary of building

an
25
26 orientation effects, e.g. X-direction (parallel) versus Y-direction (perpendicular), on tensile
27
28
properties for several materials are listed in Table 3 [42,44,57]. This anisotropic behavior,
M
29
30
31 as evidenced in Table 3, can be explained by the presence of weak interfacial layers parallel
32
33
d

34 to the crack, in Z- and Y-directions, providing an easy path for shear bands [24,32].
35
te

36 Different cooling rates for these two deposition orientations may also influence the
37
38
microstructure and mechanical properties. For samples built in the longitudinal direction
p

39
40
41 (X-direction), the laser passing time between each successive layer is longer compared to
ce

42
43 materials built in the Y- or Z-directions. This diffference causes higher cooling rates, and
44
45
Ac

46 consequently, finer microstructures for specimens deposited in the X-direction [42]. As


47
48 shown in Table 3, tensile properties are more sensitive to build orientations for stainless
49
50
51 steel specimens than for nickel-based superalloys. These observations can be attributed to
52
53 the difference in thermal conductivity between these materials. It is known that the thermal
54
55
56 conductivity of nickel-based superalloys critically depends on temperature and its
57
58 metallurgical condition; this is in contrast to stainless steels which are less temperature-
59
60
61
62
63
64 Page 20 of 103
65
1
2
3
4 dependent [106–108]. Realizing the effects of building orientation is more involved for
5
6
7 nickel-based alloys and requires further investigation.
8
9 The Vickers macro-hardness of Ti-6Al-4V thick-wall (i.e. multi-track) parts is
10
11
12 reported to be higher than those of thin-wall parts [44]. As a result, flat tensile specimens

t
13

ip
14 typically exhibit higher yield and ultimate stresses as well as lower elongation-to-failure
15
16
than round specimens. These effects may originate from microstructural variation caused

cr
17
18
19 by time-variant cooling rates during fabrication [44].

us
20
21 Charpy impact tests have revealed that the impact resistance of DLD parts is
22
23
24 comparable to that of their conventionally-built counterparts [41,43,47]. The Charpy

an
25
26 impact test measures the impact resistance of the material against stress concentration,
27
28
which, according to ASTM E-23 [109], also implies the resistance of material against crack
M
29
30
31 propagation. The Charpy impact energy of laser-deposited AISI 316L SS parts has been
32
33
d

34 reported to be in the range of 90-110 J, which is comparable to the wrought material in the
35
te

36 annealed condition (105 J) [41,110]. Inconel 625 specimens exhibit impact energy of 46-
37
38
49 J, which is lower than the as-rolled materials (65-70 J). Annealing treatments at 1,223 K
p

39
40
41 have been reported to result in ~10% improvement in the impact energy of laser-deposited
ce

42
43 Inconel 625 specimens [110].
44
45
Ac

46 Hot Isostatic Pressing (HIP) is a manufacturing process used for reducing the
47
48 porosity, and increasing the density, of materials through elevated temperature and
49
50
51 constant/isostatic gas pressure. The HIP process has noticeable effects on the tensile
52
53 strength of laser-deposited parts. As shown in Fig. 7, adequate densification of LENS Ti-
54
55
56 6Al-4V can be accomplished through HIP - leading to a higher ductility and elongation-to-
57
58 rupture. Furthermore, the high temperatures experienced during the HIP process causes an
59
60
61
62
63
64 Page 21 of 103
65
1
2
3
4 increase in the alpha-platelet thickness of Ti-6Al-4V, which consequently reduces the
5
6
7 strength of the material [24].
8
9 Kobryn & Semiatin alternated the DLD laser path by 90º with each subsequent
10
11
12 layer of Ti-6Al-4V [24]. As a result, the tensile properties of specimens in X- and Y-

t
13

ip
14 directions were found to be similar, as shown in Fig. 7. Figure 7 also demonstrates that
15
16
more anisotropy exists in the as-built specimens as compared to the ones after the HIP

cr
17
18
19 process [24]. This indicates that porosity may play an important role in the anisotropic

us
20
21 behavior of laser-deposited parts, especially when measuring their ultimate strength.
22
23
24 The effects of mechanical or crystallographic texture on a part’s anisotropic

an
25
26 behavior are still noticeable after the HIP process, as evidenced when measuring yield
27
28
strength. Crystallographic textures represent the crystalline structure and give information
M
29
30
31 about crystal orientation of different phases, while mechanical texture deals with effective
32
33
d

34 grain size and slip length. Anisotropy in mechanical texture may be due to the columnar
35
te

36 grain morphology - having a smaller effective grain size in directions perpendicular to the
37
38
columnar grains than in the parallel directions. Anisotropy in crystallographic texture may
p

39
40
41 be caused by high-level of anisotropy in the hexagonal crystal structure, as evidenced in
ce

42
43 titanium alloys. The <100> fiber texture which is formed during columnar solidification
44
45
Ac

46 may be responsible for anisotropy in the crystallographic texture of LENS Ti-6Al-4V parts
47
48 [24].
49
50
51 In summary, mechanical properties such as tensile and impact resistance of DLD
52
53 parts are generally similar or higher than those of wrought and cast materials. The relatively
54
55
56
high cooling rates experienced during the DLD process favors the production of finer
57
58 microstructures and this results in acceptable mechanical properties. However, pores and
59
60
61
62
63
64 Page 22 of 103
65
1
2
3
4 defects within DLD parts, which are typical ‘side-effect’ of laser deposition, may
5
6
7 deteriorate their mechanical properties. Anisotropy in tensile properties of DLD parts is
8
9 attributed to layer orientations and is of concern. Higher tensile strength is usually observed
10
11
12 for specimens oriented horizontallyas compared to parts oriented vertically during

t
13

ip
14 fabrication. Weak orientations, resulting from the lack of perfect bounding between layers
15
16
in horizontally-oriented specimens, are parallel to the loading direction, and therefore, do

cr
17
18
19 not significantly affect the tensile properties of the DLD part. In horizontally-oriented

us
20
21 specimens, voids are oriented parallel to the loading axis, hindering defect growth, and
22
23
24 therefore, retarding inter-layer de-bonding. The weak orientation of laser-deposited parts

an
25
26 is perpendicular to the loading direction in vertically-oriented specimens which results in
27
28
considerably lower tensile strengths. Moreover, the different thermal histories for
M
29
30
31 vertically-oriented specimens lead to more porosity and residual stress (in tension) which
32
33
d

34 results in lower tensile strength compare to horizontally-oriented specimens. Performing a


35
te

36 Hot Isostatic Pressing (HIP) procedure after DLD can significantly reduce the anisotropic
37
38
mechanical behavior of the laser-deposited parts by reducing manufacturing-induced
p

39
40
41 porosity.
ce

42
43
44
3.2 Fatigue Resistance
45
Ac

46
47
48 3.2.1. Background
49
50
51 Understanding and quantifying fatigue is critical for the effective analysis and
52
53 design of many engineering materials. Fatigue is one of the main failure modes of materials
54
55
56
in mechanical applications. During the fatigue process, cyclic damage leads to local
57
58 cracking and causes fracture after a sufficient number of fluctuations. According to ASTM
59
60 E1823, fatigue is defined as: “the process of progressive localized permanent structural
61
62
63
64 Page 23 of 103
65
1
2
3
4 change occurring in a material subjected to conditions that produce fluctuating stresses and
5
6
7 strains at some point or points that may culminate in cracks or complete fracture after a
8
9 sufficient number of fluctuations [111].”
10
11
12 Fatigue failure can be characterized by three stages: crack initiation (nucleation),

t
13

ip
14 crack propagation, and final fracture. Cracks tend to initiate at grain boundaries, inclusions,
15
16
pores and other microstructural defects. In this stage, cracks typically propagate along

cr
17
18
19 maximum shear stress planes and grow to lengths of several hundred micrometers (e.g.

us
20
21 microscopic growth), as presented in Fig. 8. Crack growth behavior at this length scale is
22
23
24 significantly affected by crack surface morphology and microstructure texture. As a

an
25
26 consequence of crack growth, the stress intensity factor, 𝐾, increases and starts to develop
27
28
in different planes. Crack growth typically consists of small crack growth in the plane of
M
29
30
31 maximum shear stress, followed by long crack growth (e.g. macroscopic growth) along the
32
33
d

34 maximum tensile plane, as depicted in Fig. 8. Finally, by approaching the maximum stress
35
te

36 intensity, 𝐾max , to plane-strain fracture toughness, 𝐾Ic , stable crack propagation changes
37
38
to unstable crack propagation. At this stage, if tensile stresses are high enough, cracks
p

39
40
41 become so large that sudden failure occurs [112,113].
ce

42
43
Fatigue life is typically quantified by using the number of load cycles/fluctuations
44
45
Ac

46 until failure, 𝑁f . Generally speaking, fatigue life that is less than 103 cycles is considered
47
48 as low cycle fatigue (or short-life regime), fatigue life in the range of 103 < 𝑁f < 105 is
49
50
51 classified as mid-cycle fatigue (or mid-life regime), and 𝑁f > 105 is categorized as high-
52
53 cycle fatigue (or long-life regime). Different analytical or computational fatigue life
54
55
56 models are available for design engineers to predict 𝑁f , such as: the nominal stress-life
57
58 (𝑆 − 𝑁) model, the local strain-life (𝜀 − 𝑁) model, and the fatigue crack growth (𝑑𝑎/𝑑𝑁 −
59
60
61
62
63
64 Page 24 of 103
65
1
2
3
4 ∆𝐾) model [112]. In addition, either the stress-life or strain-life method can be combined
5
6
7 with the fatigue crack growth method to consider both crack initiation and all growth stages
8
9 inherent to the fatigue process – and this is the so-called ‘two-stage model’ [112].
10
11
12

t
13

ip
14
15
16

cr
17
18 3.2.2. Fatigue Crack Initiation
19

us
20 Low cycle fatigue (LCF) [105] and high cycle fatigue (HCF) [104] behavior of
21
22
23 double annealed TC18 titanium alloy (Ti-5Al-5Mo-5V-1Cr-1Fe) fabricated by DLD have
24

an
25 been recently investigated. A thick plate-like specimen of TC18 titanium alloy was
26
27
28 manufactured by a ‘laser melting deposition’ (LMD) process. For the LCF specimens
M
29
30 [105], the plate was stress-relieved at 750 ºC for 1 h, followed by air cooling (AC). It was
31
32
33
then annealed at 880 ºC for 1 h, followed by furnace cooling (FC) to 750 ºC for another 2
d

34
35 h, and air cooled to room temperature. Finally, it was aged at 580 ºC for 4 h and again air
te

36
37 cooled. The HCF specimens [104] also underwent a similar heat treatment process.
38
p

39
40 Moreover, The LCF specimens were machined by electric discharge wire cutting (EDM)
41
ce

42 from the plate according to the GB/T-15248 standard [114] in a vertical direction, while
43
44
45 the HCF specimens were machined in a horizontal direction. The surfaces of all specimens
Ac

46
47 were mechanically polished. Tension LCF tests were performed at a strain ratio, 𝑅s =
48
49
50 𝜀𝑚𝑖𝑛 ⁄𝜀𝑚𝑎𝑥 = -1 (e.g. fully reversed), and HCF tests were performed at a stress ratio 𝑅s =
51
52 𝜎𝑚𝑖𝑛 ⁄𝜎𝑚𝑎𝑥 = 0.1.
53
54
55 According to Li et al. [105], cracks initiated from pores within most of the
56
57 specimens tested during LCF tests. Two crack initiation sites for DLD TC18 specimens
58
59
60
were observed on the fractured surface in the LCF regime with different features. Since
61
62
63
64 Page 25 of 103
65
1
2
3
4 crack progress takes more time in the LCF regime, there might be an opportunity for other
5
6
7 cracks to initiate [115]. The primary crack (main crack) origin site was found to be flat and
8
9 smooth with radial stripes, while the secondary crack was a circular pit with a limited area
10
11
12 for propagation [105]. Due to the different sizes and locations of pores leading to distinct

t
13

ip
14 levels of stress concentration, there were different initiating and propagating rates for the
15
16
macroscopic cracks resulting in distinctions between primary and secondary crack origins

cr
17
18
19 [116]. According to Li et al. [105], the fatigue life of a material subjected to a given strain

us
20
21 range depends on ductility, ultimate tensile strength, and elastic modulus - with ductility
22
23
24 being the dominant property in high strains (e.g. 1% to 2%). Therefore, higher ductility

an
25
26 leads to a longer life in LCF, and this was confirmed by Shamsaei and Fatemi for medium
27
28
carbon steels [117].
M
29
30
31 For lower strain ranges (e.g. HCF), increasing the tensile strength may be more
32
33
d

34 beneficial than increasing the ductility [105]. The microstructure of double annealed DLD
35
te

36 TC18 consists of fine lamella-like primary α phase and transformed β matrix, in contrast
37
38
to bi-modal (e.g. a mixture of both lamellar and equiaxed) microstructure of wrought TC18.
p

39
40
41 Due to more tortuous crack paths in the lamellar, which leads to higher energy dissipation
ce

42
43 during fracture, the microstructure of DLD TC18 leads to slower fatigue crack propagation
44
45
Ac

46 rates as compared to the bi-modal microstructure [104,118]. Accordingly, the lamellar


47
48 microstructure of DLD TC18 favors LCF resistance where the behavior is dominated by
49
50
51 crack propagation. Furthermore, Li et al. [105] reported that the continuous grain boundary
52
53 of α phase leads to straight-wise propagation while the discontinuous one (α phase) causes
54
55
56
non-straight propagation, and consequently, reduces the propagation rate during the crack
57
58
59
60
61
62
63
64 Page 26 of 103
65
1
2
3
4 growth along the grain boundary. Hence, reducing the continuity of the grain boundary by
5
6
7 heat treatments may enhance the overall fatigue resistance of DLD parts [105].
8
9 The fatigue strength of DLD TC18 with Widmanstätten (e.g. lamellar)
10
11
12 microstructure at 107 cycles, and its comparison to wrought TC18 with lamellar and bi-

t
13

ip
14 modal microstructures, is presented in Fig. 9 [33,104]. It may be seen that DLD TC18
15
16
shows significantly lower HCF strength as compared to wrought TC18. Moreover, the HCF

cr
17
18
19 strength of wrought TC18 with Widmanstätten (e.g. lamellar) microstructure is lower than

us
20
21 the one with bi-modal microstructures. Apart from different heat treatment processes for
22
23
24 wrought materials, this observation is mainly due to the presence of micro-pores in the

an
25
26 DLD specimens and different microstructures, respectively. Pores may greatly affect the
27
28
fatigue behavior by creating stress concentrations at their walls.
M
29
30
31 Local microscopic stresses larger than the yield strength cause local plastic
32
33
d

34 deformation and can lead to fatigue crack initiation under cyclic loading. It worth noting
35
te

36 that in contrast to LCF, crack initiation contains the dominant part of total fatigue lifetime
37
38
in HCF, as presented in Fig. 10(a). Therefore, any variations in microstructure
p

39
40
41 discontinuity caused by defects, such as cracks, porosity and un-melted powder can
ce

42
43 accelerate the nucleation of cracks and may then decrease the total fatigue life in the HCF
44
45
Ac

46 regime [104,118]. It has been found that the micro-pore size and its distance from the
47
48 surface strongly affects HCF life of titanium alloys [116]. As reported by Wang et al. [104],
49
50
51 at the same stress amplitude, fatigue life of DLD TC18 decreases with an increase in pore
52
53 size and a decrease in its distance from the surface. Moreover, a pore’s location is a more
54
55
56
influential parameter than the pore size on the fatigue resistance at lower stress levels (e.g.
57
58 HCF).
59
60
61
62
63
64 Page 27 of 103
65
1
2
3
4 Microstructural features also have a significant (and complicated) influence on
5
6
7 fatigue behavior. Finer microstructures, as a result slip bands with higher density, typically
8
9 exhibit better crack initiation resistance as compared to coarser microstructures. However,
10
11
12 materials with finer microstructure promote a flatter crack path that results in higher crack

t
13

ip
14 growth rates. Therefore, materials with finer microstructure usually exhibit less resistance
15
16
to crack propagation resulting in inferior LCF behavior, as compared to coarser

cr
17
18
19 microstructures with a rougher crack path, as presented in Fig. 10(b). For instance, bi-

us
20
21 modal microstructure exhibits better crack initiation resistance, resulting in better HCF
22
23
24 behavior, compared to lamellar microstructures, as shown in Fig. 9. This can be attributed

an
25
26 to lower crystallographic relationships between grains of bi-modal microstructures -
27
28
resulting in more resistance to slip activity by phase interfaces [104,119].
M
29
30
31 The Widmanstӓtten (lamellar) microstructure resulted from double annealing of
32
33
d

34 DLD TC18 has low crack growth resistance, and therefore, is not favored for LCF
35
te

36 applications. Although grain boundaries improve the resistance to deformation due to their
37
38
blocking effects on the dislocations’ slip, the soft zone formed along the grain boundary α
p

39
40
41 may accelerate the crack propagation. This resistance against deformation may also cause
ce

42
43 higher stress and extra hardening in the materials. The main crack in DLD TC18 tends to
44
45
Ac

46 propagate glidingly and easily along the grain boundary α where a softer α precipitated-
47
48 free zone is typically formed. Heat treatment processes can eliminate the negative effect of
49
50
51 the soft zone and provide the incoherent grain boundary α [104].
52
53 The HCF performance of Inconel 718 fabricated by LENS is presented in Fig. 11
54
55
56 [44]. Tensile fatigue tests were performed at a stress ratio of 𝑅𝑠 = 0.1. Before machining
57
58 and polishing specimens to a roughness 𝑅a < 0.2 μm, a heat treatment schedule was
59
60
61
62
63
64 Page 28 of 103
65
1
2
3
4 performed on a plate of DLD Inconel 718: 1093 °C for 1 h, cooling to 718 °C (with 84
5
6
7 °C/h), then 718 °C for 4 h, cooling to 620 °C (at 56 °C/h), then 620 °C for 16 h, cooling to
8
9 below 371 °C (at 139 °C/h), and finally AC to room temperature [44]. Results indicate that
10
11
12 the fatigue strength of a laser-deposited Inconel 718 specimen in midlife and long life

t
13

ip
14 regimes may be comparable with wrought Inconel 718. Better fatigue behavior is expected
15
16
in high cycle regime for DLD materials since they possess finer microstructures resulting

cr
17
18
19 from relatively high cooling rates, as presented in Fig. 10(b). However, a significant drop

us
20
21 in the tolerable stress in very high cycle fatigue regime (Nf > 107) is noticeable in Fig. 11
22
23
24 which may be explained by fatigue behavior being more sensitive to defects (e.g. pores and

an
25
26 inclusion) than microstructure. Lack of bonding between layers and the presence of defects
27
28
have been reported in [44] as the main reasons for the lower fatigue resistance of LENS
M
29
30
31 Inconel 718 in very high-cycle regime - where a large portion of the total fatigue life is
32
33
d

34 spent towards crack initiation, as shown in Fig. 10(a).


35
te

36 Contradictory results have been reported in the literature for the fatigue behavior of
37
38
laser-deposited Ti-6Al-4V specimens (DLD Ti-6Al-4V). The HCF data from various
p

39
40
41 studies on DLD Ti-6Al-4V, printed in different directions as presented in Fig. 6, are
ce

42
43 summarized in Fig. 12 [24,28,44]. Results indicate that LENS Ti-6Al-4V has comparable
44
45
Ac

46 fatigue behavior to its wrought form [28,44]. However, Kobryn & Semiatin report
47
48 significantly lower fatigue strength for LENS Ti-6Al-4V in midlife regimes [24]. This
49
50
51 variation in published fatigue results may be attributed to different test set-ups,
52
53 manufacturing parameters, and post manufacturing processes.
54
55
56 Grylls [28] and Amsterdam & Kool [44] used a stress ratio of 𝑅𝑠 = 0.1 in their
57
58 tensile fatigue tests; however, the stress ratio in Kobryn & Semiatin’s study was not
59
60
61
62
63
64 Page 29 of 103
65
1
2
3
4 reported. Specimen size and surface condition can also contribute to different fatigue lives
5
6
7 observed in these studies. In Amsterdam & Kools’ study [44], fatigue specimens were
8
9 machined and polished to a roughness 𝑅a < 0.2 μm ; however, no surface condition
10
11
12 description was provided in the other fatigue studies. The DLD process parameters have a

t
13

ip
14 strong influence on the thermal history, microstructure, bonding between layers, pore size
15
16
and shape, and therefore, on the fatigue resistance of the fabricated part.

cr
17
18
19 Residual stresses accumulated in parts produced by DLD is another thermal-

us
20
21
22
history-dependent phenomenon that can greatly affect the fatigue behavior of LBAM
23
24 products [56]. Post-manufacturing process parameters, such as heat treatment, are also

an
25
26 influential factors on the fatigue behavior. In Amsterdam & Kool’s study [44], Ti-6Al-4V
27
28
specimens were heat treated at 970 °C for 1 h, water quenched and aged at 538 °C for 4 h,
M
29
30
31 and then air cooled, while Kobryn & Semiatin’s specimens [24] were stress relieved in
32
33
d

34 vacuum for 2 h at 700 – 730 °C. Grylls [28] does not report any heat treatments for the
35
te

36 specimens. Comprehensive studies discussing the effects of process parameters, print


37
38
orientations, and post manufacturing treatments on the fatigue behavior of DLD parts are
p

39
40
41 still needed.
ce

42
43 The Ti-V phase diagram at 6% Al is shown in Fig. 13. It can be observed that single-
44
45
Ac

46 phase β exists above 1,000 ºC. At room temperature, the microstructure of Ti-6Al-4V,
47
48 which consists of both HCP α and BCC β phases, varies based on thermal history. The
49
50
51 cooling rate is the determining factor for the nature of the resultant phase (e.g. martensite,
52
53 lamellar, equiaxed, or bi-modal) during the cooling from the β-phase region to a
54
55
56 temperature significantly below the β transus. Heat treating below the transus temperature
57
58 for Ti-6Al-4V (approximately 1,000 ºC) will not result in a full β annealed phase.
59
60
61
62
63
64 Page 30 of 103
65
1
2
3
4 Therefore, heat treatment of DLD Ti-6Al-4V specimens within the α+β phase field (for
5
6
7 example, point A in Fig. 13) does not change the mechanical properties. However, it may
8
9 increase the ultimate tensile strength and decrease the ductility within the β-phase field
10
11
12 (e.g. over 1000 ºC, marked as point B in Fig. 13) [120]. Heat treatments affect the size,

t
13

ip
14 distribution and morphology of the α phase; however, they have no significant effect on
15
16
the prior-β grains [31]. Heat treatment of laser-deposited Ti-6Al-4V specimens close to the

cr
17
18
19 solutionizing temperature (900 - 970 ºC) changes the acicular - microstructure to

us
20
21
22 columnar grains and results in coarser Widmanstätten grains and lower porosities
23
24 [32,121]. This coarser microstructure increases ductility; and therefore, improves LCF life

an
25
26
27 of Ti-6Al-4V, as presented in Fig. 10 (b).
28
M
29 Anisotropic fatigue resistance of laser-deposited parts has been reported [24]. As
30
31
32
shown in Fig. 14, the fatigue behavior of LENS Ti-6Al-4V specimens depends on the
33
d

34 fabrication/deposition direction (previously shown in Fig. 6). Specimens that are vertically-
35
te

36 oriented (Z-direction) during fabrication show significantly lower fatigue strength than
37
38
horizontally-oriented (X-direction) and laterally-oriented (Y-direction) specimens [24].
p

39
40
41
ce

This can be attributed to the various laser paths inducing inhomogeneous cooling behavior
42
43
44 of the melt pool which results in different microstructure features, residual stresses, and
45
Ac

46 consequently, fatigue behavior. Specimens fabricated in the Z-direction contain extensive


47
48
49
porosity compared to those built in the X- and Y-direction [24]. In addition, parts fabricated
50
51 by alternating the laser deposition path by 90º with each subsequent layer have similar
52
53 fatigue behavior as specimens fabricated in X- and Y- directions. Post-DLD processes,
54
55
56 such as HIP, improve the fatigue strength of laser-deposited parts up to the level of their
57
58 wrought counterparts, especially in the long life regime where crack initiation constitutes
59
60
61
62
63
64 Page 31 of 103
65
1
2
3
4 most of the total fatigue life, as shown in Fig. 14. In addition, HIP’ed specimens exhibit
5
6
7 less anisotropic behavior than heat treated ones, as shown in Fig. 14. Such observations
8
9 indicate that porosity may play a significant role in the anisotropic behavior of laser-
10
11
12 deposited parts. In general, the effects of different manufacturing and post manufacturing

t
13

ip
14 parameters on the anisotropic behavior of laser-deposited parts are not yet well understood;
15
16
thus, further investigations are required.

cr
17
18
19

us
20
21
22
23
3.2.3. Fatigue Crack Growth
24

an
25 Fracture toughness, or the ability of material containing pre-existing flaws to resist
26
27
28 fracture, is an important material property for many design applications. The stress-
M
29
30 intensity factor (𝐾I ) may be used to calculate the fracture toughness of many materials in
31
32
33
‘Mode I’ (tensile or opening mode) of fracture. Beyond a critical dimension, the value of
d

34
35 𝐾I becomes relatively constant and is then referred to as the plane-strain fracture toughness,
te

36
37
38
𝐾Ic . According to ASTM E1823, the plane-strain fracture toughness is defined as “the
p

39
40 crack-extension resistance under conditions of crack-tip plane-strain in mode I for slow
41
ce

42 rates of loading under predominantly linear-elastic conditions and negligible plastic-zone


43
44
45 adjustment; 𝐾Ic provides for the measurement of crack-extension resistance at the onset
Ac

46
47 (2% or less) of crack extension [111].”
48
49
50 Kobryn & Semiatin [24] extracted Compact Tension (CT) specimens from DLD
51
52 cubes according to ASTM E399 [122]. Some specimens were heat treated in vacuum for 2
53
54
55 h at 700 – 730 °C, while the others underwent HIP for 2 h at 900 °C and 100 MPa. It was
56
57 found that heat treated specimens fabricated in Y- and Z-directions exhibited higher
58
59
60
fracture toughness than heat treated specimens built in the X-direction. This finding may
61
62
63
64 Page 32 of 103
65
1
2
3
4 be attributed to the crack growth path with respect to layers’ orientations. For instance, the
5
6
7 crack growth path was perpendicular to deposited layers for specimens fabricated in Y-
8
9 and Z-directions, whereas the crack growth path was parallel to layers in specimens
10
11
12 fabricated in the X-direction. Furthermore, the weak interfacial layers of specimens

t
13

ip
14 fabricated in X-direction may favor crack propagation. Specimens built in Z-direction
15
16
showed the highest toughness when compared to ones built in other directions; for both

cr
17
18
19 heat treatment and HIP. The HIP process, however, was shown to improve the toughness

us
20
21 of LENS Ti-6Al-4V specimens towards that of wrought and cast Ti-6Al-4V regardless of
22
23
24 the building orientation [24].

an
25
26 Figure 15 presents 𝐾Ic for LENS Ti-6Al-4V specimens deposited in different
27
28
building orientations with 𝐾Ic measured from cast and wrought Ti-6Al-4V specimens
M
29
30
31 [24,123]. It may be seen that the HIP process reduces the effects of anisotropy on the
32
33
d

34 fracture toughness of LENS specimens. Similar fracture toughness values measured for
35
te

36 HIP’ed specimens fabricated in all three directions indicate the dominant effect of porosity,
37
38
mainly due to lack-of-fusion, on the anisotropic behavior of laser-deposited parts with
p

39
40
41 respect to tensile and fatigue behavior.
ce

42
43
Fatigue crack growth rates (FCGRs) for LENS Inconel 625 and AISI 316L SS are
44
45
Ac

46 compared with the FCGRs for their corresponding wrought materials [124] in Fig. 16 –
47
48 using the crack growth rate, d𝑎/d𝑁, versus stress intensity factor, ∆𝐾 [41,43,110]. The 25-
49
50
51 mm-thick CT specimens for fracture toughness and fatigue tests were extracted from LENS
52
53 samples according to ASTM E647 [125]. Fracture toughness tests were conducted based
54
55
56 on the ASTM E1820 [126] standard at a stress ratio of 𝑅s = 0.3. The direction of crack
57
58 growth in the CT specimens was shown to be parallel to the direction of laser deposition
59
60
61
62
63
64 Page 33 of 103
65
1
2
3
4 and perpendicular to the specimen’s build up direction. As shown in Fig. 16(a), the FCGR
5
6
7 for a laser-deposited Inconel 625 specimen is lower than that of the wrought Inconel 625
8
9 [127] for the stress intensity factor range of ∆𝐾 = 20 - 25 MPa√m. The FCGR for laser-
10
11
12 deposited and wrought Inconel 625 specimens tend to coincide for ∆𝐾 values only above

t
13

ip
14
15
25 MPa√m. However, the FCGR for the laser-deposited AISI 316L SS specimen is similar
16

cr
17 to the ones reported for their wrought counterparts, as presented in Fig. 16(b).
18
19 The J-integral fracture toughness (𝐽Ic ) for laser-deposited Inconel 625 and AISI

us
20
21
22 316L SS specimens is reported to be, respectively, in the range of 194-254 KJ/m2 and 143-
23
24 259 KJ/m2 and lower than their wrought counterparts [110]. Transgranular fatigue crack

an
25
26
27 propagation is exhibited for both LENS Inconel 625 and LENS AISI 316L SS [110]. The
28
M
29 crack opening concept can also be used to measure the toughness of a material. According
30
31
32 to ASTM E1290 [128], the crack tip opening displacement (CTOD) is defined as “the crack
33
d

34 displacement due to elastic and plastic deformation at variously defined locations near the
35
te

36
37 original (prior to an application of force) crack tip”. The CTOD fracture toughness for both
38
p

39 LENS Inconel 625 and LENS AISI 316L stainless steel is found to be in the range of 0.28-
40
41
ce

42
0.54 mm [110]. The CTOD fracture toughness values of LENS Inconel 625 and LENS
43
44 AISI 316L SS are lower than that of their wrought counterparts which are 1.03-1.16 mm
45
Ac

46 and 1.84–1.67 mm, respectively [41,47]. Heat treatment at 950 °C for 1 h slightly improves
47
48
49 the CTOD fracture toughness of LENS Inconel 625 to the range of 0.34 – 0.54 mm [41,47].
50
51 Aging alone does not significantly affect the microstructure of nickel-based
52
53
54 superalloys, while solution heat treatment followed by aging may result in a thin
55
56 recrystallized layer at the surface. Solution treatment and aging may cause an increased γ'
57
58
59
60
61
62
63
64 Page 34 of 103
65
1
2
3
4 precipitate size and a higher cubic morphology, enhancing the facture toughness of Inconel
5
6
7 625 [57].
8
9
10 3.2.4. Fatigue Modeling
11
12

t
13 Fatigue behavior of laser-deposited metallic parts mainly depends on the

ip
14
15
microstructure resulting from DLD processing conditions. Therefore, a microstructure
16

cr
17
18 sensitive fatigue model that can incorporate microstructural features may be appropriate
19

us
20 for modeling the cyclic behavior of DLD parts. Recent advancements in electron
21
22
23 microscopy have resulted in additional fatigue stages being observed, resulting in multi-
24

an
25 stage fatigue (MSF) models being proposed. In such models, the crack initiation stage is
26
27
28 typically divided into crack incubation and microstructurally small crack growth [129].
M
29
30 The multi-stage fatigue (MSF) model of McDowell et al. [129] can be a good candidate to
31
32
33
capture the effect of microstructural content (e.g. porosity, pores size their distance, grain
d

34
35 size and orientation, inclusion size) on fatigue lives. This model was used by Xue et al.
te

36
37 [38] to incorporate the microstructural features in the fully-reversed (R=-1) strain-
38
p

39
40 controlled fatigue behavior of LENS AISI 316L stainless steel. The MSF model comprises
41
ce

42 of three regimes of fatigue damage: crack incubation, microstructurally small crack (MSC)
43
44
45 and/or physically small crack (PSC) growth, and long crack (LC) growth [38,110] as shown
Ac

46
47 in Eq. (1):
48
49
50 𝑁𝑇𝑜𝑡𝑎𝑙 = 𝑁𝐼𝑁𝐶 + 𝑁𝑀𝑆𝐶/𝑃𝑆𝐶 + 𝑁𝐿𝐶 (1)
51
52 where 𝑁Total is the total fatigue life. Here, 𝑁Inc is the number of cycles to incubate a crack
53
54
55
(including nucleation and small crack growth around an inclusion on order of ½ the
56
57 inclusion pore size or particle diameter) and 𝑁MSC is the number of cycles required for
58
59 propagation of a microstructurally small crack. The MSC regime can extend to almost a
60
61
62
63
64 Page 35 of 103
65
1
2
3
4 millimeter depending on the texture of the matrix and microstructural inclusion
5
6
7 morphology. The 𝑁PSC term is the number of cycles required for propagation of a physically
8
9 small crack (PSC), and 𝑁LC is the number of cycles required for long crack propagation
10
11
12 which is applicable to growth in the range of a > (10–20) MS. Here, MS is the characteristic

t
13

ip
14 length scale of interaction with microstructural (MS) features which can be defined as the
15
16
dendrite cell size (DCS) or the smallest grain size (GS) [38,129–131].

cr
17
18
19 The MSF model predicts fatigue damage in materials based on microstructural

us
20
21 inclusions, porosity, dendrite cell size (DCS), and nearest neighbor distance (NND)
22
23
24 [129,132]. This microstructure-sensitive capability was used to formulate different stages

an
25
26 of fatigue damage in LBAM components [38]. Xue et al. [38] used X-ray computed
27
28
tomography to capture average and maximum values of pore size, pore density (number of
M
29
30
31 pore per mm2), and porosity fractions, as well as the shape of each pore. The pore shape is
32
33
d

34 characterized by the sphericity coefficient, λ, which quantifies the resemblance of the pore
35
te

36 to a perfect sphere [38]. They investigated the fatigue behavior as well as the upper and
37
38
lower bounds for the predicted strain–life curves using the MSF model by varying the
p

39
40
41 incubation pore porosity and size. The largest and smallest pore sizes were used to predict
ce

42
43 the lower and upper bounds, respectively. Their results in comparison with the ASME Code
44
45
Ac

46 Sect. III design curve [133] are presented in Fig. 17. It may be observed that the fatigue
47
48 performance of LENS 316L stainless steel is lower than the ASME design curve as a result
49
50
51 of inherent porosity in DLD specimens. The upper bound predicted by the MSF model is
52
53 close to the ASME design curve and this indicates that pore size is the main reason affecting
54
55
56
the fatigue behavior of DLD AISI 316L SS. According to these results, fatigue cracks are
57
58 formed mostly at the relatively large pores located at or near the specimens’ surface. They
59
60
61
62
63
64 Page 36 of 103
65
1
2
3
4 also reported occasional crack initiations from the partially melted particles on the surface.
5
6
7 Various pore sizes in different specimens were found to cause scatter in fatigue data. The
8
9 adjacent large discontinuities in some specimens also reduced the fatigue incubation life,
10
11
12 dominating the HCF regime [38]. Xue et al. [38] attributed the scatter in total fatigue life

t
13

ip
14 data to the scatter in the fatigue incubation lives, as they did not observe much
15
16
microstructural effects on the fatigue crack growth stage for crack lengths of a > 1.4 mm.

cr
17
18
19 Experimental results have shown that DLD has the ability to produce parts with

us
20
21 fatigue properties comparable to conventionally-built ones. However, the inherent
22
23
24 porosity, inclusions, or any defect in DLD parts may result in poor fatigue properties,

an
25
26 especially in high cycle fatigue (HCF) regime. Pores serve as appropriate cites for crack
27
28
initiation and strongly reduce the HCF life, where crack initiation life dominates the
M
29
30
31 majority of the total life. The fine microstructures typical of DLD parts provide better crack
32
33
d

34 initiation resistance than coarse microstructure due to higher density of slip bands, thus
35
te

36 enhanced fatigue resistance during the HCF regime. At the low cycle fatigue (LCF)
37
38
regime, coarse microstructures are favored due to their better crack propagation resistance.
p

39
40
41 Therefore, heat treatments may be employed to improve the LCF lives of DLD materials
ce

42
43 by increasing the grain size and ductility. Hot isostatic pressing (HIP) can improve both
44
45
Ac

46 LCF and HCF lives up to the level of wrought materials by coarsening the microstructure
47
48 and reducing the porosity. Moreover, the HIP process reduces the effect of building
49
50
51 orientation, and therefore, anisotropy in the fatigue behavior of DLD parts by removing the
52
53 effect of directional porosity. Directionality in cooling behavior of the melt pool, porosity,
54
55
56
and the orientation of weak interfacial layers toward crack propagation path are mainly
57
58 responsible for the observed anisotropy in the fatigue behavior of DLD parts. However, the
59
60
61
62
63
64 Page 37 of 103
65
1
2
3
4 effect of building direction on crack incubation, short and long crack growth, and final
5
6
7 fracture is not yet fully understood.
8
9 It should be emphasized that there is still a need for more comprehensive studies to
10
11
12 understand the effects of LBAM process parameters (e.g. laser power, beam traverse speed,

t
13

ip
14 and powder feed rate) and design strategies (e.g. building orientation and scanning
15
16
patterns) on the post-manufacture mechanical behaviors of the LBAM parts. This may

cr
17
18
19 include investigating the effects of such parameters and strategies on the multi-axial fatigue

us
20
21 behavior, mean stress effect, fatigue behavior under variable amplitude loading, etc.
22
23
24

an
25
26 4. Process Optimization for Enhanced Mechanical Properties
27
28
Since the mechanical integrity of DLD parts depends on the process parameters –
M
29
30
31 which affect the microstructural distribution via thermal history – it is important to
32
33
d

34 optimize the DLD process parameters to generate near-net-shaped parts with minimal
35
te

36 defects. The optimized process parameters can then be utilized for effectively ‘seeding’ a
37
38
thermally-monitored, feedback-controlled DLD process. Optimal process parameters are
p

39
40
41 typically determined via extensive experiments, which usually require high experimental
ce

42
43 costs and a significant time investment. It is challenging to develop a comprehensive and
44
45
Ac

46 general methodology for DLD process optimization because many interacting process
47
48 parameters are involved. The existing research efforts aimed at optimizing the mechanical
49
50
51 properties of DLD parts are now summarized.
52
53
54
55
56 4.1. Dimension Reduction
57
58
59
60
61
62
63
64 Page 38 of 103
65
1
2
3
4 The DLD process is affected by many process parameters, as well as the
5
6
7 interactions among them. The resulting high dimensionality, known as the “curse of
8
9 dimension”, imposes significant challenges in modeling how process parameters affect
10
11
12 DLD part features. To resolve this issue, one seeks to reduce dimension of process models

t
13

ip
14 by grouping parameters – typically in dimensionless form. These parameter groups are
15
16
then investigated to determine their role on the DLD process and their interaction amongst

cr
17
18
19 each other. Buckingham’s Π (Pi) theorem provides a guideline in selecting the number of

us
20
21 non-dimensionalized parameters. However, dimensionless parameters generated via
22
23
24 Buckingham’s theory are not unique as the theory only suggests the number of

an
25
26 dimensionless parameters that can be constructed. In most cases, researchers select
27
28
dimensionless parameters that are useful in understanding the underlying process physics.
M
29
30
31 Possible choices of dimensionless parameters include melting efficiency, deposition
32
33
d

34 efficiency, process efficiency [112,113], laser absorptivity [136], specific energy, and
35
te

36 others. The utilization of common thermo-fluidic dimensionless numbers - such as


37
38
Reynold’s number, Re, Prandtl’s number, Pr and the Bond number, Bo - is also sought as
p

39
40
41 these classical numbers can provide physical insight into various aspects of the DLD
ce

42
43 process. As an example, one can find Re for the melt pool to help in assessing if laminar
44
45
Ac

46 or turbulent flow exists.


47
48 The advantage of using dimensionless parameters is that these parameters are
49
50
51 usually defined based on DLD processes, instead of materials to be processed. Thus,
52
53 dimensionless parameters are less dependent on material properties. In practice, when a
54
55
56
material is to be processed, dimensionless parameters can be used to suggest the range of
57
58 parameters in which good metal parts are formed. Furthermore, the use of dimensionless
59
60
61
62
63
64 Page 39 of 103
65
1
2
3
4 parameters can alleviate the confusion of dealing with various units for a single process
5
6
7 parameter. For example, the units utilized for traverse speed can vary substantially (e.g.
8
9 inches/minute, cm/s, etc.). Utilization of a dimensionless traverse speed allows for easier
10
11
12 communication of process parameters amongst the international community.

t
13

ip
14 Reduced-dimension parameter groupings are just as useful. For instance, Peng et
15
16
al. [137] studied nickel alloy and, similar to quantifying welding processes, they utilized

cr
17
18
19 specific energy - a dimensional parameter grouping of the laser power, traverse speed, and

us
20
21 laser diameter, e.g.:
22
23
24 Laser Power
Specific Energy =

an
(2)
25 Laser Scanning Speed ∗ Laser Diameter
26
27 The specific energy is an important factor for quantifying many laser/welding
28
M
29
30 processes for sufficient bonding strength and joint/layer resistance to collapsing.
31
32
33
d

34
35 4.2. Process Maps
te

36
37 Understanding the DLD process parameters and their impact on the microstructure
38
p

39
40 of the part motivates the creation and use of ‘process maps’ – a methodology particularly
41
ce

42 useful for optimization due to its common use of dimensionless parameters [138–140].
43
44
Such plots possess non-dimensionalized ordinates to help in determining effects related to
45
Ac

46
47 changing scan rate, part preheating and laser power – and are generalized based on
48
49 analytical, numerical or experimental results. In many cases, a simple geometry is modeled
50
51
52 or investigated such as a thin wall structure; for utilizing analytical models, the deposition
53
54 process is typically neglected. The process maps can be used as tools to aid DLD users in
55
56
57 ascertaining the appropriate, initial process parameters for a given material for fabrication
58
59 via DLD.
60
61
62
63
64 Page 40 of 103
65
1
2
3
4 Process maps have been developed for predicting the steady-state melt pool size in
5
6
7 parts for any practical combination of DLD/LENS process variables [86,141–143]. A
8
9 typical steady-state thermal process map can demonstrate how melt pool length is affected
10
11
12 by normalized height of substrate and melt temperature [139]. An advantage of steady-

t
13

ip
14 state process maps are their ease of implementation: all predictions are contained in the
15
16
form of a single three-dimensional or two-dimensional plot that process engineers can use

cr
17
18
19 directly. Vasinonta et al. [139] demonstrate that the predicted melt pool lengths agree well

us
20
21 with the actual measurements during the LENS process. Bontha et al. [68,70] further
22
23
24 generalized the thermal process maps. They investigated solidification cooling rates and

an
25
26 thermal gradients in laser deposition processes, namely solidification maps, with the
27
28
eventual goal of predicting the microstructure of DLD parts. The simultaneous control of
M
29
30
31 residual stress and melt pool size has also been addressed by Vasinonta et al. [144].
32
33
d

34 Thermal process maps have also been developed to conduct transient analyses – to
35
te

36 account for dynamic DLD processes and their effects on, for example, melt pool/track
37
38
morphology. One such dynamic DLD process is the so-called “boundary issue” - which
p

39
40
41 results in an increase in the melt pool size as the layer boundary is approached [145]. The
ce

42
43 boundary problem is exacerbated by the fact that a reduction in laser velocity (and thus an
44
45
Ac

46 increase in thermal energy imparted to the part per distance moved in the deposition
47
48 direction) may be needed to accurately deposit material near the boundary, where a velocity
49
50
51 reversal is needed to continue with deposition of the next layer of material.
52
53 To efficiently and effectively control the melt pool size/morphology and other
54
55
56
process parameters, it is important to understand melt pool size control over a range of
57
58 process size scales and to understand the transient response of melt pool size to step
59
60
61
62
63
64 Page 41 of 103
65
1
2
3
4 changes in laser power or velocity. Specifically, a transient analysis allows one to
5
6
7 determine how changes in process control variables affect the process parameters such as
8
9 the melt pool size, cooling rates, thermal gradients, etc. A transient analysis on the DLD
10
11
12 process also allows one to determine how long it takes for the process parameters to reach

t
13

ip
14 a new steady-state due to a step change in control variables. This information can be used
15
16
to determine the lead/lag time of control actions.

cr
17
18
19 Birnbaum et al. [142] demonstrated how the normalized melt pool length (l) varies

us
20
21 as a function of normalized distance traveled by the laser (X), as shown in Fig. 18. Each
22
23
24 curve represents a process map of the corresponding melt pool temperature, Tm. The

an
25
26 vertical gaps between two curves represent changes in the (normalized) melt pool length
27
28
when the melt pool temperature changes due to various DLD dynamics (e.g., the melt pool
M
29
30
31 temperature may increase when the laser reaches the boundary of the part). The horizontal
32
33
d

34 gaps indicate the (normalized) distance needed to be traveled by the laser to obtain constant
35
te

36 melt pool length when the melt pool temperature changes, which can be converted to the
37
38
laser travel time by dividing by the traverse speed. These times may be useful for
p

39
40
41 establishing lower bounds of response times for existing DLD thermal feedback control
ce

42
43 systems.
44
45
Ac

46 The approaches and results pertaining to process maps can be used to determine, in
47
48 general, how to modify process variables in order to obtain an ideal melt pool size, control
49
50
51 maximum residual stresses and control microstructure. However, the methodology of
52
53 process maps was initially developed specifically for application to LENS, which is
54
55
56
typically equipped with a 500 W Nd:YAG laser, or other lasers with similar power. There
57
58 is a limited fundamental understanding of how to apply deposition knowledge acquired
59
60
61
62
63
64 Page 42 of 103
65
1
2
3
4 from small-scale systems to analogous large-scale systems (e.g., AeroMet, which
5
6
7 manufactures components for the aerospace industry and uses an 18 kW CO2 laser) [142].
8
9 Whenever a new LBAM system is developed at a different size scale, engineers must
10
11
12 perform a large number of experiments to characterize their process. Multiple process maps

t
13

ip
14 with various scales have been developed for predicting part features for the large scale
15
16
process via extrapolation [142,146]. Although the resulting prediction can be used to

cr
17
18
19 provide a possible range for the optimal process parameters in large-scale DLD processes,

us
20
21 the prediction may be inaccurate due to the error caused by model extrapolation. Therefore,
22
23
24 there exists a great need to fill the gap between industrial applications that demand the use

an
25
26 of large-scale deposition processes and the process development that occurs on small-scale
27
28
processes in laboratory conditions.
M
29
30
31 Process maps are advantageous in that they provide a fundamental way to
32
33
d

34 control/predict melt pool size, stress and material properties by presenting results in a form
35
te

36 that process engineers can readily use. However, there also exist two major limitations in
37
38
the existing process map methods. First, the current process maps are for limited part
p

39
40
41 shapes - developed for thin-wall and bulk shapes only. In other words, these process maps
ce

42
43 do not hold for other common shapes, let alone parts with complex geometry. One
44
45
Ac

46 promising application of AM is to fabricate parts whose geometry is so complex that they


47
48 cannot be produced using traditional manufacturing methods. Therefore, in order to make
49
50
51 process maps useful in real-world manufacturing applications, future work is needed to
52
53 develop process maps that characterize the thermal behaviors and mechanical properties of
54
55
56
parts with various shapes. Secondly, process maps do not consider temperature-dependent
57
58 material properties. Current process maps are based on Rosenthal’s analytical solution for
59
60
61
62
63
64 Page 43 of 103
65
1
2
3
4 temperature with moving-heat-source boundary [147], and material properties are idealized
5
6
7 as being independent of part temperature. This may not be a realistic assumption in real
8
9 world applications. In fact, process maps are used to approximate the underlying
10
11
12 fabrication process when the temperature remains in a certain range.

t
13

ip
14
15
16
4.3. Data-Driven Models

cr
17
18
19 Although physics-based models such as process maps are essential for thoroughly

us
20
21 understanding the underlying DLD processes, their development is extremely challenging
22
23
24 due to the complexity associated with DLD. Some research efforts have circumvented this

an
25
26 challenge by utilizing data-driven methods that directly model how the process parameters
27
28
affect the quality of final parts. These methods include, but are not limited to: design of
M
29
30
31 experiments, artificial intelligence and more. A common feature shared by these methods
32
33
d

34 is that process parameters and part features are empirically related based on experimental
35
te

36 data sets. In other words, the developed methods are not completely dependent on the
37
38
domain knowledge of a specific process and thus may be applied to other AM processes
p

39
40
41 by redefining process control parameters and quality measures. Some data-driven methods
ce

42
43 developed for DLD process are now described.
44
45
Ac

46 Research for increasing the quality of DLD parts primarily rely on a trial-and-error
47
48 procedure to determine optimal process parameters and achieve some desired properties of
49
50
51 the fabricated product. That said, statistical design of experiments (DOE) provides a
52
53 systematic framework to utilize the previous experimental data and plan future
54
55
56
experimental trials with the minimal cost. Also, simplified relations between part features
57
58 and process control parameters can be learned in an empirical way.
59
60
61
62
63
64 Page 44 of 103
65
1
2
3
4 Kummailil et al. [29,148] have applied a two-level fractional factorial design to
5
6
7 determine the effects of build parameters on the deposition of Ti-6Al-4V. By analyzing the
8
9 experimental data, a power relationship between the deposition height, mass flow rate and
10
11
12 energy density is reported. In the same way, response surface modeling has been adopted

t
13

ip
14 to optimize process parameters and achieve better geometric accuracy, building speeds and
15
16
surface finishes [149,150]. The key limitation to DOE is that the resulting designs of

cr
17
18
19 experiments are usually tailored for each specific process. Existing applications of DOE

us
20
21 methods do not take into account underlying physics knowledge or results from similar
22
23
24 studies. Therefore, when a new material or a new process is used, extensive experiments

an
25
26 have to be conducted for process optimization.
27
28
Another category of data-driven methods that have been applied for AM process
M
29
30
31 optimization is artificial intelligence (AI), which aims at “training” a black-box model
32
33
d

34 based on a large training data set. Theses algorithms include but are not limited to Support
35
te

36 Vector Machine, Neural Network, Bayesian Network, and their extensions. With a large
37
38
training data set, AI algorithms usually provide accurate estimations of parameter-feature
p

39
40
41 relations. For instance, Lu et al. [151] applied the method of least square support vector
ce

42
43 machine (LS-SVM) to investigate the relation between mechanical properties of parts and
44
45
Ac

46 process parameters such as laser power, traverse speed, and the powder feed rate in LENS.
47
48 Validated by using fabricated thin-walled parts, the method of LS-SVM is reported to
49
50
51 accurately predict the deposition height when a large sample of parts is used to train the
52
53 model. Similar successful applications of AI methods were performed by Casalino and
54
55
56
Ludovico [152], which uses Feed Forward Neural Network (FFNN) to model a laser
57
58
59
60
61
62
63
64 Page 45 of 103
65
1
2
3
4 sintering process, and by Wang et al. [153] which adopts Bayesian Probability Network to
5
6
7 characterize a laser bending process.
8
9 The DLD process optimization for controlled layer height can also be accomplished
10
11
12 using advanced/intelligent computational methods such as the mutable smart bee algorithm

t
13

ip
14 (MSBA) and fuzzy interference system (FIS), and unsupervised machine learning
15
16
approaches such as self-organizing maps (SOMs) [154]. Despite these successful studies,

cr
17
18
19 the application of AI methods is rather rare in the literature of AM at large. This is because

us
20
21 the key to a successful application of AI methods is enormous training data that can be
22
23
24 used to estimate the process model, which usually results in extremely high experimental

an
25
26 costs. Moreover, due to the proprietary nature of DLD experiment data, data sets are often
27
28
hard to obtain.
M
29
30
31
32
33
d

34 5. Process Control
35
te

36 Direct Laser Deposition (DLD) was born as an ‘open-loop’ control manufacturing


37
38
process – with an end-user providing, and responding to, process parameters. As
p

39
40
41 mentioned, these process parameters are often learned via trial-and-error for a given
ce

42
43 material/build-pattern/machine combination and then modifications to the process
44
45
Ac

46 parameters are made manually as needed. Traditionally, process parameters (such as the
47
48 traverse speed, laser power, etc.) are held constant with time while the end-user may control
49
50
51 the idle time between layers. Based on Part I and Part II of this overview, it has been shown
52
53 that time-invariant process parameters typically result in parts with an anisotropic
54
55
56
microstructure that typically requires a post-DLD process such as heat treatment or HIP.
57
58 Utilization of constant DLD process parameters can lead to thermal distortion, increased
59
60
61
62
63
64 Page 46 of 103
65
1
2
3
4 residual stress and cracking [155]. As a result of such observations, an element of DLD
5
6
7 control has emerged over the years and has been researched extensively. With feedback
8
9 (or closed-loop) control, DLD machines are able to monitor and automatically respond to
10
11
12 faulty or low-quality tracks/builds, as measured via (i) near net shape, (ii) highly-dense

t
13

ip
14 (low porosity), low dilution (concentration gradients controlled) and (iv) precise geometry,
15
16
in real-time. The DLD feedback control problem then consists of accurately

cr
17
18
19 monitoring/detecting faults and subsequently responding to such faults as to correct them

us
20
21 in a timely fashion as the process moves forward. Note that controlled process parameters
22
23
24 are selected based on their responsiveness to changes and primarily include the laser power,

an
25
26 traverse speed, hatch distance, process time intervals and more. Certain DLD process
27
28
parameter combinations can result in various microstructures. Selection of process
M
29
30
31 parameters for ensuring that this condition is met imposes an unprecedented challenge in
32
33
d

34 DLD process optimization and control - requiring layer process parameters to vary during
35
te

36 deposition. These ‘layer process parameters’ may interact with other process parameters.
37
38
For example, due to heat retention in the part, an elevated temperature may persist in the
p

39
40
41 previous layer [156,157]. This effect will be magnified during idling events that occur near
ce

42
43 corners and sharp edges, causing a considerable and progressive build-up. Therefore, laser
44
45
Ac

46 parameters must be adjusted according to other process parameters, such as powder flow
47
48 rate, to ensure that retained heat from previous layers will be stationary during the
49
50
51 fabrication process [157]. Most of the existing control methods can be categorized into two
52
53 groups: melt pool control and deposition (layer height) control. In what follows, the recent
54
55
56
developments in these two control methods are reviewed.
57
58
59
60
61
62
63
64 Page 47 of 103
65
1
2
3
4 5.1. Melt Pool Control
5
6
7 Like many laser-based/welding manufacturing processes [158], thermal monitoring
8
9 is employed (see Part I) via infrared thermography and two-color pyrometry to actively
10
11
12 determine part quality based on temperature fields, gradients and cooling rates. Such

t
13

ip
14 ‘thermal control’ is established by means of using this thermal data as input for an
15
16
employed feedback control system. As shown in Fig. 19, proportional-integral-derivative

cr
17
18
19 (PID) control (via LabVIEW) of the DLD process can be accomplished by using the

us
20
21 monitored melt pool temperature distribution for altering laser power [159]. Based on
22
23
24 thermal feedback during the DLD process, there are two means as to alter temperature

an
25
26 behavior (and thus microstructural formation); either the input parameters can be varied
27
28
(e.g. laser power, scan rate, etc.) or thermal energy can be directed to or away from the part
M
29
30
31 via active thermal management. Closed-loop control during DLD is of importance,
32
33
d

34 because it can lead to hands-off, unsupervised manufacturing that produces high-quality,


35
te

36 repeatable parts. However, the wide-scale application of control is a challenge since DLD
37
38
machines and environments can vary and the thermal control is limited by the current state-
p

39
40
41 of-the-art in thermography and diagnostics [158,160].
ce

42
43 The melt pool morphology, while in its liquid phase, is paramount to the integrity
44
45
Ac

46 and shape of each solidified track/layer. As described in Part I, due to bulk heating effects
47
48 and other variables, the melt pool can elongate, shrink, splash and/or become excessively
49
50
51 superheated and unstable. To ensure consistent melt pool morphology during DLD and for
52
53 each deposition layer, process parameters should be varied appropriately. The challenge
54
55
56
is detecting variation in melt pool size/temperature and ensuring that the DLD machine
57
58 automatically and effectively responds to such changes.
59
60
61
62
63
64 Page 48 of 103
65
1
2
3
4 Salehi et al. used a DLD-control system that employed a pyrometer and photodiode
5
6
7 in conjunction with LabVIEW software [159]. The melt pool size and temperature were
8
9 monitored and this data was fed to LabVIEW for accomplishing feedback control of laser
10
11
12 process parameters such as speed and power. For this study, it was found that employing

t
13

ip
14 melt pool temperature/size control can result in more uniform layers. Separate experiments
15
16
were also conducted to monitor/control dilution and the HAZ size on the part, but the

cr
17
18
19 cladding results were not as enhanced. This indicates the need for multi-parameter process

us
20
21 control; for instance, by controlling melt pool size and traverse speed.
22
23
24 Raghavan et al. [161] related the real-time melt pool temperature with sub-surface

an
25
26 temperatures, cooling rates and the post-fabricated mechanical properties of Ti-6Al-4V
27
28
parts fabricated via DLD. A welding model was used for approximating the LENS process
M
29
30
31 and the heat transfer and liquid metal flow was modeled to calculate melt pool shapes and
32
33
d

34 thermal cycles during laser processing of a Ti-6Al-4V alloy [161]. The results indicate that
35
te

36 a more comprehensive control method for LBAM is required, since feedback control that
37
38
is based solely on maintaining a target top surface geometry can be limited. It was shown
p

39
40
41 that monitoring/controlling only the melt pool surface area can be insufficient in producing
ce

42
43 targeted part quality. This is mainly because monitoring the top surface melt geometry does
44
45
Ac

46 not provide sufficient information to accurately predict the melt pool depth. Despite similar
47
48 top surface contours, the overall pool geometry can vary considerably, as presented in Fig.
49
50
51 20. Furthermore, since a clear correlation between the peak temperature and melt pool
52
53 geometry is not readily apparent, it may be difficult to accurately implement process
54
55
56
control based solely on thermal imaging of the temperature profiles on the top surface of
57
58 the melt pool [161]. Raghavan et al. also demonstrated that as the bulk temperature of the
59
60
61
62
63
64 Page 49 of 103
65
1
2
3
4 part increases, and the laser power is modified to control the melt pool profile, variation in
5
6
7 localized solidification was observed – indicating a drawback in utilizing melt pool aerial-
8
9 shape/laser control with regard to microstructure [161].
10
11
12 Deposited layers may possess cracks and porosities which are known to alter part

t
13

ip
14 functionality. Thus, when these defects go undetected, they can lead to part failure or low
15
16
part performance during application. A low-cost alternative to melt pool monitoring and

cr
17
18
19 control is to detect defects in layers as they are deposited. Barua et al. [162] developed a

us
20
21 monitoring system consisting of an SLR camera for obtaining images of clad lines
22
23
24 immediately behind the melt pool. The temperature of each image pixel was approximated

an
25
26 using calibrated surface temperature and resultant RGB values. A defect-free deposit
27
28
gradually decreases in temperature and provides a means to generate a reference cooling
M
29
30
31 curve for a given set of process parameters. Defects, such as a cracks and porosity, can lead
32
33
d

34 to an increased temperature gradient in the vicinity of the defective region due to thermal
35
te

36 constriction/spreading resistance. This then leads to deviation from the reference cooling
37
38
curve – alerting end-users to the presence of possible defects. This method can be used to
p

39
40
41 halt a faulty DLD process in order to save both material and labor costs.
ce

42
43 Due to sensible bulk heating in the part during DLD, the laser power should
44
45
Ac

46 decrease with successive layers in order to achieve better melt pool shape control. Wang
47
48 and Felicelli [79] numerically predicted a drop on-the-order of 5% of laser power per pass
49
50
51 after a certain degree of bulk heating is achieved. They also demonstrated that the required
52
53 laser power for melt pool control decreases with time during the deposition of a layer.
54
55
56
Wang et al. [163] further found that a melt pool with constant shape for the entire duration
57
58 of a thin wall build during DLD provides for a laser power that depends linearly (with
59
60
61
62
63
64 Page 50 of 103
65
1
2
3
4 negative slope) with deposition layer number for a wide range of traverse speeds. This
5
6
7 trend is illustrated in Fig. 21 for various traverse speeds. For LENS of stainless steel, it
8
9 was found that the laser power decrease interval is near constant and independent of travel
10
11
12 speed and was approximately 16.7 W/layer [163]. It was also found that as the laser travel

t
13

ip
14 speed increases, the supplied laser power should increase. Peyre et al. [164] demonstrated
15
16
that the melt pool size increases with growing layers for a DLD thin wall of Ti-6Al-4V –

cr
17
18
19 further indicating the need for layer-variant laser power.

us
20
21 Tang and Landers [155] investigated closed-loop thermal control for single-nozzle
22
23
24 DLD of H13 tool steel. The melt pool temperature was monitored using a pyrometer and

an
25
26 an empirical model was developed to adjust laser power to abide a reference temperature
27
28
and maintain a less time-variant temperature to increase part quality. An empirical thermal
M
29
30
31 controller was designed based on a first-order melt pool temperature transfer function with
32
33
d

34 the form:
35
te

36 𝐾𝑡
37 𝑇(𝑠) = 𝑉 𝛼 𝑄𝛽 𝑀𝛾 (𝑠) (3)
𝜏𝑠 + 1
38
p

39
40 where the thermal time constant, τ, was experimentally measured to be approximately 30
41
ce

42 ms. The parameters V, Q and M correspond to the traverse speed, laser power and powder
43
44
45 feed rate, respectively. Experimental data was fitted to determine the unknown parameters:
Ac

46
47 𝐾𝑡 , α, β and γ. Experiments demonstrated that the thermal control method could track time-
48
49
50
varying and constant reference temperature for both constant and transient operating
51
52 parameters. Using the thermal control, bulk heating effects along the traverse and build-
53
54 height direction were reduced. However, when tested for multi-track deposition, it was
55
56
57 found that a wavy, non-uniform morphology was produced and that the melt pool width
58
59 changed along the track path. This was attributed to the first-order nature of the controlling
60
61
62
63
64 Page 51 of 103
65
1
2
3
4 transfer function (e.g. Eq. (3)). More robust control is required for multi-track DLD, and
5
6
7 higher-order modeling terms are needed to describe more complex melt pool morphologies
8
9 with regard to net heat input. They also indicate that dissimilar melt pool sizes can have
10
11
12 the same temperature and this can further impact the controller effectiveness in ensuring

t
13

ip
14 uniform track morphology [155].
15
16
Gaumann et al. [165] investigated the control of a single-nozzle DLD for various

cr
17
18
19 supperalloys for ensuring epitaxial growth of columnar dentrites during melt pool

us
20
21 solidification. The process parameters were selected for partial re-melting of the substrate
22
23
24 (build plate) to ensure epitaxy (deposition of a crystalline layer on a crystalline substrate)

an
25
26 and the lack of nucleating/growing equiaxed grains. Such control on microstructure and
27
28
epitaxy is of importance when using DLD for part repair or cladding.
M
29
30
31
32
33
d

34 5.2. Deposition Control


35
te

36 The powder flow rate is crucial for maintaining uniform layer height. The difficulty
37
38
in real-time powder flow control arises from the lack of sensors, or techniques for reliable
p

39
40
41 measurement of the powder flow during the build, Many sensing systems have been
ce

42
43 developed to solve the powder flow rate measurement problem, including the weight-loss
44
45
Ac

46 metering system, optoelectronic sensor [166], compact pressure sensor [167], etc. Each of
47
48 these methods has their own limitations: the weight-loss metering system may not be
49
50
51 accurate for measuring low powder flow rates (e.g., 5–15 g/min); the optoelectronic sensor
52
53 cannot be used for in-process measurement since it cannot sustain the high temperatures;
54
55
56
the compact pressure sensor relies on using a screw feed mechanism with compressed air
57
58 as the carrier gas.
59
60
61
62
63
64 Page 52 of 103
65
1
2
3
4
5
6
7 Closed loop optical feedback may have first been integrated into DLD by
8
9 Mazumder et al. [67,78,168] using a ‘reflective topography’ technique. The optical
10
11
12 feedback assembly consisted of three evenly-spaced photodetectors within the laser focal

t
13

ip
14 plane and aligned to observe the melt pool region. The primary goal was to control the
15
16
deposit thickness/height. This height controller demonstrated the ability to sense when the

cr
17
18
19 layer is building up higher in the overlapping region - shuting off the laser until it has

us
20
21 passed the area of excess build-up.
22
23
24 Kummailil et al. [169] studied the effects of process parameters on the deposition

an
25
26 of Ti-6Al-4V. It was found that deposition can be related to powder feed rate and energy
27
28
per unit area through a power relationship – and this can help in choosing an optimal
M
29
30
31 deposition layer thickness. Neela and De [170] proposed a non-dimensional number,
32
33
d

34 termed a non-dimensional specific energy index – specific to LENS, as:


35
te

36 𝑄
37 𝜋𝑟𝑏2 𝑉
38 𝑁sp = (4)
𝜌𝐶(𝑇L − 𝑇a ) + 𝜌𝐿
p

39
40
41 where rb is the radius of the laser beam, 𝜌 is density, C is the specific heat capacity,
ce

42
43
44 TL is the liquidus temperature, Ta is the ambient temperature and L is the latent heat of
45
Ac

46 fusion [171]. They demonstrated that this non-dimensional index should be reduced as the
47
48
49
build height increases [170] – in order to reduce over-melting. In essence, the heat input
50
51 per unit length needs to be decreased. Note that variable traverse speed can also be utilized
52
53 for controlling track morphology during DLD. Pi et al. [172] used a variable speed strategy
54
55
56 during DLD builds to counteract bulk heating effects on middle-region deposition
57
58 morphology. Currently, most DLD research employs a constant powder flow rate. For
59
60
61
62
63
64 Page 53 of 103
65
1
2
3
4 example, Xing et al. [173] used an infrared photodetector and CCD camera for monitoring
5
6
7 and controlling the melt pool size and thermal history during the DLD process for improved
8
9 layer height control. The CCD camera was used with a line-structure laser to monitor layer
10
11
12 height, while the powder feed rate was also monitored, and held constant, by employing a

t
13

ip
14 customized optoelectronic sensor. This approach to feedback control of DLD was shown
15
16
to effectively improve clad tolerances and build features. However, using the constant

cr
17
18
19 powder flow rate may result in non-uniform layer thickness where the powder nozzle

us
20
21 accelerates/decelerates at the end of each layer. Tang et al. [167] developed a variable
22
23
24 powder flow rate control to maintain uniform/consistent track morphology even as the

an
25
26 DLD system decelerates and accelerates. The authors demonstrated that by adjusting the
27
28
powder flow at the corners and on the edges of the part, layer height variation was reduced
M
29
30
31 substantially compared to the parts fabricated with a constant powder flow rate.
32
33
d

34 Other than controlling the powder flow rate, Fearon and Watkins [174] provide a
35
te

36 means to accomplish non-feedback control for consistent layer height by altering the
37
38
powder stream shape during the DLD process – for various laser powers, feed rates and
p

39
40
41 traverse speeds. This method can control the layer height by altering the powder efficiency
ce

42
43 within a fixed vertical plane located near the deposition head. Thus, the position of the
44
45
Ac

46 deposition head can affect the layer height, allowing the layer height to be controlled by a
47
48 program that indicates the incremental step height per deposition layer. However, by
49
50
51 limiting the amount of powder that is deposited, the deposition efficiency of DLD will
52
53 significantly reduce, and the process time will be prolonged.
54
55
56
57
58
59
60
61
62
63
64 Page 54 of 103
65
1
2
3
4 6. Conclusions and Ongoing Challenges
5
6
7 An overview of the microstructure and post-manufacture, mechanical
8
9 characteristics of parts fabricated via Direct Laser Deposition (DLD), as well as the control
10
11
12 and optimization of the DLD process has been presented. The inter-related relationships

t
13

ip
14 among process parameters, thermal history, microstructure, and fatigue behavior of DLD
15
16
components are presented schematically in Fig. 22.

cr
17
18
19 As presented in Fig. 22, process parameters affect the cooling rate, thermal

us
20
21 gradients, and generally, the thermal history of DLD parts. The complex thermal history of
22
23
24 the deposited part governs solidification, and consequently, the resultant microstructure,

an
25
26 porosity, and residual stress formation. Mechanical properties, especially fatigue
27
28
resistance, are extremely sensitive to microstructure, porosity, and residual stress within
M
29
30
31 DLD components. Therefore, microstructure sensitive fatigue models, relating the
32
33
d

34 microstructural features to fatigue resistance of material, can be readily employed to predict


35
te

36 the fatigue life of DLD parts. Such analytical models can be complemented by using finite
37
38
element analysis (FEA), in which critical elements with higher stresses/strain and
p

39
40
41 possibility of fatigue failure can be determined.
ce

42
43 In order to achieve enhanced fatigue life for the critical elements of the DLD parts,
44
45
Ac

46 it is essential to understand and characterize how DLD process parameters affect thermal
47
48 history, solidification, and eventually microstructural/mechanical properties. This remains
49
50
51 an open research area due to the large number of process parameters (e.g., laser power,
52
53 traverse speed, powder feed rate, layer thickness, hatching pitch, scanning pattern, etc.)
54
55
56
involved during DLD. Most of the existing studies seek only optimal process parameters
57
58 via extensive experimental or simulation. A major limitation of this approach is that the
59
60
61
62
63
64 Page 55 of 103
65
1
2
3
4 resulting optimal process parameters may not be useful when experimental conditions (e.g.,
5
6
7 process or material) change – resulting in new experiments to-be-conducted from scratch.
8
9 Further research is needed to (1) leverage the information from prior similar studies and
10
11
12 (2) systematically characterize the relation between process parameters and part features

t
13

ip
14 so that the DLD process can be optimized in a more efficient manner.
15
16
This challenge of process optimization is further compounded by the interactions

cr
17
18
19 among DLD parameters. In reality, it may not be practical to incorporate all process

us
20
21 parameters in either experimental or analytical studies. Ignorance of such higher-order
22
23
24 interaction effect, taking place during the DLD process, causes systematic uncertainty in

an
25
26 the resulting models and experimental results. Process uncertainty is associated with the
27
28
initial (latent heat exchange) and evolutionary (dendritic) solicitation, and conductive,
M
29
30
31 convective, and radiative heat transfer. For instance, the spatial/temporal scale for DLD is
32
33
d

34 relatively small for conduction heat transfer and thus thermal responses can be difficult to
35
te

36 measure and model — especially for the material (which is not detectable). Such
37
38
uncertainties will not only affect the microstructure but also the mechanical features of the
p

39
40
41 fabricated parts. In addition, there lies considerable uncertainty in melt pool depth (and
ce

42
43 other dimensions) due to uncertain heat transfer and fluid/part wetting behavior (contact
44
45
Ac

46 angles unknown). Further research is needed to incorporate uncertainty when optimizing


47
48 the DLD process.
49
50
51 Due to the many process parameters and the accompanied process uncertainty,
52
53 deposited parts experience a complex/hard-to-measure thermal history. This non-uniform
54
55
56
thermal history, accompanied with high thermal gradients near the HAZ, results in
57
58 undesirable, material-dependent residual stresses and distortions within the DLD parts.
59
60
61
62
63
64 Page 56 of 103
65
1
2
3
4 Microstructure evolution away from the HAZ is typically unavoidable due to thermal
5
6
7 cycling effects incurred during open-loop controlled DLD. The majority of research has
8
9 focused on controlling process parameters (e.g. laser power, traverse speed, etc.) to
10
11
12 maintain an optimal melt pool size – for enabling a more uniform temperature distribution

t
13

ip
14 and microstructure to reduce residual stress and distortions. However, more recent studies
15
16
show that the melt pool area alone may not be a sufficient indicator to achieve desired part

cr
17
18
19 quality. Other process variables, such as layer height, should be monitored and controlled

us
20
21 simultaneously.
22
23
24 In general, mechanical properties – including: tensile properties, hardness, impact

an
25
26 resistance, etc. - of DLD parts are comparable to conventionally-fabricated materials due
27
28
to the relatively high cooling rates experienced during DLD resulting in finer
M
29
30
31 microstructure. However, defects such as pores and inclusions - inherent consequences of
32
33
d

34 non-optimal DLD techniques - may be detrimental to mechanical properties, in particular,


35
te

36 fatigue resistance.
37
38
Anisotropy has been reported in both tensile and fatigue behavior of DLD
p

39
40
41 components during mechanical testing. Samples deposited in the direction along (e.g.
ce

42
43 parallel to) the length of the tensile specimens (e.g. horizontal) typically exhibit higher
44
45
Ac

46 tensile strength than the layers deposited perpendicular to the length of the tensile
47
48 specimens (vertical). However, the effect of building orientation on fatigue properties of
49
50
51 DLD parts are not yet well understood. Non-uniform, unsteady cooling rates, which can
52
53 result in orientation defects and weak metallurgical bonds between layers, are the main
54
55
56
reason for the anisotropy observed in DLD parts. Hot Isostatic Pressing (HIP) can
57
58
59
60
61
62
63
64 Page 57 of 103
65
1
2
3
4 significantly reduce the effect of building orientation on tensile and fatigue properties of
5
6
7 laser-deposited parts by eliminating directional porosities.
8
9 Microstructure (e.g. grain morphology and size) and defects (e.g. inclusions and
10
11
12 pores) are influential parameters on the fatigue behavior of DLD parts. Microstructural

t
13

ip
14 features have a complicated influence on the fatigue behavior. Finer microstructures, due
15
16
to higher density of slip bands, usually show better crack initiation resistance during the

cr
17
18
19 high cycle fatigue (HCF) regime as compared to coarser microstructures. However, coarser

us
20
21 microstructures exhibit more resistance to crack propagation, and therefore, have better
22
23
24 low cycle fatigue (LCF) behavior by promoting a rougher crack path. Defects behave as

an
25
26 crack initiation sites by creating stress concentrations at their walls; thus, significantly
27
28
reducing the fatigue strength especially during HCF where the total fatigue lifetime is
M
29
30
31 dominated by crack initiation life. More detrimental effects on fatigue strength are expected
32
33
d

34 from larger pores and/or the ones closer to the surface.


35
te

36 Microstructure sensitive fatigue models have the potential to predict fatigue


37
38
behavior of laser-deposited parts based on microstructure contents (e.g. grain size and
p

39
40
41 orientation) and defects (e.g. pores and inclusion size, etc.). However, more investigation
ce

42
43 is needed to better understand the effects of process parameters on the microstructural
44
45
Ac

46 evolution and resultant mechanical behavior of DLD components.


47
48 Based on this overview, there are still significant gaps in the literature for fully
49
50
51 understanding the relationship between the process parameters, thermal history,
52
53 solidification, microstructure, and mechanical properties of DLD techniques. Some
54
55
56
potential research topics to advance the state-of-the-art in DLD with respect to
57
58
59
60
61
62
63
64 Page 58 of 103
65
1
2
3
4 enhancing/understanding DLD part mechanical behavior and modeling are listed as
5
6
7 follows:
8
9 1) Since fatigue cracks usually initiate from plastic straining in localized regions, plastic
10
11
12 deformation is an important aspect of the fatigue process. Therefore, cyclic strain-

t
13

ip
14 controlled tests can better characterize the fatigue behavior of a material, especially in
15
16
the LCF regime. Less attention has been paid to cyclic strain-controlled testing as

cr
17
18
19 compared to traditional load or stress controlled testing as demonstrated in the available

us
20
21 literature. Therefore, generating comprehensive experimental data to better understand
22
23
24 the cyclic strain behavior and develop strain-based fatigue models for DLD

an
25
26 components appears to be critical.
27
28
2) Building orientation of parts fabricated via DLD may affect their thermal history, and
M
29
30
31 consequently, microstructure - as well as their post-manufactured mechanical behavior.
32
33
d

34 As a result, a comprehensive study on the effects of building orientation on tensile and


35
te

36 fatigue behavior is needed.


37
38
3) Multiaxial loading is common in many industrial applications. Even under uniaxial
p

39
40
41 nominal loading, multiaxial stresses may exist due to stress concentrations or residual
ce

42
43 stresses [175]. Therefore, it is essential to account for the cyclic multiaxial
44
45
Ac

46 stresses/strains in design by means of an appropriate multiaxial fatigue model. Based


47
48 on an exhaustive survey of the open literature, there is no report available that addresses
49
50
51 the multiaxial fatigue behavior of DLD components.
52
53 4) The mean stress, which causes cyclic ratcheting and additional undesirable excess
54
55
56
deformation, has significant influence on fatigue behavior. Various in-service
57
58 components and structures are subjected to variable amplitude spectrum loading with
59
60
61
62
63
64 Page 59 of 103
65
1
2
3
4 mean stresses. Therefore, effects of mean stresses and variable amplitude stress/strain
5
6
7 loading on fatigue behavior of DLD parts should be investigated.
8
9 5) Developing a microstructural sensitive fatigue models, such as the multi-stage fatigue
10
11
12 (MSF) model, that relates the effect of surface condition and microstructural content

t
13

ip
14 (e.g. porosity, pores and inclusions size, shape and their distance, grain size and
15
16
orientation, etc.) to fatigue properties of DLD parts is essential.

cr
17
18
19 6) There is a significant lack of combined studies relating the thermal history during

us
20
21 deposition to fatigue behavior of deposited component. Such studies can connect the
22
23
24 thermal history and solidification to microstructure, and mechanical behavior of DLD

an
25
26 parts. Knowing such relationships and by applying a microstructural sensitive fatigue
27
28
model, it may be possible to optimize DLD process parameters in order to reach a
M
29
30
31 specific fatigue behavior based on the application, as presented in Fig. 22.
32
33
d

34 7) Online control of DLD will require some degree of feed-forward prediction and thus
35
te

36 high-fidelity computational models must be used in order to solve the complex DLD
37
38
physics within a very short time frame.
p

39
40
41 8) Real-time control and the effective modeling of the DLD process, and inherent melt
ce

42
43 pool, will depend on the current-state of computational modeling and resources. In
44
45
Ac

46 order to better model the melt pool morphology and coupled heat transfer, Direct
47
48 Numerical Simulation (DNS) can be used, but significant computer resources are
49
50
51 required.
52
53 9) One drawback of imposing real-time control over deposition is that the required control
54
55
56
systems are not part of the equipment and must be built or purchased by the user—a
57
58 nontrivial task. Although adding a level of control to DLD is desirable for enhanced
59
60
61
62
63
64 Page 60 of 103
65
1
2
3
4 part quality, an added benefit is the use of the monitoring devices for data mining. Data
5
6
7 collected from various DLD processes can be utilized for various research and
8
9 development purposes, e.g. – model verifications, process characterization and process
10
11
12 optimization. Such data, collected via use of monitoring equipment can then be used

t
13

ip
14 to significantly reduce the time required to ‘learn’ about optimal process parameters
15
16
for an effective DLD part.

cr
17
18
19

us
20
21
Acknowledgements
22
23
24 The authors would like to acknowledge the support from Mississippi State

an
25
26 University’s (MSU) Center for Advanced Vehicular Systems (CAVS) and Office of
27
28
Research and Economic Development (ORED).
M
29
30
31
32
33
d

34
35
te

36
37
38
p

39
40
41
ce

42
43
44
45
Ac

46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 61 of 103
65
1
2
3
4 References
5
6
7
8 [1] M.L. Griffith, D.M. Keicher, C.L. Atwood, J.A. Romero, J.E. Smugeresky, L.D.
9 Harwell, et al., Free Form Fabrication of Metallic Components Using Laser
10
11
Engineered Net Shaping (LENSTM), in: Proc. 7th Solid Free. Fabr. Symp., Austin,
12 USA, 1996: pp. 125–132.

t
13

ip
14 [2] V. Weerasinghe, W. Steen, Laser cladding by powder injection, Proc. 1st Int.
15 Conf. Lasers Manuf. (1983) 125–132.
16

cr
17
18 [3] V. Weerasinghe, W. Steen, Laser cladding with blown powder, Met. Constr. 19
19 (1987) 581–5.

us
20
21
22
[4] J. Sears, Direct laser powder deposition-’State of the Art', No. KAPL-P-000311;
23 K99089. Knolls At. Power Lab., Nis. NY. (1999).
24

an
25 [5] M. McLean, Laser direct casting high nickel alloy components, Adv. Powder
26 Metall. Part. Mater. 3 (1997) 21..
27
28
M
29 [6] J. Mazumder, J. Choi, J. and Nagarathnam, K., Koch, D. Hetzner, The direct metal
30 deposition of H13 tool steel for 3D components, JOM. 49 (1997) 55–60.
31
32
33
[7] G. Lewis, R. Nemec, J. Milewski, D. Thoma, Directed light fabrication, No. LA-
d

34 UR--94-2845; CONF-9410189--2. Los Alamos Natl. Lab., NM (United States).


35 (1994).
te

36
37 [8] J. Milewski, G. Lewis, D. Thoma, Directed light fabrication of a solid metal
38
hemisphere using 5-axis powder deposition, J. Mater. Process. Technol. 75 (1998)
p

39
40 165–172.
41
ce

42 [9] X. Wu, J. Liang, J. Mei, C. Mitchell, P.S. Goodwin, W. Voice, Microstructures of


43
44
laser-deposited Ti–6Al–4V, Mater. Des. 25 (2004) 137–144.
45
Ac

46 [10] F. Arcella, F. Froes, Producing titanium aerospace components from powder using
47 laser forming, Jom. 52 (2000) 28–30.
48
49
50 [11] J. Fessler, R. Merz, Laser deposition of metals for shape deposition manufacturing,
51 Proc. Solid Free. Fabr. Symp. Univ. Texas Austin. (1996) 117–124.
52
53 [12] D.M. Keicher, W.D. Miller, LENS moves beyond RP to direct fabrication, Met.
54
55
Powder Rep. 53 (1998) 26–8.
56
57 [13] M. Griffith, M. Schlienger, L. Harwell, Thermal behavior in the LENS process,
58 No. SAND--98-1850C; CONF-980826--. Sandia Natl. Labs., Albuquerque, NM
59 (United States). (1998).
60
61
62
63
64 Page 62 of 103
65
1
2
3
4 [14] L. Xue, M. Islam, Free-form laser consolidation for producing functional metallic
5
6 components, Laser Inst. Am. Laser Mater. Process. 84 (1998).).
7
8 [15] L. Xue, M. Islam, Free-form laser consolidation for producing metallurgically
9 sound and functional components, J. Laser Appl. (2000).
10
11
12 [16] J. Choi, Process and Properties Control in Laser Aided Direct Metal/Materials

t
13 Deposition Process, in: Manufacturing, ASME, 2002: pp. 81–89.

ip
14
15 [17] S. Ghosh, J. Choi, Three-dimensional transient finite element analysis for residual
16
stresses in the laser aided direct metal/material deposition process, J. Laser Appl.

cr
17
18 17 (2005) 144–158.
19

us
20 [18] J. Choi, Y. Chang, Characteristics of laser aided direct metal/material deposition
21
22
process for tool steel, Int. J. Mach. Tools Manuf. 54 (2005) 597–607.).
23
24 [19] R. Dwivedi, R. Kovacevic, An expert system for generation of machine inputs for

an
25 laser-based multi-directional metal deposition, Int. J. Mach. Tools Manuf. 46
26 (2006) 1811–1822.
27
28
M
29 [20] R. Dwivedi, R. Kovacevic, Process Planning for Multi-Directional Laser-Based
30 Direct Metal Deposition, Proc. Inst. Mech. Eng. Part C J. Mech. Eng. Sci. 219
31 (2005) 695–707.
32
33
d

34 [21] F. Liou, K. Slattery, M. Kinsella, Applications of a hybrid manufacturing process


35 for fabrication of metallic structures, Rapid Prototyp. J. 13 (2007) 236–244..
te

36
37 [22] J. Zhang, F. Liou, Adaptive slicing for a five-axis laser aided manufacturing
38
process, Proc. 2001 ASME Des. Autom. Conf. (2001).
p

39
40
41 [23] J. Zhang, Adaptive Slicing for a Multi-Axis Laser Aided Manufacturing Process, J.
ce

42 Mech. Des. 126 (2004) 254.


43
44
45 [24] P.A. Kobryn, S.L. Semiatin, Mechanical Properties of Laser-Deposited Ti-6Al-4V,
Ac

46 Solid Free. Fabr. Proceedings. Austin. (2001) 179–186.


47
48 [25] P. Kobryn, S. Semiatin, Microstructure and texture evolution during solidification
49
50 processing of Ti–6Al–4V, J. Mater. Process. Technol. (2003).
51
52 [26] P.C. Collins, R. Banerjee, S. Banerjee, H.L. Fraser, Laser deposition of
53 compositionally graded titanium–vanadium and titanium–molybdenum alloys,
54
55
Mater. Sci. Eng. A. 352 (2003) 118–128.
56
57 [27] S. Kelly, S. Kampe, Microstructural evolution in laser-deposited multilayer Ti-
58 6Al-4V builds: Part II. Thermal modeling, Metall. Mater. Trans. A. 35 (2004)
59 1869–1879.
60
61
62
63
64 Page 63 of 103
65
1
2
3
4 [28] R. Grylls, LENS Process White Paper : Fatigue Testing of LENS Ti-6-4, Optomec
5
6 Intern. Report, Addit. Manuf. Syst. from Nano to MACRO. (2005) 2–6.
7
8 [29] J. Kummailil, C. Sammarco, D. Skinner, C. a. Brown, K. Rong, Effect of Select
9 LENSTM Processing Parameters on the Deposition of Ti-6Al-4V, J. Manuf.
10
11
Process. 7 (2005) 42–50.
12

t
13 [30] H.C.D.G. Iii, Development of Laser Fabricated Ti-6Al-4V, NASA Tech. Rep.

ip
14 (2006).
15
16
[31] G. Dinda, L. Song, J. Mazumder, Fabrication of Ti-6Al-4V scaffolds by direct

cr
17
18 metal deposition, Metall. Mater. Trans. A. 39 (2008) 2914–2922.
19

us
20 [32] J.O. J. Alcisto, A. Enriquez, H. Garcia, S. Hinkson, T. Steelman, E. Silverman, P.
21
22
Valdovino, H. Gigerenzer, J. Foyos, and O.S.E.-S. J. Dorey, K. Karg, T.
23 McDonald, Tensile Properties and Microstructures of Laser-Formed Ti-6Al-4V,
24 JMEPEG. 20 (2011) 203–212.

an
25
26 [33] G.Q. Wu, C.L. Shi, W. Sha, A.X. Sha, H.R. Jiang, Microstructure and high cycle
27
28 fatigue fracture surface of a Ti–5Al–5Mo–5V–1Cr–1Fe titanium alloy, Mater. Sci.
M
29 Eng. A. 575 (2013) 111–118.
30
31 [34] N. Bian, L., Thompson, S.M., Shamsaei, Mechanical Properties and
32
33
Microstructural Features of Direct Laser Deposited Ti-6Al-4V, JOM. (2015) in
d

34 press.
35
te

36 [35] J. Brooks, C. Robino, T. Headley, S. Goods, M. Griffith, Microstructure and


37 property optimization of LENS deposited H13 tool steel, Proc. Solid Free. Fabr.
38
Symp. (1999).
p

39
40
41 [36] L. Costa, R. Vilar, T. Réti, Simulating the Effects of Substrate Pre-Heating on the
ce

42 Final Structure of Steel Parts Built by Laser Powder Deposition, in: Proc. 15th
43
44
Solid Free. Fabr. Symp., Austin, USA, 2004: pp. 643–654.
45
Ac

46 [37] H. El Kadiri, L. Wang, M.F. Horstemeyer, R.S. Yassar, J.T. Berry, S. Felicelli, et
47 al., Phase transformations in low-alloy steel laser deposits, Mater. Sci. Eng. A. 494
48 (2008) 10–20.
49
50
51 [38] Y. Xue, a. Pascu, M.F. Horstemeyer, L. Wang, P.T. Wang, Microporosity effects
52 on cyclic plasticity and fatigue of LENSTM-processed steel, Acta Mater. 58 (2010)
53 4029–4038.
54
55
56 [39] Ј.К. Mickovski, I.Ј. Lazarev, Ј. Lazarev, Microstructure case study of LENStm
57 processed cilinder from AISI H13 steel, J. Technol. Plast. 35 (2010) 61–74.
58
59
60
61
62
63
64 Page 64 of 103
65
1
2
3
4 [40] D. Wu, X. Liang, Q. Li, L. Jiang, Laser Rapid Manufacturing of Stainless Steel
5
6 316L/Inconel718 Functionally Graded Materials: Microstructure Evolution and
7 Mechanical Properties, Int. J. Opt. 2010 (2010) 1–5. doi:10.1155/2010/802385.
8
9 [41] P. Ganesh, R. Kaul, G. Sasikala, H. Kumar, S. Venugopal, P. Tiwari, et al., Fatigue
10
11
Crack Propagation and Fracture Toughness of Laser Rapid Manufactured
12 Structures of AISI 316L Stainless Steel, Metallogr. Microstruct. Anal. 3 (2014)

t
13 36–45.

ip
14
15 [42] P.L. Blackwell, The mechanical and microstructural characteristics of laser-
16
deposited IN718, J. Mater. Process. Technol. 170 (2005) 240–246.

cr
17
18
19 [43] C.P. Paul, P. Ganesh, S.K. Mishra, P. Bhargava, J. Negi, a. K. Nath, Investigating

us
20 laser rapid manufacturing for Inconel-625 components, Opt. Laser Technol. 39
21
22
(2007) 800–805.
23
24 [44] E. Amsterdam, G.A. Kool, High Cycle Fatigue of Laser Beam Deposited Ti-6Al-

an
25 4V and Inconel 718, ICAF, Bridg. Gap between Theory Oper. Pract. Springer
26 Netherlands. (2009) 1261–1274.
27
28
M
29 [45] G.P. Dinda, a. K. Dasgupta, J. Mazumder, Laser aided direct metal deposition of
30 Inconel 625 superalloy: Microstructural evolution and thermal stability, Mater. Sci.
31 Eng. A. 509 (2009) 98–104.
32
33
d

34 [46] P. Ganesh, R. Kaul, C.P. Paul, P. Tiwari, S.K. Rai, R.C. Prasad, et al., Fatigue and
35 fracture toughness characteristics of laser rapid manufactured Inconel 625
te

36 structures, Mater. Sci. Eng. A. 527 (2010) 7490–7497.


37
38
[47] G. Puppala, A. Moitra, S. Sathyanarayanan, R. Kaul, G. Sasikala, R.C. Prasad, et
p

39
40 al., Evaluation of fracture toughness and impact toughness of laser rapid
41 manufactured Inconel-625 structures and their co-relation, 13th Int. Conf. Fract.
ce

42 Beijing, China, June 16–21. 625 (2013).


43
44
45 [48] G.D.J. Ram, B.E. Stucker, A Feasibility Study of LENS® Deposition of CoCrMo
Ac

46 Coating on a Titanium Substrate, J. Manuf. Sci. Eng. 130 (2008) 024503.


47
48 [49] G.D. Janaki Ram, C.K. Esplin, B.E. Stucker, Microstructure and wear properties of
49
50 LENS deposited medical grade CoCrMo., J. Mater. Sci. Mater. Med. 19 (2008)
51 2105–11.
52
53 [50] F.A. España, V.K. Balla, S. Bose, A. Bandyopadhyay, Design and fabrication of
54
55
CoCrMo alloy based novel structures for load bearing implants using laser
56 engineered net shaping, Mater. Sci. Eng. C. 30 (2010) 50–57.
57
58 [51] M. Hedges, N. Calder, Near net shape rapid manufacture & repair by LENS, Cost
59 Eff. Manuf. via Net-Shape Process. Meet. Proc. RTO-MP-AVT-139. 13 (2006).
60
61
62
63
64 Page 65 of 103
65
1
2
3
4 [52] R. Grylls, Laser Engineered Net Shaping LENS ® Phase II, 2005.
5
6
7 [53] W. Liu, J.N. DuPont, Fabrication of functionally graded TiC/Ti composites by
8 Laser Engineered Net Shaping, Scr. Mater. 48 (2003) 1337–1342.
9
10
11
[54] Y.T. Pei, J.T.M. De Hosson, Functionally graded materials produced by laser
12 cladding, Acta Mater. 48 (2000) 2617–2624.

t
13

ip
14 [55] M.T. Ensz, M.L. Griffith, D.E. Reckaway, Critical issues for functionally graded
15 material deposition by laser engineered net shaping (LENS), Proc. MPIF Laser
16
Met. Depos. Conf. San Antonio TX. (2002).

cr
17
18
19 [56] L. Costa, R. Vilar, Laser powder deposition, Rapid Prototyp. J. 15 (2009) 264–

us
20 279.
21
22
23 [57] C. Selcuk, Laser metal deposition for powder metallurgy parts, Powder Metall. 54
24 (2011) 94–99.

an
25
26 [58] L. Xue, J. Chen, M. Islam, J. Pritchard, Laser consolidation of Ni-base IN-738
27
28 superalloy for repairing gas turbine blades, ICALEO 2000 Laser Mater. Process.
M
29 Conf. (2000).
30
31 [59] M. Bohrer, H. Basalka, W. Birner, K. Emiljanow, Turbine blade repair with laser
32
33
powder fusion welding and shape recognition, Int. Conf. Met. Powder Depos.
d

34 Rapid Manuf. San Antonio. (2002).


35
te

36 [60] M. Gäumann, S. Henry, F. Cleton, Epitaxial laser metal forming: analysis of


37 microstructure formation, Mater. Sci. Eng. 271 (1999) 232–241.
38
p

39
40 [61] G. Baskes, E.W. Kreutz, A. Gasser, K. Wissenbach, R. Poprawe, Laser-shape
41 reconditioning and manufacturing of tools and machine parts, in: ICALEO’98,
ce

42 1998.
43
44
45 [62] D.D. Gu, W. Meiners, K. Wissenbach, R. Poprawe, Laser additive manufacturing
Ac

46 of metallic components: materials, processes and mechanisms, Int. Mater. Rev. 57


47 (2012) 133–164.
48
49
50 [63] L. Costa, R. Vilar, Laser Powder Deposition, Rapid Prototyp. J. 15 (2009) 264–
51 279.
52
53 [64] K. Weisheit, A., Gasser, A., Backes, G., Jambor, T., Pirch, N., Wissenbach, Direct
54
55
Laser Cladding, Current Status and Future Scope of Application, in: Laser-
56 Assisted Fabr. Mater. Springer Berlin Heidelb., 2013: pp. 221–240.
57
58 [65] R. Vilar, Laser Powder Deposition, Compr. Mater. Process. Vol. 10 Adv. Addit.
59 Manuf. Tool. 10 (2014) 163–216.
60
61
62
63
64 Page 66 of 103
65
1
2
3
4 [66] W.E. Frazier, Metal Additive Manufacturing: A Review, J. Mater. Eng. Perform.
5
6 23 (2014) 1917–1928.
7
8 [67] J. Mazumder, D. Dutta, N. Kikuchi, A. Ghosh, Closed loop direct metal
9 deposition: art to part, Opt. Lasers Eng. 34 (2000) 397–414.
10
11
http://www.sciencedirect.com/science/article/pii/S0143816600000725 (accessed
12 May 6, 2014).

t
13

ip
14 [68] S. Bontha, N.W. Klingbeil, P. a. Kobryn, H.L. Fraser, Thermal process maps for
15 predicting solidification microstructure in laser fabrication of thin-wall structures,
16
J. Mater. Process. Technol. 178 (2006) 135–142.

cr
17
18
19 [69] B. Zheng, Y. Zhou, J.E. Smugeresky, J.M. Schoenung, E.J. Lavernia, Thermal

us
20 Behavior and Microstructural Evolution during Laser Deposition with Laser-
21
22
Engineered Net Shaping: Part I. Numerical Calculations, Metall. Mater. Trans. A.
23 39 (2008) 2228–2236.
24

an
25 [70] S. Bontha, N.W. Klingbeil, P. a. Kobryn, H.L. Fraser, Effects of process variables
26 and size-scale on solidification microstructure in beam-based fabrication of bulky
27
28 3D structures, Mater. Sci. Eng. A. 513-514 (2009) 311–318.
M
29
30 [71] R. Vilar, Laser cladding, J. Laser Appl. 11 (2001) 64–79.
31
32
33
[72] W. Hofmeister, M. Wert, J. Smugeresky, Investigation of solidification in the laser
d

34 engineered net shaping (LENSTM) process, JOM. 51 (1999) 1–6.


35
te

36 [73] W. Hofmeister, M. Griffith, M. Ensz, J. Smugeresky, Melt pool imaging for


37 control of LENS processing, Proc. Int. Conf. Met. Powder Depos. Rapid Manuf.
38
San Antonio, Texas, USA. (2002) 188–194.
p

39
40
41 [74] F.-J. Kahlen, A. Kar, Tensile Strengths for Laser-Fabricated Parts and Similarity
ce

42 Parameters for Rapid Manufacturing, J. Manuf. Sci. Eng. 123 (2001) 38.
43
44
45 [75] B. Zheng, Y. Zhou, J.E. Smugeresky, J.M. Schoenung, E.J. Lavernia, Thermal
Ac

46 Behavior and Microstructure Evolution during Laser Deposition with Laser-


47 Engineered Net Shaping: Part II. Experimental Investigation and Discussion,
48 Metall. Mater. Trans. A. 39 (2008) 2237–2245.
49
50
51 [76] J. Chen, L. Xue, Microstructural characteristics of laser-clad AISI P20 tool steel,
52 Proc. 1st Int. Surf. Eng. Congr. 13th IFHTSE Congr. Mater. Park ASM Int. (2002)
53 198.
54
55
56 [77] G. Link, Layered manufacturing of laser-deposited carbon steels, 2000.
57
58 [78] J. Mazumder, A. Schifferer, J. Choi, Direct materials deposition: designed macro
59 and microstructure, Mater. Res. Innov. 3 (1999) 118–131.
60
61
62
63
64 Page 67 of 103
65
1
2
3
4 [79] L. Wang, S. Felicelli, Process Modeling in Laser Deposition of Multilayer SS410
5
6 Steel, J. Manuf. Sci. Eng. 129 (2007) 1028.
7
8 [80] B. Baufeld, O. Van Der Biest, R. Gault, K. Ridgway, Manufacturing Ti-6Al-4V
9 Components by Shaped Metal Deposition: Microstructure and Mechanical
10
11
Properties, IOP Conf. Ser. Mater. Sci. Eng. 26 (2011) 012001.
12

t
13 [81] L. Costa, R. Vilar, T. Reti, A. Deus, Rapid tooling by laser powder deposition:

ip
14 Process simulation using finite element analysis, Acta Mater. (2005).
15
16
[82] ASTM Standard, ASTM E6-09be1 Standard Terminology Relating to Methods of

cr
17
18 Mechanical Testing, (2009).
19

us
20 [83] P. Rangaswamy, M.L. Griffith, M.B. Prime, T.M. Holden, R.B. Rogge, J.M.
21
22
Edwards, et al., Residual stresses in LENS ® components using neutron diffraction
23 and contour method, Mater. Sci. Eng. A. 399 (2005) 72–83.
24

an
25 [84] F. Liu, X. Lin, G. Yang, M. Song, J. Chen, W. Huang, Optics & Laser Technology
26 Microstructure and residual stress of laser rapid formed Inconel 718 nickel-base
27
28 superalloy, Opt. Laser Technol. 43 (2011) 208–213.
M
29
30 [85] K. Dai, L. Shaw, Distortion minimization of laser-processed components through
31 control of laser scanning patterns, Rapid Prototyp. J. 8 (2002) 270–276.
32
33
d

34 [86] J. Beuth, N. Klingbeil, The Role of Process Variables in Laser-Based Direct Metal
35 Solid Freeform Fabrication, (2001).
te

36
37 [87] M.L. Griffith, M.E. Schlienger, L.D. Harwell, M.S. Oliver, M.D. Baldwin, M.T.
38
Ensz, et al., Understanding thermal behavior in the LENS process, Mater. Des. 20
p

39
40 (1999) 107–113.
41
ce

42 [88] Z. Shuangyin, L. Xin, C. Jing, H. Weidong, Influence of Heat Treatment on


43
44
Residual Stress of Ti-6Al-4V Alloy by Laser Solid Forming, Rare Met. Mater.
45 Eng. 38 (2009) 774–778.
Ac

46
47 [89] A.J. Pinkerton, M. Karadge, W. Ul Haq Syed, L. Li, Thermal and microstructural
48 aspects of the laser direct metal deposition of waspaloy, J. Laser Appl. 18 (2006)
49
50 216.
51
52 [90] D. Keicher, J. Smugeresky, The laser forming of metallic components using
53 particulate materials, JOM. 49 (1997) 51–54..
54
55
56 [91] Y. Li, H. Yang, X. Lin, W. Huang, J. Li, Y. Zhou, The influences of processing
57 parameters on forming characterizations during laser rapid forming, Mater. Sci.
58 Eng. A. 360 (2003) 18–25.
59
60
61
62
63
64 Page 68 of 103
65
1
2
3
4 [92] W. Liu, J.N. Dupont, In-Situ Reactive Processing of Nickel Aluminides by Laser-
5
6 Engineered Net Shaping, 34 (2003) 2633–2641.
7
8 [93] J. Yu, X. Lin, L. Ma, J. Wang, X. Fu, J. Chen, et al., Influence of laser deposition
9 patterns on part distortion , interior quality and mechanical properties by laser
10
11
solid forming (LSF), Mater. Sci. Eng. A. 528 (2011) 1094–1104.
12

t
13 [94] A. Nickel, D. Barnett, F. Prinz, Thermal stresses and deposition patterns in layered

ip
14 manufacturing, Mater. Sci. Eng. A. (2001) 59–64.
15
16
[95] H. Wenbin, L. Yong Tsui, G. Haiqing, A study of the staircase effect induced by

cr
17
18 material shrinkage in rapid prototyping, Rapid Prototyp. J. 11 (2005) 82–89.
19

us
20 [96] A. Dolenc, I. Mäkelä, Slicing procedures for layered manufacturing techniques,
21
22
Comput. Des. 26 (1994) 119–126.
23
24 [97] Y. Yang, J.Y.H. Fuh, H.T. Loh, Y.S. Wong, A Volumetric Difference-based

an
25 Adaptive Slicing and Deposition Method for Layered Manufacturing, J. Manuf.
26 Sci. Eng. 125 (2003) 586.
27
28
M
29 [98] P. Singh, D. Dutta, Multi-Direction Slicing for Layered Manufacturing, J. Comput.
30 Inf. Sci. Eng. 1 (2001) 129.
31
32
33
[99] Y. Yang, J.Y.H. Fuh, H.T. Loh, Y.S. Wong, Minimizing staircase errors in the
d

34 orthogonal layered manufacturing system, IEEE Trans. Autom. Sci. Eng. 2 (2005)
35 276–284.
te

36
37 [100] J. Ruan, L. Tang, F.W. Liou, R.G. Landers, Direct Three-Dimensional Layer
38
Metal Deposition, J. Manuf. Sci. Eng. 132 (2010) 064502.
p

39
40
41 [101] W. Hofmeister, M. Griffith, Solidification in direct metal deposition by LENS
ce

42 processing, JOM. 53 (2001) 30–34.


43
44
45 [102] ASM Handbook, ASM Handbook Volume 3, Alloy Phase Diagrams., Mater. Park.
Ac

46 OH ASM Int. (1992).


47
48 [103] Nickel, Cobalt, and Their Alloys, ASM International, 2000.
49
50
51 [104] Y. Wang, S. Zhang, X. Tian, H. Wang, High-cycle fatigue crack initiation and
52 propagation in laser melting deposited TC18 titanium alloy, Int. J. Miner. Metall.
53 Mater. 20 (2013) 665–670.
54
55
56 [105] Z. Li, X. Tian, H. Tang, H. Wang, Low cycle fatigue behavior of laser melting
57 deposited TC18 titanium alloy, Trans. Nonferrous Met. Soc. China. 23 (2013)
58 2591–2597.
59
60
61
62
63
64 Page 69 of 103
65
1
2
3
4 [106] J.N. Sweet, E.P. Roth, M. Moss, Thermal conductivity of Inconel 718 and 304
5
6 stainless steel, Int. J. Thermophys. 8 (1987) 593–606.
7
8 [107] M. Zieliñska, M. Yavorska, M. Porêba, J. Sieniawski, Thermal properties of cast
9 nickel based superalloys, 44 (2010) 35–38.
10
11
12 [108] L. Avala, M. Bheema, P.K. Singh, R.K. Rai, S. Srivastava, Measurement of

t
13 Thermo Physical Properties of Nickel Based Superalloys, (2013) 108–112.

ip
14
15 [109] ASTM Standard, ASTM E23 - 12c, Standard Test Methods for Notched Bar
16
Impact Testing of Metallic Materials, (2012).

cr
17
18
19 [110] P. Ganesh, R. Kaul, H. Kumar, Metallurgical Characterization of Laser Fabricated

us
20 Structures of Engineering Alloys, Interact. Meet Util. Laser Technol. Ind. Med.
21
22
RRCAT. 22 (2011) 18–23.
23
24 [111] ASTM Standard, ASTM E1823 - 13, Standard Terminology Relating to Fatigue

an
25 and Fracture Testing, (2013).
26
27
28 [112] R.I. Stephens, A. Fatemi, R.R. Stephens, H.O. Fuchs, Metal Fatigue in
M
29 Engineering, 2nd Edition, Wiley. (2000) 496.
30
31 [113] R.A. Mesquita, G.E. Totten, Failure Analysis of Heat Treated Steel Components,
32
33
ASM Int. (2008).
d

34
35 [114] Chinease Standard, GB/T 15248, The test method for axial loading constant-
te

36 amplitude low-cycle fatigue of metallic materials, (2008).


37
38
[115] N. Shamsaei, A. Fatemi, Small Crack Growth Path and Rate under Combined
p

39
40 Stresses, CP2012. (2013) 171–182.
41
ce

42 [116] C. Lin, C. Ju, J.C. Lin, A comparison of the fatigue behavior of cast Ti–7.5 Mo
43
44
with cp titanium, Ti–6Al–4V and Ti–13Nb–13Zr alloys, Biomaterials. (2005).
45
Ac

46 [117] N. Shamsaei, a. Fatemi, Effect of hardness on multiaxial fatigue behaviour and


47 some simple approximations for steels, Fatigue Fract. Eng. Mater. Struct. 32
48 (2009) 631–646.
49
50
51 [118] F. Wang, Mechanical property study on rapid additive layer manufacture
52 Hastelloy® X alloy by selective laser melting technology, Int. J. Adv. Manuf.
53 Technol. 58 (2011) 545–551.
54
55
56 [119] O. Jin, S. Mall, Effects of microstructure on short crack growth behavior of Ti–
57 6Al–2Sn–4Zr–2Mo–0.1 Si alloy, Mater. Sci. Eng. A. (2003).
58
59
60
61
62
63
64 Page 70 of 103
65
1
2
3
4 [120] B. Baufeld, O. van der Biest, Mechanical properties of Ti-6Al-4V specimens
5
6 produced by shaped metal deposition, Sci. Technol. Adv. Mater. 10 (2009)
7 015008.
8
9 [121] K. Dragolich, N. DiMatteo, S. Henry, Fatigue data book: light structural alloys,
10
11
ASM Int. (1994).
12

t
13 [122] ASTM Standard, ASTM E399-09, Standard Test Method for Linear-Elastic Plane-

ip
14 Strain Fracture Toughness K Ic of Metallic Materials, (2009).
15
16
[123] M. Donachie, Titanium: a technical guide, in: ASM Int., 2000.

cr
17
18
19 [124] A. Handbook, Vol. 19, Fatigue and Fracture, ASM Int. (1996).

us
20
21
22
[125] ASTM Standard, ASTM E647-08, Standard Test Method for Measurement of
23 Fatigue Crack Growth Rates, (2008).
24

an
25 [126] ASTM Standard, ASTM E1820-11, Standard Test Method for Measurement of
26 Fracture Toughness, (2011).
27
28
M
29 [127] ASM Handbook, Aerospace Structural Metals, Purdue Univ. West Lafayette.
30 (1997).
31
32
33
[128] ASTM Standard, ASTM E1290-08e1, Standard Test Method for Crack-Tip
d

34 Opening Displacement (CTOD) Fracture Toughness Measurement (Withdrawn


35 2013), (2008).
te

36
37 [129] D.. McDowell, K. Gall, M.. Horstemeyer, J. Fan, Microstructure-based fatigue
38
modeling of cast A356-T6 alloy, Eng. Fract. Mech. 70 (2003) 49–80.
p

39
40
41 [130] J.D. Bernard, J.B. Jordon, M. Lugo, J.M. Hughes, D.C. Rayborn, M.F.
ce

42 Horstemeyer, Observations and modeling of the small fatigue crack behavior of an


43
44
extruded AZ61 magnesium alloy, Int. J. Fatigue. 52 (2013) 20–29.
45
Ac

46 [131] M. Lugo, J.B. Jordon, K.N. Solanki, L.G. Hector, J.D. Bernard, a. a. Luo, et al.,
47 Role of different material processing methods on the fatigue behavior of an AZ31
48 magnesium alloy, Int. J. Fatigue. 52 (2013) 131–143.
49
50
51 [132] Y. Xue, D. McDowell, M. Horstemeyer, Microstructure-based multistage fatigue
52 modeling of aluminum alloy 7075-T651, Eng. Fract. Mech. 74 (2007) 2810–2823.
53
54
55 [133] J.P. Strizak, H. Tian, P.K. Liaw, L.K. Mansur, Fatigue properties of type 316LN
56 stainless steel in air and mercury, J. Nucl. Mater. 343 (2005) 134–144.
57
58 [134] R.R.R. Unocic, A fundamental investigation of process efficiencies in the Laser
59
Engineered Net Shaping ( LENS ) solid freeform fabrication process, 2002.
60
61
62
63
64 Page 71 of 103
65
1
2
3
4 [135] R.R. Unocic, J.N. Dupont, Process Efficiency Measurements in the Laser
5
6 Engineered Net Shaping Process, 35 (2004) 143–152.
7
8 [136] J. Liu, L. Li, Effects of process variables on laser direct formation of thin wall,
9 Opt. Laser Technol. 39 (2007) 231–236. doi:10.1016/j.optlastec.2005.08.012.
10
11
12 [137] L. Peng, Y. Taiping, L. Sheng, L. Dongsheng, H. Qianwu, X. Weihao, et al.,

t
13 Direct laser fabrication of nickel alloy samples, Int. J. Mach. Tools Manuf. 45

ip
14 (2005) 1288–1294.
15
16
[138] A. Vasinonta, J. Beuth, M. Griffith, Process Maps for Laser Deposition of Thin-

cr
17
18 Walled Structures, in: Proc. 10th Solid Free. Fabr. Symp., Austin, USA, 1999: pp.
19 383–392.

us
20
21
22
[139] A. Vasinonta, J.L. Beuth, M.L. Griffith, A Process Map for Consistent Build
23 Conditions in the Solid Freeform Fabrication of Thin-Walled Structures, J. Manuf.
24 Sci. Eng. 123 (2001) 615.

an
25
26 [140] A. Vasinonta, Process maps for controlling residual stress and melt pool size in
27
28 laser-based SFF processes, Olid Free. Fabr. Proceedings. Proc. 2000 Solid Free.
M
29 Fabr. Symp. Austin. (2000) 200–208.
30
31 [141] A. Vasinonta, J.L. Beuth, R. Ong, Melt Pool Size Control in Thin-Walled and
32
33
Bulky Parts via Process Maps, Solid Free. Fabr. Proceedings. Proc. 2001 Solid
d

34 Free. Fabr. Symp. Austin. (2001) 432–440.


35
te

36 [142] A. Birnbaum, P. Aggarangsi, J. Beuth, Process Scaling and Transient Melt Pool
37 Size Control in Laser-Based Additive Manufacturing Processes, Solid Free. Fabr.
38
Proceedings. Proc. 2003 Solid Free. Fabr. Symp. Austin. (2003) 328–339.
p

39
40
41 [143] P. Aggarangsi, J.L. Beuth, D.D. Gill, Transient Changes in Melt Pool Size in Laser
ce

42 Additive Manufacturing Processes, (2000) 163–174.


43
44
45 [144] A. Vasinonta, J.L. Beuth, M. Griffith, Process Maps for Predicting Residual Stress
Ac

46 and Melt Pool Size in the Laser-Based Fabrication of Thin-Walled Structures, J.


47 Manuf. Sci. Eng. 129 (2007) 101.
48
49
50 [145] P. Aggarangsi, J.L. Beuth, M. Griffith, Melt Pool Size and Stress Control for
51 Laser-Based Deposition Near a Free Edge, in: Proc. 14th Solid Free. Fabr. Symp.,
52 Austin, USA, 2003: pp. 196–207.
53
54
55
[146] A.J. Birnbaum, J.L. Beuth, J.W. Sears, Scaling Effects in Laser-Based Additive
56 Manufacturing Processes, Solid Free. Fabr. Proceedings. Proc. 2004 Solid Free.
57 Fabr. Symp. Austin. (2004) 151–162.
58
59
60
61
62
63
64 Page 72 of 103
65
1
2
3
4 [147] D. Rosenthal, The Theory of Moving Sources of Heat and Its Application to Metal
5
6 Treatments, Trans. Am. Soc. Mech. Eng. 68 (1946) 849–866.
7
8 [148] J. Kummailil, Process Models for Laser Engineered Net Shaping, 2004.
9
10
11
[149] J.G. Zhou, D. Herscovici, C.C. Chen, Parametric process optimization to improve
12 the accuracy of rapid prototyped stereolithography parts, Int. J. Mach. Tools

t
13 Manuf. 40 (2000) 363–379.

ip
14
15 [150] C. Lynn-Charney, D.W. Rosen, Usage of accuracy models in stereolithography
16
process planning, Rapid Prototyp. J. 6 (2000) 77–87.

cr
17
18
19 [151] Z. Lu, D. Li, B. Lu, A. Zhang, G. Zhu, G. Pi, The prediction of the building

us
20 precision in the Laser Engineered Net Shaping process using advanced networks,
21
22
Opt. Lasers Eng. 48 (2010) 519–525.
23
24 [152] G. Casalino, A.D. Ludovico, Parameter selection by an artificial neural network

an
25 for a laser bending process, Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 216
26 (2002) 1517–1520.
27
28
M
29 [153] L. Wang, S.D. Felicelli, J.E. Craig, Experimental and Numerical Study of the
30 LENS Rapid Fabrication Process, J. Manuf. Sci. Eng. 131 (2009) 041019.
31
32
33
[154] A. Fathi, A. Mozaffari, Vector Optimization of Laser Solid Freeform Fabrication
d

34 System Using a Hierarchical Mutaable Smart Bee-Fuzzy Inference System and


35 Hybrid NSGA-II/Self-Organizing Map, J. Intell. Manuf. 25 (2014) 775–795.
te

36
37 [155] L. Tang, R.G. Landers, Melt Pool Temperature Control for Laser Metal Deposition
38
Processes—Part I: Online Temperature Control, J. Manuf. Sci. Eng. 132 (2010)
p

39
40 011010.
41
ce

42 [156] Weerasinghe V.M, Laser cladding of flat plates, University of London, 1985.
43
44
45 [157] A. Vasinonta, J. Beuth, M. L. Griffith., Process Maps for Laser Deposition of
Ac

46 Thin-Walled Structures, Solid Free. Fabr. Proceedings. Proc. 1999 Solid Free.
47 Fabr. Symp. Austin. (1999).
48
49
50 [158] T. Purtonen, A. Kalliosaari, A. Salminen, Monitoring and Adaptive Control of
51 Laser Processes, Phys. Procedia. 56 (2014) 1218–1231.
52
53 [159] D. Salehi, M. Brandt, Melt pool temperature control using LabVIEW in Nd:YAG
54
55
laser blown powder cladding process, Int. J. Adv. Manuf. Technol. 29 (2005) 273–
56 278.
57
58
59
60
61
62
63
64 Page 73 of 103
65
1
2
3
4 [160] I.. Smurov, M. Doubenskaia, Temperature Monitoring by Optical Methods in
5
6 Laser Processing, in: D. Majumdar, I. Manna (Eds.), Laser-Assisted Fabr. Mater.,
7 Springer, Berlin, 2013: pp. 373–422.
8
9 [161] a. Raghavan, H.L. Wei, T. a. Palmer, T. DebRoy, Heat transfer and fluid flow in
10
11
additive manufacturing, J. Laser Appl. 25 (2013) 052006. doi:10.2351/1.4817788.
12

t
13 [162] S. Barua, F. Liou, J. Newkirk, T. Sparks, Vision-based defect detection in laser

ip
14 metal deposition process, Rapid Prototyp. J. 20 (2013) 77–85.
15
16
[163] L. Wang, S. Felicelli, Y. Gooroochurn, P.T. Wang, M.F. Horstemeyer,

cr
17
18 Optimization of the LENS Process for Steady Molten Pool Size, Mater. Sci. Eng.
19 A. 474 (2008) 148–156.

us
20
21
22
[164] P. Peyre, P. Aubry, R. Fabbro, R. Neveu, A. Longuet, Analytical and Numerical
23 Modelling of the Direct Metal Deposiion Process, J. Phys. D. Appl. Phys. 41
24 (2008) 025403.

an
25
26 [165] M. Gaumann, C. Bezencon, P. Canalis, W. Kurz, Single-Crystal Laser Deposition
27
28 of Superalloys : Processing – Microstructure Maps, Acta Mater. 49 (2001) 1051–
M
29 1062.
30
31 [166] D. Hu, R. Kovacevic, Sensing, modeling and control for laser-based additive
32
33
manufacturing, Int. J. Mach. Tools Manuf. 43 (2003) 51–60.
d

34
35 [167] L. Tang, J. Ruan, R.G. Landers, F. Liou, Variable Powder Flow Rate Control in
te

36 Laser Metal Deposition Processes, J. Manuf. Sci. Eng. 130 (2008) 041016.
37
38
[168] J. Koch, J. Mazumder, Apparatus and Methods for Monitoring and Controlling
p

39
40 Multi-Layer Laser Cladding, 2000.
41
ce

42 [169] J. Kummailil, C. Sammarco, D. Skinner, Effect of select LENSTM processing


43
44
parameters on the deposition of Ti-6Al-4V, J. Manuf. Process. 7 (2005) 42–50.
45
Ac

46 [170] V. Neela, a. De, Three-dimensional heat transfer analysis of LENSTM process


47 using finite element method, Int. J. Adv. Manuf. Technol. 45 (2009) 935–943.
48
49
50 [171] A. De, T. DebRoy, A Smart Model to Determine Effective Thermal Conductivity
51 and Viscosity in the Weld Pool, J. Appl. Phys. 95 (2004) 5320–5240.
52
53 [172] G. Pi, A. Zhang, G. Zhu, D. Li, B. Lu, Research on the forming process of three-
54
55
dimensional metal parts fabricated by laser direct metal forming, Int. J. Adv.
56 Manuf. Technol. 57 (2011) 841–847.
57
58 [173] F. Xing, W. Liu, T. Wang, Real-time Sensing and Control of Metal Powder Laser
59 Forming, Intell. Control Autom. WCICA. (2006) 6661–6665.
60
61
62
63
64 Page 74 of 103
65
1
2
3
4 [174] E. Fearon, K.G. Watkins, Optimisation of layer height control in direct laser
5
6 deposition, in: 23rd Int. Congr. Appl. Laser Electro-Optics (ICALEO 2004)., 2004.
7
8 [175] A. Fatemi, N. Shamsaei, Multiaxial fatigue: An overview and some approximation
9 models for life estimation, Int. J. Fatigue. 33 (2011) 948–958.
10
11
12

t
13

ip
14
15
16

cr
17
18
19

us
20
21
22
23
24

an
25
26
27
28
M
29
30
31
32
33
d

34
35
te

36
37
38
p

39
40
41
ce

42
43
44
45
Ac

46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64 Page 75 of 103
65
Table

LIST OF TABLES

Table 1 Different powder-based laser deposition additive manufacturing techniques


[2–23].

t
ip
Table 2 Comparison of the tensile properties of materials formed by Laser Powder
Deposition (DLD) and their wrought counterparts at room temperature
[57,63,102–105].

cr
Table 3 Effect of building orientation on the tensile properties of the direct laser-

us
deposited materials [42,44,57].

an
M
d
p te
ce
Ac

Page 76 of 103
Table 1 Different powder-based laser deposition additive manufacturing techniques [2–23].

Process Acronym References


Laser Cladding LC [2,3]

t
Laser Direct Casting LDC [4,5]

ip
Direct Metal Deposition DLD [6]
Directed Light Fabrication DLF [7–9]
Laser Forming Lasform [10]

cr
Shape Deposition Manufacturing SDM [11]
Laser Engineering Net Shaping LENS [12,13]
Laser Powder Fusion LPF [4]

us
Freeform Laser Consolidation LC [14,15]
Laser-aided Direct-Metal/Material Deposition DMD [16–18]
Laser-based Multi-Directional Metal Deposition- LBMDMD [19,20]
Laser-aided Manufacturing LAMP [21–23]

an
M
d
p te
ce
Ac

Page 77 of 103
t
ip
cr
Table 2 Comparison of the tensile properties of materials formed by Laser Powder Deposition (DLD) and their wrought

us
counterparts at room temperature [57,63,102–105].
Elongation to failure
Ultimate Stress (MPa) Yield Stress (MPa)
Alloys (%) AM Process

an
Wrought DLD Wrought DLD Wrought DLD
1 1
316 SS 586 758 2341 4341 50 1
461 LENS
*2 3
316L SS 480 540-560 170*2 330-3453 40 *2
35-433 Laser consolidation
… 6551 2761 3241 551 701

M
404L SS LENS
1 1
AISI H-13 1,725 1,703 1,4481 1,4621 12 1
1-31 LENS
CPM-9V … 1,315 3
… 8213 … >23 Laser consolidation
Ti-6Al-4V 931‡1 896-1,000‡1 855‡1 827-965‡1 10‡1 1-16‡1 LENS

ed
TC-18 1,157‡5 1,147-1,188‡5,6 1,119‡5 1,095‡5,6 14 ‡5
4.5-5.75‡5,6 Laser melting deposition
IN-7183 1,379†1 1,400†1 1,158†1 1,117†1 20†1 16†1 LENS
1 1
IN-625 834 931 4001 6141 37 1
381 LENS
IN-600 660‡1 7311 285‡1 4271 45‡1 401 LENS
pt
4 3
IN-690 725 665 3484 4503 41 4
493 DLF
IN-738 1,0954 1,2003 950 8703 6.54 183 Laser consolidation
*
Hot finished-annealed.
ce


Annealed.

Solution treated and annealed.
1
[57]
Ac

2
[102]
3
[63]
4
[103]
5
[104]
6
[105]

Page 78 of 103
t
ip
cr
Table 3 Effect of building orientation on the tensile properties of the direct laser-deposited materials [42,44,57].

us
Ultimate Stress (MPa) Yield Stress (MPa) Elongation to failure (%)
Alloys Perpendicular Parallel Perpendicular Parallel Perpendicular Parallel References

an
(Y-direction) (X-direction) (Y-direction) (X-direction) (Y-direction) (X-direction)
316 SS 793 807 448 593 66 30 [57]
304L SS 655 710 324 448 70 59 [57]
IN-625 931 938 614 517 38 37 [57]

M
IN-690 607 745 386 434 45 48 [57]
IN-718 650 1,000 612 640 4.6 38 [42]
Ti-6Al-4V 1,065 1,073 941 973 11.4 10 [44]

ed
pt
ce
Ac

Page 79 of 103
Figure

LIST OF FIGURES
Fig. 1 Microstructure of a direct laser-deposited Ti-6Al-4V part in Y-Z plane at
(a) the top region and (b) the bottom region [80].

Fig. 2 (a) Macrostructure with fine columnar grains and (b) microstructure
containing very fine Widmanstätten structure for LENS Ti-6Al-4V.

t
ip
Fig. 3 Powder injection point, A, (a) ahead of, (b) in-line with, and (c) behind the
laser spot center, O [91].

cr
Fig. 4 Different deposition patterns: (a) raster, (b) bi-directional, (c) offset-out,
(d) offset-in, and (e) fractal patterns.

us
Fig. 5 Tool paths using paralleled and 3D layer approaches [100].
Fig. 6 Schematic view of a laser-deposited Ti-6Al-4V specimens, (a) as-deposited
surface in the x-direction (horizontally-oriented), (b) as-deposited surface

an
in the Y-direction (laterally-oriented), and (c) as-deposited surface in the Z-
direction (vertically-oriented).
Fig. 7 Tensile properties of LENS Ti-6Al-4V in different building orientations
M
after heat treatment and HIP process [24].

Fig. 8 Schematic of three stages of fatigue failure: crack initiation (incubation),


d

crack propagation, and final fracture.


te

Fig. 9 Comparison of fatigue strength of direct laser-deposited TC18 specimens


[104] to wrought TC18 with Widmanstätten (lamellar) and bi-modal
p

microstructures [33] at 107 cycles.


Fig. 10 (a) Strain life schematic of crack nucleation, growth, and final fracture, (b)
ce

Schematic of fatigue life for fine and coarse microstructure [112].


Fig. 11 The S-N data for the Inconel 718 specimens fabricated with the LENS
Ac

technique [44].
Fig. 12 Comparison of S-N data of different study for the LENS Ti-6Al-4V
specimens in different directions [24,28,44] to the cast, cast plus HIP, and
wrought materials [123].
Fig. 13 Ti-V phase diagram at 6% Al.
Fig. 14 Comparison of fatigue strengths for LENS Ti-6Al-4V in different building
orientations for both heat treated and HIP’ed specimens [24].

Fig. 15 Comparison of KIC values for LENS Ti-6Al-4V [24] to the conventionally
fabricated Ti-6Al-4V [123].

Page 80 of 103
Fig. 16 Comparison of fatigue crack growth rate results, da/dN, versus intensity
factor, ∆K, for LENS (a) Inconel 625 [46] and (b) AISI 316 L SS [41]
specimens to wrought materials [124].
Fig. 17 Fatigue life predictions using the multistage fatigue model versus the

t
strain–life data for the LENS AISI 316L SS specimens [38] and ASME

ip
design curve [133].
Fig. 18 Normalized melt pool length (L) vs. normalized distance from the location

cr
of a step change in laser power (X) for a thin-walled structure for various,
normalized melt pool temperatures, Tm [143].

us
Fig. 19 Proportional-integral-derivative (PID) control of DLD melt pool
temperature based on thermal measurement [159].
Fig. 20 Temperature contours of melt pools during DLD at traverse speed of 500

an
cm/s (a)-(b): similar in shape in XY (build) plane and (c)-(d):
corresponding XZ (profile) plane views demonstrating different melt pool
depths [161].
M
Fig. 21 Nominal laser power vs. layer/pass number for uniform melt pool geometry
on stainless steel thin wall build for various traverse velocities [163].
Fig. 22 Schematic relationships among process parameters, thermal history,
d

microstructure, and fatigue behavior in Direct Laser Deposition process.


p te
ce
Ac

Page 81 of 103
t
ip
(a) (b)

cr
us
an
M
Fig. 1. Microstructure of a direct laser-deposited Ti-6Al-4V part in Y-Z plane at (a) the
top region and (b) the bottom region [80].
d
p te
ce
Ac

Page 82 of 103
t
ip
cr
1000µm

us
(a) (b)

Fig. 2. (a) Macrostructure with fine columnar grains and (b) microstructure containing

an
very fine Widmanstätten structure for LENS Ti-6Al-4V.
M
d
p te
ce
Ac

Page 83 of 103
t
ip
cr
us
Fig. 3. Powder injection point, A, (a) ahead of, (b) in-line with, and (c) behind the laser

an
spot center, O [91].
M
d
p te
ce
Ac

Page 84 of 103
t
ip
cr
us
an
M
d
te

Fig. 4. Different deposition patterns: (a) raster, (b) bi-directional, (c) offset-out, (d)
offset-in, and (e) fractal patterns.
p
ce
Ac

Page 85 of 103
t
ip
cr
Fig. 5. Tool paths using paralleled and 3D layer approaches [100].

us
an
M
d
p te
ce
Ac

Page 86 of 103
t
ip
cr
us
an
M
d
p te
ce
Ac

Fig. 6. Schematic view of a laser-deposited Ti-6Al-4V specimens, (a) as-deposited


surface in the x-direction (horizontally-oriented), (b) as-deposited surface in the Y-
direction (laterally-oriented), and (c) as-deposited surface in the Z-direction (vertically-
oriented).

Page 87 of 103
t
ip
cr
us
an
Fig. 7. Tensile properties of LENS Ti-6Al-4V in different building orientations after heat
M
treatment and HIP process [24].
d
p te
ce
Ac

Page 88 of 103
t
ip
cr
us
an
M
d
p te

Fig. 8. Schematic of three stages of fatigue failure: crack initiation (incubation), crack
ce

propagation, and final fracture.


Ac

Page 89 of 103
t
ip
cr
us
an
Fig. 9. Comparison of fatigue strength of direct laser-deposited TC18 specimens [104] to
wrought TC18 with Widmanstätten (lamellar) and bi-modal microstructures [33] at 107
M
cycles.
d
p te
ce
Ac

Page 90 of 103
t
ip
cr
us
an
M
d
p te
ce
Ac

Fig. 10. (a) Strain life schematic of crack nucleation, growth, and final fracture, (b)
Schematic of fatigue life for fine and coarse microstructure [112].

Page 91 of 103
900
LENS Inconel 718 data [44]
800 Curve Fitted to LENS Inconel 718
Stress Amplitude (MPa)

700

t
ip
600

cr
500

us
400

300

an
1E+03 1E+04 1E+05 1E+06 1E+07 1E+08 1E+09
Cycles to Failure (Nf)
Fig. 11. The S-N data for the Inconel 718 specimens fabricated with the LENS technique [44].
M
d
p te
ce
Ac

Page 92 of 103
1,600
"Kobryn & Semiatin [24] in X-direction (Horizontal)
1,400 Amsterdam & Kool [44] in X-direction (Horizontal)

t
"Kobryn & Semiatin [24] in Y-direction (Lateral)

ip
Maximum Stress (MPa)

1,200 Kobryn & Semiatin [24] in Z-direction (Vertical)


Cast+HIP [123] Grylls [28] in Z-direction (Vertical)

cr
1,000 Cast [123] Wrought [123]

800

us
600

an
400

200
M
0
1E+02 1E+03 1E+04 1E+05 1E+06 1E+07 1E+08 1E+09
Cycles to Failure (Nf)
d
te

Fig. 12. Comparison of S-N data of different study for the LENS Ti-6Al-4V specimens in
p

different directions [24,28,44] to the cast, cast plus HIP, and wrought materials [123].
ce
Ac

Page 93 of 103
t
ip
cr
us
an
M
Fig. 13. Ti-V phase diagram at 6% Al.
d
te
p
ce
Ac

Page 94 of 103
1,600
X-direction (Horizontal) Heat treated
1,400 Y-direction (Lateral) Heat treated
Z-direction (Vertical) Heat treated
X-direction (Horizontal) HIP'ed
Maximum Stress (MPa)

1,200 Wrought
Cast+HIP Y-direction (Lateral) HIP'ed
Cast Z-direction (Vertical) HIP'ed
1,000

t
ip
800

cr
600

400

us
200

an
0
1E+02 1E+03 1E+04 1E+05 1E+06 1E+07 1E+08
Cycles to Failure (Nf)
M
Fig. 14. Comparison of fatigue strengths for LENS Ti-6Al-4V in different building
orientations for both heat treated and HIP’ed specimens [24].
d
p te
ce
Ac

Page 95 of 103
t
ip
cr
us
an
M
d

Fig. 15. Comparison of KIC values for LENS Ti-6Al-4V [24] to the conventionally
te

fabricated Ti-6Al-4V [123].


p
ce
Ac

Page 96 of 103
t
ip
cr
us
(a)
an
M
d
p te
ce
Ac

(b)

Fig. 16. Comparison of fatigue crack growth rate results, da/dN, versus intensity factor,
∆K, for LENS (a) Inconel 625 [46] and (b) AISI 316 L SS [41] specimens to wrought
materials [124].

Page 97 of 103
0.8
LENS-produced AISI 316L SS Data [38]
0.7 MSF Prediction Curve [38]
MSF-Lower Bound [38]

t
Strain Amplitude (%)

0.6 MSF-Upper Bound [38]

ip
ASME Design Curve [133]
0.5

cr
0.4

us
0.3

0.2

0.1
1E+02 1E+03 1E+04 an 1E+05 1E+06 1E+07
M
Cycles to Failure (Nf)

Fig. 17. Fatigue life predictions using the multistage fatigue model versus the strain–life
d

data for the LENS AISI 316L SS specimens [38] and ASME design curve [133].
p te
ce
Ac

Page 98 of 103
3

2.5
Tm=0.86
2
L=l/(2k/ρcV)

t
ip
1.5
Tm=1.18

cr
1
Tm=1.44

us
0.5
Tm=2.59

an
0
0 1 2 3 4 5 6
X=x/(2k/ρcV)
M
Fig. 18. Normalized melt pool length (L) vs. normalized distance from the location of a
step change in laser power (X) for a thin-walled structure for various, normalized melt
pool temperatures, Tm [143].
d
p te
ce
Ac

Page 99 of 103
t
ip
cr
us
Fig. 19. Proportional-integral-derivative (PID) control of DLD melt pool temperature
based on thermal measurement [159].

an
M
d
p te
ce
Ac

Page 100 of 103


t
ip
cr
us
an
M
d

Fig. 20. Temperature contours of melt pools during DLD at traverse speed of 500 cm/s
(a)-(b): similar in shape in XY (build) plane and (c)-(d): corresponding XZ (profile) plane
te

views demonstrating different melt pool depths [161].


p
ce
Ac

Page 101 of 103


t
800

ip
Velocity=2.5 mm/s
Velocity=7.62 mm/s

cr
700
Velocity=20 mm/s
Nominal Laser Power (W)

us
600

500

400
an
M
300

200
d

1 2 3 4 5 6 7 8 9 10
te

Layer Number
p

Fig. 21. Nominal laser power vs. layer/pass number for uniform melt pool geometry on
stainless steel thin wall build for various traverse velocities [163].
ce
Ac

Page 102 of 103


t
ip
cr
us
an
M
d
te

Fig. 22. Schematic relationships among process parameters, thermal history,


microstructure, and fatigue behavior in Direct Laser Deposition process.
p
ce
Ac

Page 103 of 103

S-ar putea să vă placă și