Sunteți pe pagina 1din 107

Imperial College London Department of Mechanical Engineering

ADVANCED FRACTURE MECHANICS


Lectures on Fundamentals of Elastic, Elastic-Plastic and Creep Fracture 20022003

Course lecturer: Dr Noel ODowd

CONTENTS Page Course Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii 1. Linear Elastic Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Denition of energy release rate, G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Strain energy, energy release rate and compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Stress analysis of cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.4 Mixed mode fracture mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 1.5 Concept of small scale yielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2. Non-linear Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.1 The J integral, (Rice, 1968) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.2 Power law hardening materialsThe HRR eld . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.3 Crack tip opening displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.4 Relationship between J and G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.5 Evaluating J for test specimens and components . . . . . . . . . . . . . . . . . . . . . . . . . . 45 2.6 Application of non-linear fracture mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 2.7 K dominance, J dominance and size requirements . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.8 Standard test to determine JIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 3. Micromechanisms for ductile and brittle fracture . . . . . . . . . . . . . . . . . . . . . . 66 3.1 Micromechanism of cleavage failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 3.2 Prediction of fracture toughness using the RKR model and the HRR eld . . 67 3.3 Micromechanism of ductile failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 3.4 Prediction of fracture toughness using the MVC model and the HRR eld . 69 3.5 Competition between brittle and ductile fracture . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4. Application of BS 7910 in failure assessments . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.1 The failure assessment diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.2 Level 1 Failure Assessment Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.3 Level 2 Failure Assessment Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.4 Level 3 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 5. Creep Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.1 Secondary creep. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.2 Estimation of C in specimens and components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 5.3 Creep solutions for short times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5.4 Characterisation of creep crack initiation and growth . . . . . . . . . . . . . . . . . . . . . . 89 5.5 Elastic-plastic creep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.6 Micrographs of creep failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 6. Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 6.1 Appendix A, Extracts from two key papers on non-linear fracture mechanics 6.2 Appendix B, List of important equations for Advanced Fracture Mechanics 6.3 Appendix C, Linear Elastic K eld distributions

September 2002

Imperial College London Department of Mechanical Engineering 4M/AME Advanced Fracture Mechanics The course, which will be given by Dr ODowd, consists of approximately 22 lectures and 9 tutorials. Examination will be by written paper at the end of the course and by problem sheets (5 in total) which will be distributed throughout the year and are worth 20% of the total course mark. 4M students must have taken the course 3M course, Fundamentals of Fracture Mechanics. Aims The principal aim of the course is to provide students with a comprehensive understanding of the stress analysis and fracture mechanics concepts required for describing failure in engineering components. In addition, the course will explain how to apply these concepts in a safety assessment analysis. The course deals with fracture under brittle, ductile and creep conditions. Lectures are presented on the underlying principles and exercises provided to give experience of solving practical problems. Objectives 1. 2. 4. 5. 6. At the end of the course the student should be able to: Understand the mechanisms of fracture under brittle, ductile and creep conditions. Explain the relationship between linear elastic and non-linear fracture concepts and the terms K, G, J and C . Establish the theoretical stress distributions ahead of a crack under brittle, ductile and creep conditions. Appreciate how to make valid fracture toughness measurements on materials. Understand the theoretical basis behind fracture mechanics design codes and know how to apply these codes to cracks in engineering components.

Relevant Textbooks Most mechanics of materials textbooks provide an introduction to fracture mechanics, e.g. Mechanical Behaviour of Materials, by N.E. Dowling. The more advanced topics covered in these lectures are dealt with by the following textbooks, which should be considered background reading and are not required texts. The texts are listed in alphabetical order. 1. T.L. Anderson Fracture mechanics: fundamentals and applications, Edwards Arnold, London, 1991. 2. R.W. Hertzberg, Deformation and fracture mechanics of engineering materials, Wiley and Sons, New York, 1989. 3. M.F. Kanninen and C.H. Popelar, Advanced fracture mechanics, Oxford University Press, 1985. 4. G.A. Webster and R.A. Ainsworth, High temperature component life assessment, Chapman and Hall, London, 1994.

ii

September 2002

Imperial College of Science Technology and Medicine, Department of Mechanical Engineering 4M/AME Advanced Fracture Mechanics Introduction Fracture mechanics concerns the design and analysis of structures which contain cracks or aws. On some size-scale all materials contain aws either microscopic, due to cracked inclusions, debonded bres etc., or macroscopic, due to corrosion, fatigue, welding aws etc. Thus fracture mechanics is involved in any detailed design or safety assessment of a structure. As cracks can grow during service due to e.g. fatigue, fracture mechanics assessments are required throughout the life of a structure or component, not just at start of life. Fracture mechanics answers the questions: What is the largest sized crack that a structure can contain or the largest load the structure can bear for failure to be avoided? How long before a crack which was safe becomes unsafe? What material should be used in a certain application to ensure safety? Studies in the US in the 1970s by the US National Bureau of Standards estimated that cost of fracture due to accidents, overdesign of structures, inspection costs, repair and replacement was on the order of 120 billion dollars a year. While fracture cannot of course be avoided, they estimated that, if best fracture control technology at the time was applied, 35 billion dollars could be saved annually. This indicates the importance of fracture mechanics to modern industrialised society. The topics of linear elastic fracture mechanics, elastic-plastic fracture mechanics and high temperature fracture mechanics (creep crack growth) are dealt with in this course. The energy release rate method of characterising fracture is introduced and the K and HRR elds which characterise the crack tip elds under elastic and plastic/creep fracture respectively are derived. The principal mechanisms of fracture which control failure in the dierent regimes are also discussed. In the later part of the course, the application of these fracture mechanics principles in the assessment of the safety of components or structures with aws through the use of standardised procedures is discussed. The approach taken in this course is somewhat dierent from that in Fundamentals of Fracture Mechanics (FFM) as here more emphasis is put on the mechanics involved and outlines of mathematical proofs of some of the fundamental fracture mechanics relationships are provided. There is some revision of the topics covered in FFM, particularly in the area of linear elastic fracture mechanics though the approach is a little dierent. iii

1. Linear elastic fracture mechanics 1.1 Denition of energy release rate, G Grith (1924) derived a criterion for crack growth using an energy approach. It is based on the concept that energy must be conserved in all processes. He proposed that when a crack grows the change (decrease) in the potential energy stored in the system, U , is balanced by the change (increase) in surface energy, S, due to the creation of new crack faces.

S = s (A)

a
Figure 1.1, Schematic of a crack growing by an amount a. Consider a through-thickness crack in a body of thickness B (see Fig. 1.1). For fracture to occur energy must be conserved so, U + S = 0. The change in surface energy, S = As where A is the new surface area created and s is the surface energy per unit area, as illustrated above. The change in area A = 2Ba, (the factor of 2 arises because there are two crack faces). Inserting these values and dividing across by Ba we get 1 U = 2s . B a

Rewriting as a partial derivitive we get Griths relation, 1 U = 2s B a 1

If this equation is satised then crack growth will occur. The energy release rate, G is dened as 1 U . B a In almost all situations U/a is negative, i.e. when the crack grows the potential G= energy decreases, so G is positive. Note that the 1/B term is often left out and U is then the potential energy per unit thickness. G has units J/m2 , N/m or MPam and is the amount of energy released per unit crack growth per unit thickness. It is a measure of the energy provided by the system to grow the crack and depends on the material, the geometry and the loading of the system. The surface energy, s , depends only on the material and environment, e.g. temperature, pressure etc., and not on loading or crack geometry. From the above, a crack will extend when G crack driving force 2s = Gc material toughness .

It was found that while Griths theory worked well for very brittle materials such as glass it could not be used for more ductile materials such as metals or polymers. The amount of energy required for crack was found to be much greater than 2s for most engineering materials. The result was therefore only of academic interest and not much attention was paid to this work outside the academic community. In 1948 Irwin and Orowan independently proposed an extension to the Grith theory, whereby the total energy required for crack growth is made up of surface energy and irreversible plastic work close to the crack tip: = s + p , where p is the plastic work dissipated in the material per unit crack surface area created (in general p >> p ). Then the criterion for fracture becomes or G= 1 U 2(s + p ) B a

1 U 2(s + p ) = Gc . B a

The Grith and Irwin/Orowan approaches are mathematically equivalent, the only dierence is in the interpretation of the material toughness Gc . In general Gc is obtained 2

directly from fracture tests which will be discussed later and not from values of s and p . The critical energy release rate, Gc , can be considered to be a material property like Youngs Modulus or yield stress. It does not depend on the nature of loading of the crack or the crack shape, but will depend on things like temperature, environment etc. We next examine how to determine the potential energy and the energy release rate for a linear elastic material. 1.2 Strain energy, energy release rate and compliance The energy release rate G can be written in terms of the elastic (or elastic-plastic) compliance of a body. Before showing this, some general denitions will be given. 1.2.1 Strain energy density Strain energy density, W , is given by, W =
0

d,

where is the stress tensor (matrix), is the strain tensor (matrix). Under uniaxial loading W is simply the area under the stress-strain curve (note: not the loaddisplacement curve)as illustrated in Fig. 1.2. In general, the strain energy will not be constant throughout the body but will depend on position.

Figure 1.2, Denition of strain energy density W under uniaxial loading. 1.2.2 Elastic and plastic materials For an elastic material all energy is recovered on unloading. For a plastic material, energy is dissipated. 3

The strain energy density of an elastic material depends only on the current strain, while for a plastic material W depends on loading/unloading history. If the material is under continuous loading, W for an elastic and plastic material are the same. However, if there is total or partial unloading there is a dierence. The response of an elastic and an elastic-plastic material such as steel is shown in Fig. 1.3. The elastic material unloads back along the loading path, i.e. no work is done in a cycle which in fact is the denition of an elastic material. For an elastic-plastic material there is generally an initial elastic regime where the stress-strain curve is linear (stress directly proportional to strain) and energy is recoverable and a nonlinear plastic regime where energy is dissipated (unrecoverable). Unloading is usually taken to follow the slope of the initial elastic region. There is then a permanent plastic deformation and the work done in a cycle is given by the area under the curve.

loading unloading

loading Work done

unloading Elastic Material Work done = 0 Elastic-plastic material

Figure 1.3, Comparison between behaviour of an elastic material (left) and an elasticplastic material (right). An elastic material need not be a linear elastic materialthere are elastic materials, e.g. rubbers, which are non-linear. However, the term elastic is often used as a shorthand for linear elastic. For a linear elastic material, = D, where D is the elasticity matrix and and are the stress and strain matrices. W = In uniaxial loading W = 4 1 . 2 2 . 2E

Power law hardening, a non-linear stress-strain law where strain is proportional to stress raised to a power, is often used to represent the plastic behaviour of materials = D1/n , where again D is a matrix of material constants and n is the strain hardening exponent, 1 n . In this case, it can be shown that, W = 1.2.3 Denition of Strain energy, Ue The strain energy of the body is a measure of how much strain energy is stored in the body, depends on the material and loading and is given by, Ue =
V

n . n+1

W dV ,

where V is the volume of the body. The strain energy, Ue , is the strain energy density integrated over the whole body and it can be shown that it is equal to the area under the load-displacement curve. For a linear elastic material the strain energy is simply, Ue = P , 2

where P and are the applied load and conjugate displacement. For a power law hardening material it can be shown that, Ue = 1.2.4 Denition of Potential energy, U : The potential energy is made up of the internal strain energy and the external work done on the body and depends on the way the body is loaded.

n P . n+1

Figure 1.4, Schematic of a loaded cracked body. 5

For the body shown in Fig. 1.4, the potential energy will have a dierent denition depending on whether it is loaded by a prescribed load or a prescribed displacement: For prescribed displacement, : For prescribed load, P : U = Ue U = Ue P

Prescribed load means that the load will be xed (constant) during an increment of crack growth, while prescribed displacement means that displacement is xed during crack growth. For prescribed displacement, no work can be done by the external loading during crack growth because the displacement remains xed (work = force displacement) so the change in potential energy is only due to the change in strain energy. 1.2.5 Denition of Compliance, C: The compliance of a body is the inverse of the stiness. It is not a material property, but depends on the loading and geometry. For a linear elastic body with a crack of length a we can write = C(a)P, where C(a) is the compliance and depends on geometry (including crack length, a), Youngs modulus, E and . For a power law hardening material we can write = C(a, n)P n , where the compliance C(a, n) also depends on the hardening exponent. 1.2.6 Derivation of G from compliance for linear elastic material: For xed load: U= 1 B P P P = 2 2 U a =
P

G= where the notation a

P 2B

emphasises that load P is held constant.


P

= C(a)P a =P
P

dC da

Therefore G= If displacement is held xed:

1 2 dC P . 2B da P 2

U = Ue =

and it can be shown (this is left as an exercise) that again G= 1 2 dC P . 2B da

In other words although the potential energy depends on mode of loading, the energy release rate, G, does not and is independent of the nature of loading whether by imposed displacement or imposed load. (This is true in general not only for linear elastic materials.) 1.2.7 Stability of Crack Growth If we examine the above equation for G and dierentiate again with respect to a, we get for xed load: G a =
P

1 2 d2 C P 2B da2

For xed displacement we get G a 1 2 = P 2B d2 C 2 2 da C dC da


2

For most geometries, d2 C/da2 is positive, so for prescribed load G increases with crack growth. Therefore crack growth is unstablean increase in a leads to an increase in G. For prescribed displacement, if 2/C(dC/da)2 > d2 C/da2 then G <0 a so G decreases with crack growth and crack growth is stable, i.e. the applied displacement must be increased to maintain crack growth. Criteria for unstable crack growth: G = Gc and G > 0. a

Stability arguments are important because while an amount of stable crack growth may be acceptable, unstable fracture must always be avoided. 7

Most materials are ductile and even under nominally linear elastic conditions, fracture toughness increases with crack growth as shown. This is what is known as a resistance curve or sometimes an R curve (see Fig. 1.5).

R curve Gr No R curve behaviour Gc

a
Figure 1.5, Schematic of material resistance curve behaviour In these circumstances, satisfying the above criteria will not necessarily lead to unstable crack growth and the criteria for unstable crack growth are G = Gr 1.3 Stress analysis of cracks 1.3.1 The K-eld for linear elastic materials, Irwin (1958) An alternative approach to the energy approach in the analysis of cracks is the stress intensity approach where the stress and strain eld at the crack tip are examined. In many situations the energy and stress intensity approaches are equivalent and give the same predictions. However, it is important to be familiar with both approaches.The energy approach is appropriate mainly for elastic (linear or non-linear elastic) materials. The stress intensity approach is perhaps more exible and can be applied to a wider range of materials. We wish to determine the stress state at the crack tip. The most general 3-D stress state is, xx yx (x, y, z) = zx 8 xy yy zy xz yz , zz and G dGr > . a da

i.e. six independent stress components and all components depend on x, y and z. To simplify we rst assume an innitely sharp, straight crack front and align the axes along the crack front (see Fig. 1.6).

z x

Figure 1.6 Schematic of a three dimensional crack Assume that as the crack tip is approached the stress variation along the crack front (in the z direction) is negligible compared to the variation in the x and y direction, xx yy zz = = . . . = 0. z z z Therefore close to the crack tip the stresses and strains do not depend on zthe stresses behave as if responding to a 2-D deformation eld. The stress state at a crack tip is therefore a combination of plane stress/strain, u(x, y), v(x, y), and an anti-plane strain problem, w(x, y), where u and v are the x and y (in plane) displacements, and w is the out of plane (z) displacement. For a linear elastic problem, we may examine these modes separately and determine the total stress by (linear) superposition. 1.3.2 The plane stress/strain crack tip elds For simplicity we use polar coordinates as shown in Fig. 1.7.

Figure 1.7 Polar coordinates centered at a sharp crack tip We focus attention on the crack tip so we do not consider the remote boundary condi9

tions. The only relevant boundary conditions are the stress free crack faces: (r, ) = 0 r (r, ) = 0. For a linear elastic material, this problem may be solved by using the Airy stress function method (covered in the Advanced Stress Analysis course) which gives an exact solution to linear elastic stress problems. The details of the proof are beyond the scope of the course, but an outline of the steps in the proof is given below. The stress eld is represented as an innite series in r, i.e.

(r, ) =
i=1

Ai ri fi ()

where Ai and i are unknown real constants and fi are unknown functions of . If sucient terms are taken the exact solution to any problem can be obtained. However, for a fracture mechanics analysis, we focus on the terms which make the largest contribution to the stress in the vicinity of the crack tip, r 0. Values of 0 can be ignored, as the stress term corresponding to this value of tends to zero as we approach the crack tip (i.e. r 0 when r 0 for > 0; r = 1 when = 0). It can also be shown that in order to have bounded strain energy at the crack tip terms of order lower than 1 must also be excluded. (Physical requirements rule out innite strain energy, though innite stress is allowed). It can then be shown that the only value of , between 1 and 0, which satises equilibrium and the above boundary conditions is = 1/2, i.e. A f () + . . . r The dots indicate that this is the rst (and most important) term in a series for the crack tip stress eld. Conventionally, the arbitrary constant A is replaced by the stress intensity factor K and the solution is divided into Mode I (tension) and Mode II (shear) solutions. The Mode I stresses at the crack tip can be derived (the proof is relatively straightforward) and are as follows: KI = cos3 + . . . 2 2r KI rr = cos + cos sin2 2 2 2 2r KI sin cos2 + . . . r = 2 2 2r 10

+ ...

Note that at = 0,

KI = rr = ; 2r

r = 0;

and at = (the crack faces), (r, ) = 0 r (r, ) = 0. For Mode II the result is: KII = 2r KII rr = 2r KII r = 2r and at = 0 = rr = 0; and again at = , (r, ) = 0 r (r, ) = 0. The value of the out of plane stress, zz , will depend on whether we assume plane stress or plane strain, zz = (rr + ) plane strain 0 plane stress cos2 + ... 2 2 + ... sin 3 sin3 2 2 cos 3 cos sin2 + ... 2 2 2 3 sin KII r = ; 2r

In a real 3-D geometry, zz will vary from plane strain at the centre to plane stress at the surface. However, the other stress components remains the same. For Mode III antiplane shear mode we get KIII cos + . . . z = 2 2r KIII zr = sin + . . . 2 2r and all other stress components are zero. These stress elds fully describe the stress elds for a sharp crack in a linear elastic material. The values of KI , KII and KIII are undetermined by the above analysis 11

and will be found by considering the remote boundary conditions. The term singular (where singular means innite) elds is often used to describe these elds as they predict innite stress stress at r = 0. The singularity is referred to as a square root singularity, because, 1/ r.

The three modes are shown in Fig. 1.8 and the full mathematical description of the three modes is provided in Appendix C.

Mode I

Mode II Y Y X Z X Z

Mode III

Y X Z

Mode I: Mode II: Mode III:

Opening or Tensile Mode Sliding or In-Plane Shear Mode Tearing or Anti-Plane Shear Mode

Figure 1.8 The three modes of crack tip deformation Note: The K eld is not the exact solution to the stress eld in a cracked body. It is the solution to the stress eld as we approach the crack tip, where the approximations used in the derivation of the K eld apply. Because this term in the stress eld is so much larger than the other terms, we are justied in neglecting them and a single parameter K then denes the stress at the crack tip (more on this point later.) 1.3.4 Crack tip opening displacement The crack tip opening displacement, u in Fig. 1.9, can be obtained directly from the K eld. 12

Figure 1.9 Denition of crack tip opening displacement, u Mathematically, the crack tip opening displacement is dened as, u(r) = uy (r, = ) uy (r, = ). In the above equation uy is the y displacement and the equation for this is given in Appendix C. Inserting the equation for uy from Appendix C gives for Mode I, u(r) = 2 where = KI E r (1 + )( + 1), 2 plane strain plane stress.

3 4 3 1+

Substituting for the above can be rewritten as u(r) = 8 where E = E 1 2 E KI E r 2

plane strain plane stress.

The crack tip opening displacement, as dened here, is not a single number but gives the relative displacement of the crack faces at a distance r from the crack tip, as dened by the K eld. Note that , 1 ; r u, CTOD r,

where r measures distance from the crack tip. The use of the CTOD (crack tip opening displacement) as a fracture parameter is based on non-linear elastic fracture mechanics and will be discussed later. 13

1.3.5 K in innite and nite bodies For some idealised geometries an exact value for KI , KII , KIII can be determined. Consider an innite plate, with a straight through thickness crack of length 2a, loaded
at innity by tension, yy , and shear, xy and yz as shown in Fig. 1.10.

yy

yz xy

Figure 1.10 An innite plate with a centre crack of length 2a. It can be shown that, KII KIII KI = yy a = xy a = yz a

This is an exact solution for K and can be determined analytically. Note that a xx stress applied at innity has no eect on K since it is parallel to the crack. Note
also that the modes are uncoupled, i.e. xy does not contribute to KI or yy to KII

etc. For a homogenous material, the crack modes are always uncoupledit is somewhat more complicated for an interface crack (a crack lying at the interface of two dierent materials). For an edge crack of length a in an innite plate we get KI = 1.12 yy a For an innite plate, (a << W ) with an edge crack under pure moment M (Fig. 1.11) we get,
W

Figure 1.11 Innite plate with an edge crack under bending 14

KI = 1.12 b a,
where b is the bending stress, b =

My W 12 6M =M = 2 , 3b I 2 W W b

with W the width of the plate and b the thickness. For a nite sized plate of width 2W with a center crack under tension and shear, KI = YI (a/W )y a KII = YII (a/W )xy a KIII = YIII (a/W )yz a where YI , YII , YIII are the shape factors (and clearly as a/W 0, Y ).

The values of YI , YII , YIII must generally be obtained numerically, e.g. from a nite element solution and have been tabulated in handbooks for many geometries, e.g. Sih, G. C. Handbook of Stress-Intensity factors, 1974. 1.3.5 Comparison of K eld with full stress eld As pointed out in Section 1.3.1 the K eld is not the full solution to the stress eld in a cracked body. It is the solution to the stress eld as we approach the crack tip, where the assumptions used in the derivation of the K eld apply. Consider the innite plate, with a crack of length 2a (Fig. 1.12). The exact 2-D solution to the stress eld can be obtained using stress functions.

2a

Figure 1.12 Innite plate with a centre crack under remote tension 15

With only tension applied remotely, the yy stress ahead of the crack along the centerline, y = 0, is given by
yy |x| yy = x2 a2

for |x a| > 0,

()

where x measures distance from the centre of the plate. Replacing x by r where r measures distance from the crack tip, i.e. r = x a x = r + a, gives yy =

yy |r + a| = = . r2 + 2ar + a2 a2 r2 + 2ar (r + a)2 a2

yy |r + a|

yy |r + a|

Focusing on the crack tip, by letting r 0; then r + a a and r2 + 2ar 2ar and we get yy yy a yy a = . 2ar 2r

If we compare this result with the K eld for a Mode I crack derived earlier, KI yy = . 2r Equating the two expressions for yy we get, KI = geometry. The comparison between the exact stress eld from equation * above and the K eld is shown on a log-log plot in Fig. 1.13(a). It is seen that for r < 0.1a the K eld and the exact stress distribution are in excellent agreement. For most other geometries the stress elds must be calculated numerically. The stress eld obtained for a center cracked panel with a/W = 0.5 loaded under tension, calculated from an FE analysis, compared with the K eld for this geometry is also shown in Fig. 1.13(b) (the Y value for the geometry is 1.15 .) If the K eld agrees with the actual stress eld over a reasonable distance the specimen is said to be K dominant. Most standard test specimens are K dominant. However, the commonly used DCB (double cantilever beam) has a small zone of K dominance and results from this specimen are generally interpreted using the energy release rate rather than the stress intensity factor. Plasticity can further aect the zone of K dominance and we will have more on that later. 16 yy a the KI value for this

Figure 1.13 Comparison between exact solution and K eld, (a) innite plate with a centre crack, (b) centre cracked plate with a/W = 0.5. Note that the analysis outlined in this section is one way of obtaining the value of K. If the full stress eld for a geometry is known, the value of K may be obtained via KI = lim and similarily KII = lim
r0

r0

2r [yy ( = 0)]

2r [xy ( = 0)]

This is a rather cumbersome way of determining K and there are many other simpler numerical methods which can be used. The simplest method is probably the use of the J integral which can be converted to K as we will see later. 1.3.5 Connection between G and K Consider the strain energy released when crack grows an amount a, i.e. go from state A to state B in Fig. 1.14. The stress normal to the crack, yy , relaxes from yy (x) to zero over a while displacement from increases from 0 to u. 17

State A

Y X a

yy

State B

yy
u a X u

Energy released

Figure 1.14 Process of crack growth from State A to State B (Alternatively, one could consider the work required (i.e. energy absorbed) to close the crack an amount athe result is the same.)
a

Energy released (per unit thickness) =


0

1 yy (x)u(x)dx 2

where u(x) is the separation of the crack faces, the crack opening displacement. For Mode I: K yy (x) = 2x (dropping the subscript I for convenience) As shown in Fig. 1.14, the crack opening displacement is measured from the new position of the crack tip, i.e. with axis X , u(x ) = 8K(a + a) E x , 2

(x , because in the equation for crack opening displacement, the distance to the crack tip, r, is always positive.) Since x = x a we can write, u(x) = 8K(a + a) E 18 a x , 2

and substituting in the above equation we get 2K(a) K(a + a) Energy released = E
a 0

a x dx. x

Now by denition, energy released (per unit thickness)= Ga, therefore, Ga = Evaluating the integral,
a 0

2K(a) K(a + a) E

a 0

a x dx. x

a x dx = a, x 2

and the equation becomes, Ga = K(a) K(a + a) a. E

Dividing by a and letting a 0 so K(a + a) K(a) = K we get G= where (as before) E = E 1 2 E K2 , E plane strain plane stress.

So there is an one-to-one relationship between G and K for a linear elastic material. Therefore for a linear elastic material, the energy approach using G and the stress intensity approach using K are equivalent. When G reaches its critical value Gc then K must reach its critical value Kc . The above analysis is for Mode I loading, K = KI . The more general relationship for KI , KII , KIII which can be proven by an identical approach is, G= where G is the shear modulus. 1.3.6 Fracture toughness, KIC Near the crack tip, r 0, the stress and strain elds are well described by K. If two geometries have the same value of K then they have the same associated stress (and strain) eld. This is known as the concept of similitudeK contains all the 19
2 2 KI + KII K2 + III E 2G

information about loading and crack geometry. Thus, if we have a Mode I test specimen which fractures at a load with stress intensity KI we designate this value to be KIC . If an engineering structure with a crack is subjected to a loading which leads to a stress intensity factor KI KIC then fracture will occur in the structure. 1.3.7 Stress based criterion for fracture The stress at the crack tip is innite regardless of the magnitude of K. Therefore to formulate a stress based fracture criterion it is necessary to incorporate a distance. The fracture criterion is then phrased in terms of the attainment of a critical stress at a critical distance. A typical distance used is on the order of a grain size, say 10 m. For example if c = 500M P a and rc = 10m, then under Mode I loading KI yy = . 2r At failure KIC yy = c | at r = rc = 2rc KIC = c 2rc = 4MPa m The use of a critical stress at a critical distance is often called the RKR model, after Ritchie, Knott and Rice who rst proposed it. (See also the extract from the paper by Ritchie and Thompson, Appendix A.) Brittle, stress induced failure is known as cleavage. An alternative failure mode is ductile tearing which is associated with large plastic crack tip strains. Note that it is not necessary to know the mechanism of fracture in order to predict whether fracture will occur. For the engineer, the principal motivation behind fracture mechanics is to develop (reasonably) simple methods to predict fracture not to analyse micromechanisms of failure. Mechanisms of fracture are discussed in more detail later in the course. 1.4 Mixed mode fracture mechanics 1.4.1 Mode I and Mode II testing The majority of fracture testing is carried out under Mode I conditions as this is generally the critical mode for failure. However, increasingly mixed mode fracture toughness tests (combination of Mode I and Mode II) are being carried out to cover the full range of a materials response to mechanical load. 20

SFD

At crack tip: bending moment only, no shear => pure Mode I, M = 0

BMD

A SFD

At crack tip: shear force only, no bending => pure Mode II, M = 1

BMD

Typical test specimens for testing in Mode I and Mode II are shown in Fig. 1.15 and 1.16 respectively. By varying the position of the specimen relative to the applied load, the asymmetric bend specimen in Fig. 1.16 can also be used for mixed mode testing. 1.4.2 Denition of Mode-Mixity For a 2-D crack under mixed mode (combination of tension and shear) loading the stress ahead of the crack is given by KI yy = ; 2r KII xy = . 2r

Mode mixity is a measure of the ratio of Mode I to Mode II. It is usually expressed as 21

an angle (the phase angle) where = tan1 = 0 Mode I or sometimes as M where M= M = 0 Mode I 1.4.3 Crack path under mixed mode loading It has been observed that a crack subjected to mixed mode loading (Figure 1.17) will often not follow a straight path but will branch out of the plane. This has been observed for ductile metals and ceramics. KII KI (since tan(/2)= )

= /2 Mode II 2

M = 1; Mode II

yy xy

Figure 1.17, Shear and normal stresses for a crack tip under mixed mode loading 1.4.4 Criteria for crack kinking A number of criteria which demonstrate good agreement with experimental observations have been suggested. These give very similar predictions but are subtly dierent in the way the problem is formulated mathematically. 1. Crack branches in direction of maximum hoop ( ) stress (stress based criterion) 2. Crack branches in direction of maximum energy release rate. (Grith Criterion) 3. Crack branches in direction of local zero Mode II, in direction such that kII = 0 1.4.5 Maximum hoop stress criterion The rst criterion is the simplest to employ. The hoop stress eld is given by (see Appendix C) KII KI cos3 (/2) + = 3 sin (/2) cos2 (/2) . 2r 2r 22

Note that the hoop stress is normal to any branched crack.

Figure 1.18 Hoop stress, at an angle For the maximum hoop stress criterion the branching angle, , is then given by the angle which satises =0 KI
df1 () df () + KII II =0 d d

i.e.

where f1 () = cos3 (/2) and fII () = 3 sin (/2) cos2 (/2). Therefore, given KI and KII we can solve for . The solution for proved in Section 1.4.8 is given by: 2 1 KI KI for KII = 0. tan (/2) = + 8 4 KII KII

Note that since we assume that the critical distance is the same for all angles, a distance rc does not enter into the solution for the branching angle. 1.4.6 Grith criterion and kII = 0 criterion The other two criteria are more dicult to solve analytically. The approach taken is to consider the actual branch and calculate the new value of KI and KII (designated kI and kII ) for the branched crack (see Fig. 1.19).

K I , K II

k I , k II

Figure 1.19 Local stress intensity factors, kI and kII ahead of a branching crack. It can be shown that the local kI and kII are given by: kI () = a()KI + b()KII 23

kII () = c()KI + d()KII where, as indicated, a, b, c and d depend on the angle . Analytical solutions have not been obtained for these values so numerical solutions have been used. The two criteria for choosing the crack path, are then either based on the path which makes kII () = 0 or the path which maximises G(). For the second case, kI ()2 + kII ()2 E Problem is to maximize G() which is done numerically. (Note that the usual G G() = only gives energy for crack growth in the plane of the crack, i.e. for = 0. Under mixed mode conditions it may be energetically more favourable for the crack to grow at some other angle.) A comparison between the three theories is shown in Fig. 1.20. Note that for phase angle M = 1 (pure Mode II), 70 for maximum hoop stress theory and 80 for kII = 0 theory. Also included is some experimental data for alumina (a brittle ceramic). The maximum hoop stress theory seems to give the best agreement with the experimental data though there is signicant scatter in the data.

Figure 1.20, Branching angle, , based on three criteriacomparison with experimental data (from Suresh et al., J. American Ceramics Soc. 1990). 1.4.7 Dependence of fracture toughness on mode mixity The previous section discussed how to determine the angle for crack growth under mixed mode loading. More importantly we can also determine the dependence of the fracture 24

toughness on mode mixity. As stated earlier, the material parameters, GC or KIC apply to Mode I only, so it is of interest to examine mixed mode conditions. Of course, one can do this experimentally, simply by testing a specimen under dierent mode mixities, using for example the symmetric and asymmetric bend specimens, in Figs. 1.15 and 1.16, and determine the values of KI and KII at fracture. The advantage of theoretical models is that they provide insight into the fracture process and assist in extrapolating from one situation to another without the need for additional testing. Consider again the hoop stress criterion. Assume that the branching angle has been determined by the earlier analysis and is designated , then, 1 () = KI fI () + KII fII () 2r Fracture occurs when the hoop stress reaches a critical value, c , at a critical distance rc so we get 1 KI fI () + KII fII () = c 2rc

KIC is the fracture toughness when KII = 0, = 0, fI () = 1, therefore 2rc c = KIC .

Substituting, we get, KI KII fI () + fII () = 1 KIC KIC This is a universal curve with only one material parameter, KIC , which can be obtained from a single experiment. This is analagous to the concept of a yield surface in plasticity where only a single yield stress needs to be determined, rather than a set of values for every load condition. Figure 1.21 shows the fracture toughness curve data for alumina along with the experimental data. As always with ceramics there is a lot of scatter in the data but the trend is quite well captured by the maximum hoop stress theory, or a maximum energy theory (Grith theory). Of course, the validity of the expression for fracture toughness depends on whether the assumptions of the models are correctwhile either the Grith or maximum hoop stress theory works well in this case, neither may hold for another material. 25

Figure 1.21, Mixed mode fracture toughness locus for alumina, (from Suresh et al., J. American Ceramics Soc. 1990)

26

1.4.8 Derivation of branching angle based on critical hoop stress = 1 (KI fI () + KII fII ()) 2r

For a given KI and KII we seek to nd the angle at which the hoop stress, is a maximum. The r dependence is the same at all angles and is given by the square root singularity but the amplitude depends on the angle, . fI () = cos3 ; 2 Criterion is : fII () = 3 sin cos2 2 2

= 0 KI fI () + KII fII () = 0

where denotes dierentiation. Denote the solution to this equation to be . i.e. KI 3 cos2 sin 2 2 2 + KII 3 sin 3 cos sin cos3 2 2 2 2 2 =0

Now

3 1 KI cos2 sin = 3KII cos sin2 cos2 2 2 2 2 2 2 2 1 1 cos2 = (1 3 cos ) 2 2 2 4 1 sin = KII (1 3 cos ) 2 2 2 sin 2 = 2 sin cos

sin2 and above becomes

KI cos and since

we rewrite this as KI sin = KII (1 3 cos ) Thus branching angle satises sin 1 3 cos Note: for positive KI and KII , is negative. 27 = KII KI

For KII = 0 (Mode I), = 0 for KI = 0 (Mode II), 1 3 cos = 0 = 70.5 . (Note that f is negative for = +70.5 , i.e. hoop stress is compressive, see Fig. 3.5 in Appendix C) More general solution for KII = 0 sin 1 3 cos = KII KI

KII KII = 0 cos sin + 3 KI KI Use tan substitution: sin 2 = cos 2 = We get a quadratic in tan (/2) i.e. 2t2 KI t1=0 KII 2 tan 1 + tan2 1 tan2 1 + tan2

where t = tan (/2). There are two solutions to this equation, the maximum hoop stress is obtained for tan (/2) = 1 KI 4 KII KI KII
2

+ 8 for KII = 0

This result is plotted in the earlier gure in terms of M = 2/(tan1 KII /K1 ). Given KI and KII or given M we can predict the angle of crack growth.

28

1.5 Concept of small scale yielding All the results so far have been for a linear elastic material. Linear Elastic Fracture Mechanics (LEFM) can be applied to plastically deforming materials provided the region of plastic deformation is small. This condition is called the small scale yielding condition. The concept may be explained by Figs. 1.221.26. Here a numerical (nite element) analysis of a cracked plate is carried out. For simplicity, the plate material is assumed to be elastic-perfectly plastic (i.e. non-strain hardening). Figure 1.22 shows the nite element mesh and boundary conditions. Figure 1.23 shows contours of plastic strain obtained in the elastic-plastic nite element analysis. The upper gure is for plane strain, the lower is for plane stress. These plastic zones have the characteristic small scale yielding shape in both cases (note the larger plastic zone under plane stress). Recall from FFM that for plane strain the plastic zone size is approximately given by, rp = and for plane stress the result is rp = 1 K2 , 2 y 1 K2 2 3 y

i.e. three times larger. Here rp is the distance from the crack tip to the plastic zone boundary and y is the material yield strength. Figure 1.24(a) shows the numerically calculated elastic-plastic stress elds directly ahead of the crack for a plane strain analysis. Here equivalent (Von Mises) stress, e , divided by yield strength, y , is plotted so in the plastic zone e /y = 1. The estimated plane strain plastic zone size is indicated on Fig. 1.24(a). The numerically calculated value is somewhat less than this approximation. Figure 1.24(b) shows the same information though a larger scale is used. It is seen that for a large region the stress elds are well represented by the K eld. Figure 1.24(c) shows the same data again plotted on a log-log plot. Here the zone of K dominance and the plastic zone can be clearly seen. Figure 1.25 shows the same results from plane stress and the same trends are seen. Finally, in Fig. 1.26 a comparison is made between two dierent specimens, (both under plane strain conditions) a center cracked plate under uniaxial tension and an edge cracked plate under bending. (In the gure stress normal to the crack, yy , is plotted rather than equivalent stress, e ). The stresses are plotted when both these 29

specimens have the same K value, which is low enough so the plastic zone remains small. It may be seen that for distances greater than r/a 0.1 the stress elds in the two specimens dier but in the K dominant zone, 0.005 < r/a < 0.1, and in the plastic zone the stress elds are identical. In other words K is controlling the deformation in the crack tip region and two specimens (e.g. a laboratory specimen and a component) with the same K value have the same stress and strain elds near the crack. Until elastic-plastic fracture mechanics was developed, the precise form of these crack elds was not knownit is not necessary to know them. Provided the small scale yielding condition holds, two specimens with the same K value have the same crack tip elds. It is therefore acceptable to work with K and to deem fracture to have occurred when KI = KIC . 1.5.1 Size requirements for small scale yielding The requirement for small scale yielding is that the plastic zone size at fracture should be much less than the crack length. From the equations given earlier it may be seen that for small scale yielding conditions to hold under plane strain, rp = In other words a 0.1
2 1 KIC 2 3 y

a.

2 KIC . 2 y

The ASTM (American Society for Testing of Materials) species that for a valid plane strain KIC test, the specimen dimensions, crack length a, specimen thickness, B, uncracked ligament, b, where b = W a, must be greater than 2.5(KIC /y )2 . This implies that the specimen dimensions are about 25 times larger than the plastic zone size. The requirement that the plate thickness, B, is much greater than the plastic zone also ensures that plane strain rather than plane stress conditions prevail. Under these conditions, specimens with the same K value will have the same crack tip elds and fracture will occur when the K value reaches the plane strain fracture toughness value, KIC . As will be seen in Section 2, as the specimen size gets smaller or the plastic zone gets bigger, the small scale yielding condition is not satised and elastic-plastic fracture mechanics must be used.

30

Material behaviour

Symmetry boundary

100 90 80 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100

Symmetry boundary

1.0

0.75

0.5

0.25

0.0 -1.0 -0.75 -0.5 -0.25 0.0 0.25 0.5 0.75 1.0

crack length = 1

31

Plastic strains PEEQ

VALUE +0.00E+00 +2.00E-03 +4.00E-03 +6.00E-03 +8.00E-03 +1.00E-02 +4.14E+00

2 3 1

Plane strain

Plastic strains PEEQ

VALUE +0.00E+00 +2.00E-03 +4.00E-03 +6.00E-03 +8.00E-03 +1.00E-02 +2.23E+01

2 3 1

Plane stress

Figure 1.23, Crack tip plastic zones for plane stress and plane strain

32

(a)

(b)

(c)

Figure 1.24, Elastic-plastic crack tip fields for plane strain

33

Figure 1.25, Elastic-plastic crack tip fields for plane stress

34

Figure 1.26 Illustration of concept of K dominance for small scale yielding

35

2. Non-linear Fracture Mechanics LEFM works well as long as the zone of non-linear eects (plasticity) is small compared to the crack size. However in many situations the inuence of crack tip plasticity becomes important. There are two main issues: 1. In order to obtain KIC in a laboratory test, small specimens are preferred (for cost and convenience). However, to obtain a valid KIC measurement for materials with high toughness or low yield strength a very large test specimen may be required (c.f. size requirements for KIC testing). 2. In real components there may be signicant amounts of plasticity, so LEFM is no longer applicable. For these reasons we need to examine non-linear fracture mechanics where the inelastic near tip response is accounted for. 2.1 The J integral, (Rice, 1968) Consider a non-linear elastic material with strain energy density W such that, = W d.)

(Recall the earlier (equivalent) denition of W , W =

Consider a body with an applied stress as shown in Fig. 2.1. Consider an area of the body A enclosed by the boundary and dene the closed line integral, I ,

Figure 2.1 Closed line contour for a loaded body u ds, I = W dy t x 36

where t is the traction on , u is the displacement and ds measures distance along the curve . 2.1.1 Denition of traction, t The traction vector t on a plane is the average force per unit area exerted by particles on the positive side of the plane on particles on the negative side of the plane. The traction vector will depend on the plane considered as illustrated in Fig. 2.2.

t
n

Figure 2.2 Traction t, dened relative to a given plane The traction is dened as follows: t = n where n is the unit normal to be plane in question and is the stress matrix. (This relationship is known as Cauchys theorem and may be stated as follows: The traction at a xed point on a surface depends linearly on the normal at the point) Note that traction is a vector and stress is a matrix (tensor). 2.1.2 Determination of path independent line integral, J Returning to the integral I . I =

W dy t

u ds x

It can be shown by using the divergence theorem that for an equilibrium stress eld and associated strain eld with W the associated strain energy density, provided there are no singularities in the region A, the integral I is zero for any path . This leads us to the denition of the path independent J integral. We now consider a cracked body and examine the path, shown in Fig. 2.3, where is split into 1 , 2 etc. 37

Figure 2.3 Paths used in denition of the J integral In Fig. 2.3 2 is the remote boundary, 1 surrounds the crack tip, + and are parallel to the top and bottom faces of the crack tip respectively. Since the region bounded by contains no singularity, I =
2 ++ +1 +

W dy t

u ds x

= 0.

+ and , are along the crack face and with the axis dened as shown dy = 0. Also, by denition for a crack, t = 0, i.e. there are no tractions on the crack face.
+

=0
2

+
1

=0

or
2

=
1

=
1

where the minus sign for 1 indicates that the direction of integration is reversed for 1 . We dene, J=

W dy t

u ds x

where is any path starting on the bottom crack face and nishing at the top. The value of J is constant no matter what path is chosen. (The direction of the integration is the same as that for 2 in the diagram.) This denition set the stage for non-linear fracture mechanics. The J integral is path independent for any non-linear elastic material. Plastically deforming materials can be represented by non-linear elasticity and thus fracture mechanics can be extended beyond linear elasticity and K. In a similar manner to K, J has an interpretation both as an energy and a stress quantity. We start with its interpretation as a measure of stress intensity. 38

2.2 Power law hardening materialsThe HRR eld We start with the path independent J integral. J=

W dy t

u ds x

Take the origin at the crack tip and choose a circular path as shown in Fig. 2.4.

Figure 2.4 Denition of axes for determination of HRR eld Along this path, y = r sin dy = r cos d , ds = rd

J =

W cos n

u x

r d.

Assume a separable solution for displacement, u, as the crack tip is approached, u = r+1 u() du r () dr

(The symbol indicates proportional towe are not interested in the precise form of the elds at this stage, only in the form of the solution.) We rst wish to determine the value of which gives us the order of the singularity at the crack tip. Since J is independent of path, we can take r as small as desired. If J is to be nite and non-zero, then we must have W cos n u x 1 r as r 0

Both terms in the bracket are of order O( ) 1 r 39 as r 0

For a linear elastic material, . Therefore at the crack tip, r . r2 = = 1/2 i.e. a square root singularity for a linear elastic material. Now consider a non-linear elastic material with power law hardening ( n ; or
1/n

1 r

) As before, let r () r n () r
n+1 n

Since =

1 r n n+1

r1/n+1 () rn/n+1 ()

This is the HRR singularity, (after Hutchinson, Rice and Rosengren, 1968. For a more general power law material with uniaxial stress-strain law, =
0

where

0,

and 0 are the reference strain and stress and is a scaling factor, then we

can write, ij /0 = Ar1/n+1 ij (; n);


ij / 0

= An rn/n+1 ij (; n),

where A is the undetermined amplitude. By substituting these expression for stress and strain into the integral expression for J on the previous page, and after some manipulation, we get, J = An+1 0 0 In 40

where In is an integral containing terms which depend only on the hardening exponent n. We can therefore replace the constant A in the expressions for stress by J above so ij /0 = J 0 0 In r J 0 0 In r
1/(n+1)

ij (; n)
n/(n+1)

ij /

ij (; n).

The terms In , ij (; n), ij (; n) are dimensionless quantities which depend on the hardening exponent, n and have been determined numerically. The subscript ij in the above equations indicate that stress and strain are matrices (tensors) with six components, i, j = 1 3. The distributions for n = 3 and n = 13 are given below. The intensity of the elds also depends on whether plane stress or plane strain conditions prevail. Plane strain is always the most severe. This is in contrast to linear elasticity where the in-plane stresses, xx and yy are the same for plane stress and plane strain.

Figure 2.5 Variations of angular stress and strain functions for a Mode I crack under plane strain. (Hutchinson, J.W., Technical University of Denmark, 1982) Note that this result also holds for an elastic-plastic power law material, i.e. E = y when < y y
n

when > y 41

Here = 1, and

and y are used instead of

and 0 and have the interpretation of

yield strain and yield stress respectively. Close to the crack tip the plastic power law term will dominate so the HRR eld applies. Thus J fullls the role of K for a non-linear power law material, J characterises the intensity of the near tip eld. The condition for fracture is simply J = JIC . While power law hardening is an approximation to the material behaviour, J may still be considered as a measure of the intensity of the crack tip elds for any material description. In the same way as K, J will depend on crack length and geometry and also on the magnitude of the load. There are also tabulated geometry factors for J similar to those for K. These will be discussed later. 2.3 Crack tip opening displacement The CTOD approach is commonly used in elastic-plastic fracture mechanics whereby failure occurs when the crack opening displacement reaches a critical value. For J dominance the J and the COD approach are equivalent. As we discussed for the K eld, the COD can be determined directly from the crack tip elds, u(r) = 2uy (r, ), J y y In r
n/(n+1)

uy (r, ) =

r uy (; n) r

u
Figure 2.6 Denition of crack opening displacement Thus, for a given material, (i.e. constant n, , In ,
0

and 0 ) the CTOD depends only

on J and distance r. Since the CTOD depends on r and is zero at r = 0,(except for perfect plasticity n ) the denition of CTOD is somewhat arbitrary. 2.3.1 Conventional denition of CTOD The CTOD has been dened mathematically as being the crack opening at the point where 45 lines drawn from the crack tip intersect the crack anks as shown in Fig. 2.7. It can be shown, using the HRR eld, that for this denition of CTOD, t = dn 42 J , y

where dn depends only on n. For moderate hardening plane stress, dn 1, for plane strain, dn 0.5. Thus there is a one-to-one relationship between CTOD and J for a given material and any J based approach can be converted to a CTOD based approach.
o

45

45

Figure 2.7 Conventional denition of crack opening displacement The above equation is consistent with the expression introduced in FFM, = G . my

Here the symbol m is used rather than dn and the elastic energy release rate G is used rather than J. The equivalence between G and J for small scale yielding conditions is discussed next. 2.4 Relationship between J and G It can be shown that J is in fact equal to the change in potential energy G for a nonlinear elastic material. The proof is dicult and beyond the scope of the course but an outline of a proof is given below. Similar arguments to those used in establishing the relationship between K and G can be employed here. (The original proof due to Rice did not follow this approach.) We consider a crack growing by an amount a. Then Energy released = Ga yy ua. For a power law material yy J n+1 and u J n+1 43
n 1

so Ga Ja and it can be shown that the proportionality constant is equal to 1. The equality of the line integral J and the energy release rate G holds for any elastic material J =G= 1 U B a

and for a linear elastic material we have the additional equality J =G= K2 1 U = . B a E

While this latter relationship strictly only applies for a linear elastic material, if the zone of nonlinearity near the crack tip is small, (less than about one twentieth of the crack length), this latter relationship between J and K may be applied. As discussed previously, the condition that the plastic zone is small is known as the small scale yielding condition and is the basis of the use of LEFM for metals. If the zone of plastic strains is small enough we can ignore J and work with K alone. However, even under LEFM in ductile metals, the stress and strain elds close to the crack tip are given by the HRR eld, not the K eld. Two cracks with the same K value, under small scale yielding conditions, will have the same J value via the above relation and therefore the same HRR stress eld at the crack tip. Therefore K or J equivalently characterise the crack eld. The relationship between J and G demonstrates that the J integral may simply be thought of as another way of obtaining the energy release rate G. However, there are some diculties in considering J to be the energy release rate for a real elasticplastic material. When a crack grows there is always elastic unloading ahead of the crack which invalidates the assumptions inherent in the derivation of the relationship between J and G. However, J still retains its meaning as a stress intensity measure, the magnitude of the stresses ahead of the crack tip. As stated by Hutchinson: Tempting though it may be to think of the criterion for initiation of crack growth based on J to be an extension of Griths energy balance criterion, it is nevertheless incorrect to do so. That is not to say that an energy balance does not exist, just that it cannot be based on (the deformation theory) J. (The meaning of the term deformation plasticity here

J.W. Hutchinson, Journal of Applied Mechanics, 1983 (see Appendix A). 44

is equivalent to the term non-linear elasticity.) Under large scale yielding conditions which we will discuss later, J is sometimes split up into elastic and plastic parts, i.e. Jtotal = Je + Jp The term Je is given by K2 Je = E i.e. it is the value of J if there was no plasticity. Sometimes (and somewhat confusingly in view of its use as a symbol for the non-linear elastic energy release rate) the symbol G is used to indicate the elastic part of J. 2.5 Evaluating J for test specimens and components We have seen the importance of J in non linear fracture mechanics. The next question is how to evaluate it. The line integral approach is rather awkward and usually requires numerical techniques such as nite element analysis, so approximate methods have been developed to estimate J in test specimens and in actual components. In this section we will discuss two of the most popular methods estimate Jthe use of the factor and the GE-EPRI J estimation methods for power law materials. Most other J-estimation procedures are based on these two methods. To evaluate J using the factor we exploit the relationship between J and G. To show how this is done, it is helpful to revisit the concept of the limit load or plastic collapse load of a specimen. 2.5.1 Limit load and the denition of We examine an elastic-perfectly plastic material. The limit load (sometimes called the collapse load) is the load at which plastic collapse occurs for such a material. Consider a beam in bending (Fig. 2.8).
y

ML =

y BW 2 4

Figure 2.8, Illustration of limit moment for a plastic beam in bending. Next consider a cracked beam in bending with a << W . 45

B M W a a M

Figure 2.9, Edge cracked beam in bending. It can be shown that for plane stress conditions, MLC = y B(W a)2 W 2B = y (1 a/W )2 4 4

Note that W is replaced by W a, i.e. the collapse moment for the cracked plate is the same as that for a plate of width W a. The additional subscript C here emphasises that it is the solution for a cracked plate. Often the C is left out. For a center cracked plate with crack length 2a and plate width 2W , in tension with a << W , subjected to a load 2P , the limit load under plane stress conditions, is given by

PLC = y (W a)B. Limit load solutions are commonly used in fracture mechanics. The ratio between load and limit load is a measure of the extent of plasticity and provides a good means of compar2W

ing two geometries. For example two dierent geometries at the same ratio of load to limit load have similar J integral values. Limit loads have been obtained analytically and numerically and have been tabulated for a wide range of geometries;

For a three point bend geometry of length 2L under plane strain conditions and using the Von Mises yield criterion: a 2 PLC = y 1.22 1 W 3 46
2

W 2B . 2L

This solution may be compared with that for a small crack under pure bending given earlier. For a three point bend bar of length 2L with applied load P , the moment at the crack plane is M = P L/2, and the 1.22 term in the equation is due to a nite sized crack. 2.5.1.1 Plane strain limit load The 2/ 3 term in the limit load denition above is due to the plane strain condition, i.e. under plane strain uniaxial tension, yy = , then for von Mises yield,
2 2vm = (xx yy )2 + (xx zz )2 + (yy zz )2

xx = 0,

zz = 0.5(xx + yy ) (in plasticity)

vm =

At yield vm = y , therefore for yielding under tension in plane strain = 2y / 3, so the limit load is higher by a factor of 2/ 3 under plane strain conditions compared to plane stress conditions when the yield condition is simply = y . 2.5.2 Denition of reference stress A quantity closely related to the limit load is the reference stress. The reference stress is a measure of the proximity to collapse of the structure, and is dened as: ref = y P PL .

3 2

Thus, plastic collapse will occur for a perfectly plastic material with yield strength y when ref = y and since PL is proportional to yield stress, ref is independent of y . The reference stress is a correction to the applied stress to take account of the eect of the crack on the response of material. Use is made of the reference stress in the application of structural integrity assessments, as will be seen later in the course. 2.5.3 Use of limit load to dene the factor Limit loads are useful in determining the parameter, which is used to relate J to the area under the load displacement curve. Recall the energy denition of J, J = 1 U . B a

47

Under applied displacement,

U = Ue =
0

P () d J =

1 B

P d. a

It can be shown that this is equivalent to writing J= B(W a)

P d =
0

A, B(W a)

where is a geometry factor related to the compliance and A is the area under the load displacement curve. The above has been derived for applied displacement. However since J is independent of the mode of loading it also holds for applied load. This form for J is convenient because, provided is known, J in an experiment can be obtained from a load displacement history, which is easy to measure. Consider a three point bend specimen with a rigid elastic, perfectly plastic material. W under plane stress made of a

PL = y

W 2B (1 a/W )2 2L

Figure 2.11, Limit load for an edge cracked beam under three point bending. Under displacement control the potential energy, U , is given by, U = Ue = P (see Fig. 2.12). Therefore U = P = Pl = y W 2B (1 a/W )2 . 2L

U = P PL P

Figure 2.12, Load-displacement behaviour for a rigid-elastic, perfectly plastic material. 48

We can dierentiate the above expression for U , keeping displacement xed, to obtain J, J = 1 U B a =

2 W2 y (1 a/W ). W 2L

Since P = PL = y

W 2B W2 PL (1 a/W )2 (1 a/W ) = 2L 2L B(1 a/W ) J = 2 A. B(W a)

i.e. = 2 for this load conguration. For a center-cracked-tension geometry,


, /2

U = P = y B(W a) J = 1 U B a = y =

1 P B(W a)

i.e. = 1 for this loading conguration. Note that if, as is often the case, the crack length is designated 2a instead of a, then in
W

the J equation the load per crack tip, i.e. P/2, must be used to have = 1.
, /2

The above is an illustration for perfect plasticity. There are more rigorous proofs, (given in Kanninen and Poplar) which show that in general for a low hardening material, is close to 1 in tension and 2 in bending.

Many crack geometries are loaded by a combination of bending and tension. e.g. for a compact tension specimen, = 2 + 0.52(1 a/W ).

2.5.4 value for a linear elastic material We can also evaluate for a linear elastic material. 49

Figure 2.14, Load-displacement curve for a linear elastic material For a linear elastic material: J =G= where C is the elastic compliance P = J= P2 = P C C 1 2 dC(a) P , 2B da

1 dC 1 1 dC P = A 2B C da B C da

and since the alternative equation for J is J= A B(W a) W a dC C da

e =

Thus if the compliance is known, e can be determined. The subscript e is used here to emphasise that this is the value for an elastic material. In general e and p are not equal even for the same geometry. 2.5.5 Evaluating J for an elastic-plastic material Most materials are elastic-plastic and when a specimen is loaded part of it will have yielded while the rest will be elastic. Under these conditions the total J value may be estimated by J = Je + Jp e p Ae + Ap B(W a) B(W a) 50

where Ae = 1/2(P e ) and Ap is the remaining portion of the load displacement curve as shown in Fig. 2.15.

Figure 2.15, Load displacement curve for an elastic-plastic material Alternatively, are more usually, the elastic part of J may be determined from the linear elastic K value, which is easily obtained, and then

J=

K2 p + Ap E B(W a)

The J value in an experiment can then be obtained simply by measuring the area under the load-displacement curve. For a deeply cracked bend specimen, e 2, so a commonly used J approximation for a deep cracked bend specimen is, 2 A, B(W a)

J=

where A is the total area under the load displacement curve. 2.5.6 GE-EPRI J Estimation Scheme The GE-EPRI scheme is an alternative approach to calculating J. GE-EPRI stands for General Electric-Electrical Power Research Institute, where the method was developed. Consider a body loaded by remote stress shown in Fig. 2.16. 51

Figure 2.16 Cracked body under remote tension First consider a linear elastic material with, = (E = 0 / 0 ) 0 0 where 0 and
0

are a normalising stress and strain (material parameters).

By dimensional analysis we can write (x, y) = [ /0 ] f (x, y) 0


0

where and are the stress and strain at any point in the body (Fig. 2.16). This simply states that stress and strain in a linear elastic body are proportional to applied load. Since J=

where L is an appropriate length. Note that because we integrate with respect to distance along the contour the dependence on x and y does not enter the expression for J. We can write J = a0
0

(x, y) = [ /0 ] g(x, y)

W dy t

u ds x

J ( /0 )2 0 0 L

0 52

H(a/W )

Here the characteristic length has been taken to be the crack length a and H is the proportionality constant which depends only on geometry. Compare with K = aY (a/W ). Since for a linear elastic material, J = K 2 /E J=
2 aY 2 (a/W ) (in plane stress) E 2 0 aY 2 (a/W ) = 0

= 0

0 0 aY 2 (a/W ).

So for linear elasticity, H(a/W ) = Y 2 (a/W ). This is (yet another) way of expressing J (or K) for a geometry. This particular normalization proves useful as it can be generalised to dierent types of geometries and materials. For a plastic (or non-linear elastic) material with power law hardening behaviour, /
0

= (/0 )n

(x, y) = [ /0 ]f (x, y, n) 0 (x, y) = [ /0 ]n g(x, y, n) 0 Note the additional dependence on the hardening exponent n. Using the same argument as before we can write J 0 0 ( /0 )n+1 L and thus J = 0 0 a 0
n+1

H(a/W, n).

It proves useful to normalise by the limit load rather than 0 so we rewrite as P J = 0 0 a P0


n+1

h(a/W, n)

where P is remote load and P0 is limit load. The function h depends only on n and a/W and can be tabulated in a similar fashion to Y . Note the dimensions of each term 53

in the expression:

J 0
0

a P/P0 h

[stress][length] dimensionless [stress] dimensionless [length] dimensionless dimensionless

Some values of h are shown in Fig. 2.17 for an edge cracked panel subjected to bending for a range of a/W ratios and strain hardening exponent n.

a/W=1/8

Figure 2.17, h function for edge cracked panel in bending. Note that except for the shallow crack, (a/W = 1/8) at high n, h is quite weakly dependent on n and close to unity. Typical values of n for metals range between 5 and 20 (0.05 < 1/n < 0.2) 2.5.7 Elastic-plastic material behaviour The function h is based on purely plastic (power law) behaviour. For elastic-plastic behaviour with E = y when < y y
n

when > y ,

J can be partitioned as before into elastic and plastic parts, J = Je + Jp 54

with Jp evaluated using the estimation scheme above (taking = 1, and Je = K 2 /E .

and 0 = y

In order to get good agreement with numerically calculated J values in the elasticplastic regime, Je is usually adjusted slightly. The precise form of Je used in the EPRI scheme is described in Section 5.4 of the book by Kanninen and Poplar. It is based on a plastic zone correction approximation for J as discussed below. 2.5.7.1 Plastic zone correction When the plastic zone size is relatively small (on the order of a tenth of the crack length) modications to the LEFM K are sucient to account for the eects of material nonlinearity. The eect of crack tip yielding is to reduce the eective load supporting area at the crack tip relative to an equivalent linear elastic material. This eect may be accounted for approximately by using an eective crack length, ae , in the denition for K, i.e., Ke = Y (ae ) ae

where Y is the LEFM geometry factor. The eective crack length ae is given by, ae = a + ry , where a is the actual crack length and ry is half the plane strain or plane stress plastic zone size. Consider the case of a small crack in a large plate under tension and use the plane stress expression for the plastic zone size, ry = For a small crack, Y = 1 2 K y
2

(i.e. independent of crack length) and we get a Ke = . 2 1 1 2 y

Using the plane stress plastic zone size ensures that we will get the highest value of Ke , Ke will be overestimated if conditions are predominantly plane strain. 2.5.8 Overall J estimation procedure 55

The nal form of the GE-EPRI J estimation scheme, is then J = Je (ae ) + Jp (a, n), with Jp evaluated using the estimation scheme and Je evaluated using the plastic zone correction as discussed above, with ry based on the unmodied crack length, a. A comparison between the numerically calculated J value and the GE-EPRI approximation is shown in Fig. 2.18 (This gure is adapted from the gure on page 318 of Kanninen and Poplar). The GE-EPRI scheme can be used to estimate J in materials which obey power law hardening in the plastic regime. However, in order for the estimate of J to be accurate, the material behaviour should be well approximated by a linear elastic-power law hardening law.

Solid line: finite element solution Dashed line: GE-EPRI approximation

Figure 2.18, Comparison between GE-EPRI J estimation scheme and nite element calculations for n = 3 and n = 10. Note that the approach and the GE-EPRI scheme are equivalent in practice. The values of and h must be determined numerically, though good approximations can often be made, e.g. taking = 2 for bending, 1 for tension. Exact values of the J integral could be obtained from a full 3-D nite element analysis of the specimen/component and calculating J using the line integral denition. However, this is expensive on time and resources so approximate schemes are preferred. Note also that occasionally the symbol G is used to indicate the linear elastic energy release rate which is equal to the elastic J value, Je . This should not be confused with 56

the non-linear elastic energy release rate, also given by the symbol, G or here G and is equal to the total J. We have shown how to calculate J for test specimens. However, to carry out a failure assessment, we still need to determine J in the actual component we are interested in. Again this must be done either numerically or by some other approximation technique. Failure assessment procedures such as BSPD 7910, to be discussed in more detail later incorporate methods to do this and in the elastic-plastic fracture regime they are simply approximation schemes for the J integral. 2.6 Application of non-linear fracture mechanics The theoretical basis behind the application of non-linear fracture mechanics is illustrated by the gures overleaf. Figure 2.19 shows numerically calculated elasticplastic stress elds in the vicinity of a crack at dierent load levels for two dierent geometriesa three point bend geometry (tpb) and a centre crack panel in tension (ccp). The tpb geometry has a plate width, W , of 40 mm and and a crack length to specimen width, a/W , of 0.5. The ccp geometry has W = 400 mm (i.e. a very large plate) and a/W = 0.1. In each gure the K value for the tpb and ccp specimen are the same. Here the material behaviour in the elastic-plastic regime is assumed to follow a power law relation with yield stress, y = 500 MPa and n = 5. If both specimens are loaded at a low load, K = 5 MPa m, then it is seen that the specimens deform predominantly elastically and the crack tip stress distributions are the same. As the load is increased to K = 30 MPa m the region of plastic deformation

increases but as small scale yielding conditions are satised both specimens still have the same stress distribution at the same K value. Note, however, that the near tip elds are represented by the HRR eld rather than the K eld. At the highest load level, (K = 85 MPa m) the zone of K dominance has almost disappeared for the smaller tpb specimen. However, though not shown, the near tip elds are still represented well by the HRR distribution. It may also be seen that the crack tip stresses are dierent for both specimens even though the elastic K value is the same. (The dierence is not very obvious on the log-log scale.) The reason for this dierence is explained by the J versus K plot shown in Figure 2.20. It is seen in Fig. 2.20(a) that, because the ccp specimen is larger, the small scale yielding condition holds up to the maximum K value applied (i.e J = K 2 /E holds for this specimen up to 85 MPa m). For the tpb specimen, however, as the plastic zone 57

size increases, the J value exceeds that given by the small scale yielding estimate. Figure 2.20(b) illustrates that, as expected, under small scale yielding conditions both specimens at the same K level have the same near tip stress distributions. However, at the largest load level, as seen from the J vs. K diagram, the J values for the tpb specimen is higher than that for the ccp specimen. Hence, the tpb specimen has a somewhat higher stress eld than the ccp specimen. Under these conditions the stress elds are not characterised by K and J (and the HRR eld) must be used instead.

tpb
20 mm 40 mm

ccp

80 mm 800 mm

Figure 2.19, Comparison of K field and numerical crack tip fields for two geometries

58

(a)

(b)

(c)

K = 85 MPa m1/2

Figure 2.20, (a) J plotted against K for ccp and tpb geometry, (b) comparison between K field and numerical stress field at K = 30 MPa m1/2 (c) comparison between HRR field and numerical stress field at K = 85 MPa m 1/2

59

2.7 K dominance, J dominance and size requirements As discussed earlier, far away from the crack tip the K solution is invalid. Close to the crack tip there will be non-linearity (plasticity) so the K-eld will not be valid there either. In a KIC test, there must be a region where the K eld is in agreement with the stress elds in the specimen. Assuming we have sucent thickness to guarantee plane strain conditions, in order to have a valid KIC test, we require a small plastic zone, rp . For Mode I, 1 rp = 3 We require a K y
2

= 0.1

K y

rp . The ASTM requirements are a, B, W/2 > 2.5 KIC y


2

i.e. specimen dimensions are about 25 times larger than the plastic zone size, e.g. For AISI 4340 steel, KIC = 65 MPa m, y = 1400MPa then minimum plate size W is 11 mm. But for A533B nuclear reactor grade steel, KIC = 180MPa m, y = 350MPa, minimum W for a valid KIC test is 1.3 m. For JIC testing size requirements are less stringent. The requirement is that the J solution (i.e. the HRR eld) dominates over a region signicantly greater than the crack tip opening displacement. (In this region the stresses are strongly aected by the blunting of the crack.) It has been found from numerical studies that bend geometries show a larger zone of J dominance than tension geometries. Therefore dierent size requirements are needed for these two types of geometries. For a centre crack tension specimen the zone of J dominance becomes vanishingly small relative to the crack tip opening displacement when J > (6 103 )a0 . In other words the size requirement for a tension specimen is that a > 150JIC /y . For a deeply cracked bend specimen, J dominance is maintained up to about J = 0.07ay giving a size requirement that a > 15JIC /y . In practice the ASTM JIC standard recommends that deeply cracked bend specimensthe compact tension specimen (which despite its name is a bend dominated 60

geometry) or the single edge bend (three point bend) specimen, are used to obtain JIC and that a, b, B > 25JIC /y . These limits have been conrmed from experiments, i.e. as long as these requirements are met the JIC value obtained is independent of the specimen size or typeit is a material property. Ii should be pointed out that since both KIC and JIC are material properties, KIC can always be calculated from JIC using the relationship KIC = JIC E .

So a valid JIC test can be used to obtain KIC . The ASTM standard E1820 designates a KIC calculated from a JIC as KJIC , though in principal, KJIC KIC The concept of J dominance is illustrated by Figure 2.21. (The geometries analysed are the same as those shown in Figs. 2.19 and 2.20, though the loads are higher). Figure 2.21(a) and (b) illustrate the stress elds for a tpb specimen at two dierent J levels (note the way the x axis is normalised). It may be seen that at J/ay = 0.004 the HRR eld gives a good representation of the stress elds in the specimen for distances on the order of the CTOD (CTOD 0.5J/y so the x-axis extends for about 10 crack tip openings.). At the higher load, Fig. 2.21(b), J dominance is lost and the HRR eld agrees with the stress elds only very close to the crack tip, at distances less than the COD. For the ccp geometry, as shown in Fig. 2.21(c), even at the lower normalised J value the HRR eld is not close to the stress elds in the specimen. The J levels at which J dominance for these two specimens are lost are consistent with the ASTM limits specied earlier. It should be pointed out that when J dominance is lost the stress eld will always fall below the HRR distribution. In other words it would be conservative to assume that the elds are J controlled. However, when determining the fracture toughness, it is of course important to test the worst case situation, which will be the J dominance case. This is why the standards specify the use of a bend geometry with size requirements to ensure J dominance. In recent years, attention has turned to taking advantage of some of the conservatism inherent in the applicant of J based fracture mechanics to cracks loaded predominantly in tension and with signicant amounts of plasticity. However, this work is still at the research stage. Note that the loss of J dominance does not imply 61

that J loses its path independence. It simply means that the crack tip elds are not well described by the HRR eld with amplitude J.
(a)

(b)

(c)

Figure 2.21, Comparison of numerical stress fields and HRR fields for (a) and (b) tpb geometry and (c) ccp geometry

The idea of K and J dominance is summed up by Fig. 2.22, which shows schemat62

ically the K and J dominance zones at two dierent loads. The process zone, indicated in the gure, is the damage zone near the crack tip, whose size is on the order of the CTOD. In order for J dominance to hold, the region where the HRR elds agree with the stress elds in the specimen, as determined by a nite element analysis for example, must be larger than this process zone size.

ILLUSTRATION OF K DOMINANCE (at low load, or very large specimen)


Process Zone

K field HRR field

Specimen

ILLUSTRATION OF J DOMINANCE (at higher load or small specimen)


Process Zone

HRR field

Specimen

Figure 2.22, Illustration of K and J dominance in elastic-plastic specimens.

63

2.8 Standard test to determine JIC In this section we discuss the ASTM E 182001 test standard for measurement of fracture toughness. A three point bend or compact tension specimen is recommended with 0.45 < a/W < 0.7 and a valid test requires that a, b, B > 25 JIC . y

The load and load point displacement of the specimen are measured during the test and J is calculated from the area under the load displacement curve. JIC is the value of J at crack initiation. For very ductile materials an obvious initiation toughness, JIC is dicult to measure as there is usually some stable ductile tearing before nal failure. For such materials a J resistance curve, (J versus a) is measured and the curve extrapolated back to a = 0 to obtain JIC . The change in crack length a may be obtained using the unloading compliance method (see section 2.8.1) or some other method. J is then plotted versus a and JIC is determined by extrapolating back to a = 0. Crack blunting can give an apparent a even though crack growth has not occurred and this must be accounted for in determining JIC . This is discussed in more detail in ASTM E 1820. 2.8.1 Compliance method to estimate a Crack growth is generally non-uniform through the thickness of the specimen. As the centre of the specimen is under plane strain conditions crack growth tends to be higher there. Therefore inspection methods which rely on observing the crack growth on the surface of the specimen may not be suciently accurate. Furthermore, such inspection requires some operator judgement and are dicult to automate. The crack compliance method avoids both of these problems (to some extent). An elastic-plastic material unloads elastically and we have the relationship = C(a)P. If a specimen is unloaded a small amount during testing the elastic compliance can be obtained from the load-displacement curve (see Fig. 2.23). If the compliance changes this can only be due to a change in crack length. The function C(a) is tabulated for standard fracture specimens, and thus if the compliance is measured the amount of crack growth a can be inferred. If the crack growth is non-uniform through the thickness, then this crack length will be the average crack length through the specimen. 64

Figure 2.23, Estimation of crack length using linear elastic compliance An alternative approach in constructing a J a resistance curve is to stop the test, break open the specimen and determine the amount of crack growth optically. The disadvantage of this method is that multiple specimens will be needed to construct the J-resistance curve. Even if the compliance method is used to determine the crack length, the specimen should be broken open at the end of the test to compare the actual initial and nal crack length with the values estimated using the crack compliance method.

65

3. Micromechanisms for ductile and brittle fracture The student is referred to the extracts from the paper by Ritchie and Thompson, On macroscopic and microscopic analyses for crack initiation and crack growth toughness in ductile alloys (see Appendix A). As discussed in previous sections, the criterion for a material to fail by fracture is that, J = JIC . Provided J can be obtained and JIC is known (or can be measured) we can determine whether or not fracture will occur. However, this does not tell us anything about the mechanism of fracture, for example, whether fracture will be by a brittle or ductile mode. We have already looked at a simple cleavage model for brittle failure. The model for cleavage is the RKR (Ritchie, Knott, Rice) model which says that brittle crack extension occurs when the local tensile opening stress ahead of the crack exceeds a local fracture stress over a microstructurally signicant distance. 3.1 Micromechanism of cleavage failure Recall from rst year mechanics of materials that the ideal strength of a crystal is on the order of E/10, which should be the stress required for cleavage. However, due to yielding and crack blunting, classical plasticity, predicts that these stress levels will never be reached even at the tip of a crackthe maximum stress at the tip of a blunting crack is about 3y . In order to explain why cleavage fracture can occur an additional micro-mechanism is required. The generally agreed micro-mechanism for steels is that the dislocations that are emitted from the crack tip build up at the adjacent grain boundary, amplifying the local stress (see Fig. 3.1).

Dislocation pile-up

Carbide particles

Figure 3.1 Schematic of cleavage failure at the microscale The stress at the head of the dislocation pileup (at the grain boundary) is n times the stress at the crack tip, where n is the number of dislocations. This stress may then be large enough to initiate failure at grain boundary inclusions (e.g. carbides for a ferritic steel) and failure of the inclusion triggers failure in the associated ferrite grain 66

(in a carbon steel) The resultant micro-crack then links up with the main crack and the crack grows unstably.
Using dislocation arguments, the stress required for cleavage, f , is given by f

Gm , 5 dg

where G is the shear modulus, m is the relevant surface energy (including local plastic work) and dg is the grain size. Since G and m are weakly dependent on temperature,
this cleavage stress f is relatively independent of temperaturefor a typical carbon

steel it is about 800 MPa. 3.2 Prediction of fracture toughness using the RKR model and the HRR eld Using the RKR model and the HRR eld we can write down a micro-mechanics based
failure equation (see Fig. 3.2). The RKR criterion is that for a Mode I crack, 22 = f ,

at r = rc where, from our micro-mechanical model above, rc is on the order of a grain


size. (Following the notation of Ritchie and Thompson, l0 is used here instead of rc ).

22

22 J = 0 0 0 I n r
22

1 / n +1

~ 22 (0; n)

Figure 3.2 RKR model for cleavage failure The failure equation becomes,
f = 0

JIC 0 0 In l0

1/n+1

22 (0; n).

Now,

= 0 /E, and and In are xed for a given material and temperature so we
f

rewrite: = A0 EJIC 2 0 l0 1 In

1/n+1

,
1/n+1

where A = 22 (0; n) 67

EJIC =

f A0

n+1 2 0 l0 .

As discussed previously, there is a one-to-one relationship between JIC and KIC ,


2 i.e. JIC = KIC /E . Then f A0 (n+1)/2

KIC =

E JIC =

l0 ,

(Taking E = E for simplicity). Rearranging, we get KIC = 1 A(n+1)/2


(f )(n+1)/2 0 (1n)/2 l0 .

For n = 10 (a typical value for steel), the values are, In = 4.54, 22 = 2.5, A 2.2 (taking = 1). It may be assumed that A is independent of temperature (n depends weakly on temperature). Inserting these values gives,
4.5 KIC = 0.01(f )5 .50 l0 .

A direct relationship between KIC , the failure stress, the yield strength and the grain
size is obtained (since in this model, l0 = dg ). Consider the eect of an increase in temperature on the fracture toughness. If temperature is increased, f is unchanged, 0 decreases, so KIC increases. If the grain-size is decreased both 0 and f increase However, the relative increase in f is higher, so the overall eect is to increase KIC .

3.3 Micromechanism of ductile failure At high temperatures the yield stress decreases so the crack tip stresses go down and cleavage fracture becomes less likely. The dominant failure mechanism is then ductile tearinga process known as micro-void coalescence. In steels, voids are initiated by debonding from large inclusions, e.g. manganese sulphide particles. These voids grow, coalesce and link up with the main crack, leading to slow and stable tearing with a large amount of absorbed energy (see Fig. 3.3).

Sulphide particles (Inclusions)

* l0

Figure 3.3 Micromechanism of ductile failure 68

Void growth is generally said to be a strain-controlled phenomenon, as opposed to cleavage which is stress controlled. Ductile fracture occurs, when the plastic strain reaches the critical plastic strain, at the critical distance, r = rc . In this case the f critical distance is taken to be the spacing between the voids (or void initiating particles), see Fig. 3.3. 3.4 Prediction of fracture toughness using the MVC model and the HRR eld Assuming that the crack tip strain is represented by the HRR eld and that failure occurs in the plane of the crack ( = 0) the failure equation is, f = 0 JIC 0 0 In l0
n/n+1

(0; n).

(Note that there is a typographical error in the Ritchie and Thompson paper.) The quantity is the equivalent (Von Mises) plastic strain as given by the HRR distribution. For large n, n/(n + 1) 1 so we can write the above equation as, JIC = Writing in terms of KIC we get, KIC = In
0 E In l0 . f

f (0 In l0 ).

This equation predicts that increasing temperature will decrease the ductile fracture toughness (since 0 decreases with increasing temperature). 3.4.1 Denition of critical plastic strain The question arises as to what value of strain to use in the equation above, i.e. will it simply be the failure strain measured in a tensile test? Considerable research on micro-mechanics models for void nucleation and growth has been carried out, see e.g. the textbook by Webster and Ainsworth. It has been shown that the rate of void growth rate is proportional to the hydrostatic stress and the plastic strain rate, i.e. void growth actually depends on stress as well as strain. A typical void growth model is that of Rice and Tracey who found that if the material containing the void is assumed to be elastic-perfectly plastic with yield strength, y , then the void growth rate is given by
r = 0.558r sinh (1.5m /y ) , p

69

where r is the void radius, r is the rate of increase of the radius, m is the remote mean

(hydrostatic) stress and is the remote plastic strain rate. (The notation is used p here to indicate that the stress and strain elds are those remote from the void, but when used to link with fracture mechanics analyses, these will be the crack tip stress and strain, i.e. the size scale of the void is understood to be considerably smaller than the size scale of the crack tip singularity) Since the void growth rate depends on the hydrostatic stress, and failure is associated with the growth and coalescence of voids, this means that the value of the plastic strain, , at failure will depend on m /y . Therefore cannot be determined directly p f from a tensile test since the stress state (m /y ) is dierent. The ratio m /y or m /e , where e is the equivalent stress, is generally known as the stress triaxiality. A direct equation for the failure strain is obtained by integrating Rice and Traceys model and assuming that failure occurs when the void reaches a certain size (e.g. large enough to link with neighbouring voids (see Fig. 3.4)).
(a) Dp

dp initial crack (b)

Figure 3.4 (a) Initial conguration of voids in a ductile material (not to scale) (b) Void coalesence leading to ductile crack growth. Given the Rice and Tracey equation,
r = 0.558r sinh (1.5m /y ) , p

we can integrate from the initial void size (or void initiating particle), Dp , to the nal void size,which is given by the initial spacing, dp (see Fig. 3.4). Assuming that triaxiality, m /y , is constant during void growth, we can integrate the above equation, dr = 0.558 sinh (1.5m /y )dp r dr = 0.558 sinh (1.5m /y ) , f r 70

dp /2

Dp /2

where is the failure strain corresponding to the particular triaxiality. f Carrying out the integration, we get = f ln (dp /Dp ) . 0.558 sinh (1.5m /y )

Generally, to allow for strain hardening, y in the above equation is replaced by the equivalent Mises stress e . An alternative relationship which is only accurate for high triaxiality conditions (m /e > 0.8) is = f ln (dp /Dp ) . 0.283 exp (1.5m /e )

We can eliminate the term ln (dp /Dp ) by considering the case of a uniaxial test. For this geometry, e = , where is the applied stress and m = 3 , so m /e = 1/3. If failure occurs in a uniaxial tension test when strain = f , then inserting this into the above equation and eliminating the ln term, we get = f 0.521f . sinh (1.5m /e )

(Note that it is not correct to use the high triaxiality equation to obtain a similar relationship to the one above, though it is used in a number of text books.) The above relationship gives an expression for the critical strain, in terms of the f uniaxial failure strain, f which can be easily measured. 3.4.2 Use of notched specimens to study triaxiality eects Notched specimens can also be used to examine experimentallythe eect of stress state (triaxiality) on failure strain , see Fig.3.5

amin

Notch radius,

Figure 3.5 Notched specimen used to determine eect of stress state on failure strain. 71

By varying amin and the triaxiality, m /, in the notch region, is varied. The Bridgeman equation gives an approximate solution for the triaxiality in the notch as m 1 amin . = + ln 1 + e 3 2 If we then measure the strain in the notch at failure we can generate a plot of failure strain, , versus triaxiality as shown schematically in Fig. 3.6. f

3.0
sharp crack

uniaxial

0.3 0.2

* f

1.2

Figure 3.6 Eect of triaxiality on failure strain as determined from notched bar tests The critical strain for a crack can be determined by extrapolation. Note that the dependence of failure strain on triaxiality will depend on the material. For some materials the Rice-Tracey expression works well but for others it may not work so well. 3.5 Competition between brittle and ductile fracture Based on our two relations for cleavage and ductile failure, we can now construct two failure curves as shown in Fig. 3.7. In general these mechanisms will be in competition and fracture occurs by the mechanism that is satised rst. At low temperatures it is seen that the cleavage criterion will be satised before the ductile tearing condition and at high temperatures the ductile mechanism is activated rst. We can dene a transition temperature at which the failure mechanism changes from a cleavage to a ductile mechanism. In reality this transition does not occur at a single temperature there is a gradual change from predominantly cleavage failure to predominantly ductile as the temperature is increased. The high toughness regime where failure is by ductile tearing is often called the upper shelf regime. The mode of failure can often be identied by examination of the broken surfaces of the fractured specimens. The fracture surface of a material which has failed in a 72

ductile fashion tends to be rough and dimpled indicative of the void growth mechanism while the fracture surface of a material which has failed by cleavage tends to be at and shiny (see Fig. 3.8).

K IC =
&WEVKNG

I n * * ~ 0 E f l0

KIC
$TKVVNG 6TCPUKVK

6GORGTCVWTG

K IC =

1 A
( n +1 )/ 2

( )
* f

(1+ n )/ 2

0(1n )/2 l0*

Figure 3.7 Competition between brittle and ductile failure.

(a)

Figure 3.8 Typical fracture surfaces in metals (a) cleavage (b) microvoid coalescence. Further discussion on this topic is provided in the paper by Ritchie and Thompson in Appendix A. 73

VGORGTCVWTG

(b)

4. Application of BS 7910 in failure assessments BS7910 is the current British standard which is used in the assessment of awed structures. Its full title is Guide on methods for assessing the acceptability of aws in metallic structures. The purpose of the standard is to provide a simple, repeatable procedure for assessing the safety of cracks. It also makes contact with other British and International Standards (e.g. ISO, ASTM) and is closely related to R6, the code used in the UK nuclear industry to assess the safety of structures with defects. BS 7910 emphasises the need for NDT to detect cracks and also provides guidance on safety factors, reliability factors and probabilistic methods. The rst procedures were written as a published document of the British Standards Institute in 1980 (PD 6493) but these have now been revised and updated as a British Standard, BS 7910, which was released in 2001. Three levels of treatment of aws are provided. Level 1 is a conservative preliminary procedure which is very easy to apply; Level 2 is the normal procedure and is more complicated than Level 1. It contains two sub-options, Level 2A and Level 2B. Level 3 is the most advanced treatment and contains three sub-options, Level 3A, 3B and 3C. Level 3 is mainly used for ductile materials which exhibit some stable amount of crack growth before fracture. All levels use the concept of a failure assessment diagram (FAD) similar to the idea of a yield surface in plasticity. If an assessment point lies within the diagram, the aw or crack is deemed to be safe. If it is outside the diagram it is deemed unsafe and action must be taken. Note that a crack which fails by the conservative Level 1 option may be safe when the more accurate Level 2 analysis is carried out. Similarly, a crack which is unsafe by Level 2 may be safe under Level 3, if a small amount of stable ductile crack growth is allowed for. 4.1 The failure assessment diagram Before discussing the procedure in detail a number of issues relevant to each level are discussed. The philosophy behind the procedure is that failure can occur either due to excessive plastic deformation (plastic collapse) or by fracture. It is now well known that these failure modes are not decoupled as fracture and deformation are closely linked, but the philosophy is maintained for historic reasons and for simplicity of presentation of the method. The fracture parameters used are K, (CTOD) and, for Level 3C, J. Although K is used in the Level 2 and Level 3A and 3B assessments, they are in fact elastic-plastic based assessment procedures. 74

The proximity to fracture and plastic collapse are specied by the ratios, Kr (or r ) and Lr (Sr for Level 1) respectively, where, Kr = K KIC r = c ref P Lr = = y PL ref P Sr = . = f PL (f )

In the above, ref is the reference stress dened in section 2.5.2. PL is the plastic collapse load and f , used in the denition of Sr , is the ow stressdened in BS 7910 as the lower of 1.2y or (y + u )/2, where u is the tensile strength of the material. When dening Sr using the limit load, the ow stress, f , rather than the yield stress is used. Hence the notation PL (f ) above. The use of the ow stress takes account of the fact that the material strain hardens so that plastic collapse does not occur when the stress reaches yield. The use of the square root sign in the quantity expression, i.e., and c Hence JIC KIC = . y Ey K = , c KIC J = y K , Ey r is for consistency with the K

though this relationship does not hold precisely at high values of ref as discussed later. In order to determine the linear elastic stress intensity factor, K, the stress in the uncracked body and the appropriate shape factor, Y must be known. If the body is of complicated shape then a nite element analysis may be required to determine the stresses. Using these stresses, K may be obtained from handbook solutions, many of which are provided in BS7910. It often proves convenient to linearise the stresses, since K solutions are typically available only for cracks under tension (constant stress) or bending (linear variation of 75

stress), but not for arbitrary non-linear stress distributions. BS7910 provides guidance on linearising the stress elds as shown in Fig. 4.1. It is these linear stress distributions which are used to obtain the K value.

Figure 4.1 Linearisation of stress distributions for (a) surface aws (b) embedded aws The procedures assume that the crack is normal to the maximum principal applied stress (i.e. it is a Mode I crack), and most of the available K and limit load solutions are for Mode I loading. However, if, for example, the crack follows a weld boundary (see below) then it will not be a Mode I crack. In this situation, the crack is projected on the plane of principal stresses (see Fig. 4.2) which, in combination with other assumptions in the procedure, is designed to give a conservative assessment.

Figure 4.2 Treatment of an inclined crack within BS7910 However, if the angle between the crack plane and the principal plane is greater than 76

20 then mode mixity eects may become important and they must be accounted for. Advice for this situation is provided in the standard. Each level of the failure assessment procedure is now discussed in turn. 4.2 Level 1 Failure Assessment Diagram This is a very simple procedure. The failure assessment diagram (FAD) is shown in Fig. 4.3. It species that the applied load must be less than 80% of the plastic collapse load, based on the ow stress and K must be less than KIC / 2. The latter inequality approximately corresponds to a factor of safety of two on crack length assuming all linear elastic behaviourthe value of K in Kr is determined from available linear elastic K solutions.

Figure 4.3 The BS7910 Level 1 Failure Assessment Diagram If the CTOD approach is to be used, then is determined from K by = K2 F (/y ), y E

where F (/y ) is an adjustment factor which accounts for plasticity eects on the CTOD. It is given by the following expression, 1 F = y
2

for 0.5y 0.25 y for > 0.5y .

The largest value of F will be when /y = 0.8 (since Sr is always less than 0.8 for a Level 1 assessment) For this case F = 0.86, so in general the correction factor 77

will be small. The above expression for F is for steel and aluminium alloys. For other materials, F is always taken to be 1. Note that, in the above, dierent results would be obtained from a K-based fracture assessment and a CTOD-based analysis if /y > 0.5. This is a consequence of the fact that originally these were separate procedures used in dierent applications and in bringing the two together in a single standard, some inconsistencies and compromises are inevitable. 4.3 Level 2 Failure Assessment Diagram This is the normal assessment procedure for general analysis which is further subdivided into Level 2A and Level 2B. It is an elastic-plastic J-based approach so some background into the derivation of the FAD is provided next. Note that this background knowledge is not required to apply the method and application of Level 2A is no more complicated than that of Level 1. 4.3.1 Derivation of Level 2 failure assessment diagram Take as a starting point, the EPRI power law plasticity solution for J, J = y y a(P/PL )n+1 h(a/w; n), where PL is the limit load, and the stress-strain relationship of the material is given by /
y

= (/y )n .

Recalling the denition of the reference stress, ref = y (P/PL ), and dening the reference strain as the uniaxial strain corresponding to the reference stress, i.e. for the power low hardening material,
ref

= y (ref /y )n ,

we can rearrange the J equation above to get J =a


ref ref h(a/w; n).

The advantage of the above equation is that it can be applied to any material, provided the uniaxial stress strain behaviour is known, and although it is strictly correct only for power law hardening materials, it has been shown that it provides a conservative estimate of J for a wide range of materials. 78

It may be seen that the power-law parameter n still enters the equation for J through the dependence of h on n. However, it is been found for most engineering materials of interest and provided that the limit load is used as the normalising load for J estimation, that h is weakly dependent on n (recall Fig. 2.17) e.g. h(a/W ; n) h(a/W ; 1). The next stage is to rewrite the above equation in the form of a FAD. The condition for a safe assessment is simply J Jc , where Jc is the valid plane strain fracture toughness. In other words, J(ref ) Jc = or inverting E 1 . 2 Kc J
2 Kc , E

Multiplying across by Je = K 2 /E we get failure when K Kc Kr , Kr = K = Kc J(ref ) Je


1/2 2

Je , J

where K is simply the linear elastic stress intensity factor. We can rewrite in terms of

In other words, the curve for the FAD can be determined directly from the relationship between the total J and Je . This provides us with a denition for the FAD in terms of ref . Dening Je using the reference stress equation, we get
2 ref h(a/W ; 1), E since for a linear elastic material, = 1 and ref = ref /E. Then we have

Je = a

J E ref = , Je ref making the assumption that h(a/W ; n) = h(a/W ; 1), and then Kr = J Je
1/2

= 79

E ref ref

1/2

().

Since Lr = ref /y and

ref

can be obtained in terms of ref via the uniaxial

stress strain law, this allows a material dependent but geometry independent FAD to be plotted in terms of Kr and Lr . The above equation is correct for linear elastic materials (trivial since the result will be J/Je = 1) and gives a good representation under large scale plasticity. However, in the intermediate region when there is no remote plasticity, ref < y , but local yielding occurs at the crack tip, the agreement with the exact J solution is poor. Therefore an additional estimate for J is required in this region. For this we used the plastic zone correction approach. 4.3.2 Estimate of K and J using a plastic zone correction As discussed in Section 2, when the plastic zone size is relatively small modications to the LEFM K are sucient to account for the eects of material non-linearity. An eective K is dened, Ke = Y (ae ) ae where Y is the LEFM geometry factor. The eective crack length ae is given by, ae = a + ry , where a is the actual crack length and ry is half the plane strain or plane stress plastic zone size as dened in Section 2.5.7.1. For a small crack under tension it was shown that, a Ke = . 2 1 1 2 y Replacing by ref so the expression is applicable to general geometries and writing in the form required for an FAD, we get J = Je Ke K
2

1 1 1 2 ref y

2.

and using the binomial theorem for small ref /y we get J 1 =1+ Je 2 80 ref y
2

(#).

2 Note that we have used the equation J = Ke /E to determine J from K . This

expression is valid as we are still within small scale yielding conditions. 4.3.3 Overall Level 2 Failure Assessment Diagram Finally, combining the two expressions for J/Je , Eq. (#) above and Eq. (*) from page 81 into a single expression which reduces to the above when ref < y and to the earlier expression for large scale plasticity, the following unied expression is obtained: E ref 1 J = + Je ref 2 The term E For ref < y ,
ref /ref

ref y

E ref ref

appears in both of the terms on the RHS of the equation.


ref /ref

ref

= ref /E so E

= 1 and the small scale yielding equation is


ref /ref

obtained. Under large scale plasticity it may be seen that E

pl / el

which

becomes very large under large scale plasticity and hence the second term becomes negligible and the equation reduces to the rst of the two J expressions. Writing the above equation in terms of Lr , the FAD becomes, Kr = J Je
1/2

E ref 1 L3 y r + Lr y 2 E ref

1/2

This, nally, is the form of the FAD for a Level 2B assessment. To use a CTOD approach Kr is replaced by r as in Level 1 and the CTOD, , is obtained from K again as in Level 1 with some modication to account for plane stress/plane strain conditions. The details of the CTOD approach are given in the standard. The above FAD will depend on the material in question, i.e. a dierent FAD will be needed for each material that is being assessed. Figure 4.4 shows a typical Level 2B FAD for a particular material. 4.3.4 The Level 2A Failure Assessment Diagram Also shown in Fig 4.4 is the Level 2A FAD which is a lower bound (i.e. conservative) Level 2 FAD for a wide range of materials. If uniaxial stress-strain data is unavailable this FAD may be used. The equation for the Level 2A FAD is r or Kr = (1 0.14L2 ) 0.3 + 0.7 exp (0.65L6 ) r r The Level 2A FAD is a geometry and material independent curve which makes it very simple to use. 81

Note that neither Level 2A nor Level 2B are exact solutions but Level 2B is closer to reality than Level 2A. For a wide range of cases, it has been shown by comparing with numerical solutions and experiments that both curves are conservative and Level 2A is more conservative than Level 2B. 4.3.5 Denition of cuto line In Fig. 4.4 it may also be seen that there is a cuto at a point dened as Lrmax . This cut-o which depends on the material is to prevent global plastic collapse. The cuto is dened as Lrmax = f , y

where f is the ow stress. If Lr > Lrmax then the crack is unsafe, regardless of the value of Kr .

Figure 4.4 BS7910 Level 2 Failure Assessment Diagrams (a) Level 2A FAD, (b) Level 2B FAD derived from material stress/strain data of (c). 4.3.6 Use of the Level 2 procedure The use of the procedure is exactly as for Level 1, i.e. Kr and Lr are determined from handbook solutions or numerical calculations (elastic calculations for Kr , perfectly 82

plastic for Lr ) and the point located on the FAD. If the point lies inside the FAD then the crack is safe. 4.4 Level 3 procedure The Level 3 procedure is divided into three methods. Level 3A and 3B are essentially the same as Level 2A and 2B respectively, apart from the fact that the resistance curve is used to determine the material parameter Kc . Therefore, even if the initial point lies outside the FAD, provided there is a sucient increase in the J-resistance curve, after a small amount of ductile tearing, the crack may be safe. Level 3C involves the use of a further, more accurate FAD, determined from a full elastic-plastic J analysis of the structure. The FAD is then determined directly from Kr = J Je
1/2

and the assessment is carried out as for Level 3A and Level 3B. There is no reason why the Level 3C FAD could not be used for a Level 2 type analysis, i.e. obtaining the FAD directly from J and using the initiation KIC value, rather than the enhanced toughness after some amount of crack growth. However, it is assumed that if one is going to the expense of a full numerical analysis, one will also want to take full advantage of the material toughness.

83

5. Creep Fracture Mechanics Creep occurs when a component is held under stress over long times or high temperatures or a combination of both (see 2M Materials notes). Creep is a time dependent process that results in permanent (non-recoverable) deformation and may ultimately lead to failure (creep rupture). Creep is particularly important in chemical process plant, electrical power generation equipment and aircraft gas turbine engines and often it is failure due to creep that is the predominant design failure mode. In metals, creep is generally divided into primary, secondary and tertiary creep. Primary creep is the creep which occurs over short times and results in a decreasing strain rate. Secondary creep generally occurs over the largest period of a components lifetime and is characterised by a constant (steady state) creep rate. Tertiary creep occurs at long times, close to the time of failure and gives a very high creep rate. These modes are illustrated by the typical creep curve shown in Fig. 5.1.

Figure 5.1, Typical strain vs time creep curve 5.1 Secondary creep Since a component will spend most of its lifetime in the secondary creep regime, creep fracture mechanics has focused on this regime. During secondary creep, cracks which were initially safe, may grow slowly, under a constant load, and lead to failure, a process analogous to crack growth by fatigue under cyclic loading. For many materials, the secondary creep deformation behaviour is well characterised by a power law creep relationship, analogous to power law plasticity, c = 0 0 84
n

where n, 0 and 0 are material constants. This equation is identical to the power law plasticity relationship but with 0 replacing
0

and = 1. It may therefore be shown

that solutions to power law plasticity are also solutions to power law creep except that displacement rate and strain rate replace displacement and strain, respectively, in the creep solution. Therefore, we immediately have the solution to the steady state creep stress and strain rate at the crack tip for a power law creep material, i.e. ij /0 = C 0 0 In r C 0 0 In r
1/(n+1)

ij (; n)
n/(n+1)

c / 0 = ij

ij (; n).

The parameter C is the creep analogy to J and is dened in the same way as J using a contour integral, C = where W () =
0

u W ()dy t ds, x

is the strain energy rate density. This is a path independent integral (the proof follows immediately from the path independence of the J integral). However, the path independence relies on the fact that steady state conditions are prevailing, i.e. the creep strain rates are much larger than the elastic strain rate in the body. This will hold after long times and when the remote applied load is constant. 5.2 Estimation of C in specimens and components The same procedures developed to estimate J can be used to estimate C under steady state conditions. For example C can be estimated from load-displacement rate data, C = A, B(W a)

where A is the area under the load displacement-rate curve. Note that analogous to pure power law plasticity, A= n P , n+1

where is the remote displacement rate corresponding to the load P , allowing the above equation to be simplied to 85

C =

n P . B(W a) n + 1

Values of for various crack geometries are available in the fracture mechanics handbooks or the ASTM standardthey are identical to the elastic-plastic factors. In the same way, the GE-EPRI scheme developed to estimate J can be used to estimate C , C = a 0 0 P P0
n+1

h(a/w; n).

Note that in creep, P0 no longer has the interpretation of a limit load. It is simply a normalising load, based on the material parameter 0 . More commonly in the UK, the reference stress approach, which is an extension of the GE-EPRI approach is used to estimate C because of its simplicity. Recall from Section 4.3.1 that under pure power law plasticity,J may be estimated using the following equation, E ref J = , Je ref with ref dened via the plastic collapse load. Similarly for power law creep, we can write, C E ref = . Je ref It is convenient in the creep case to replace Je by K and the equation becomes C = and we then write C = ref ref R , where R is a length scale which depends only on the geometry of the component, R = (K/ref )2 , e.g. for a crack in an innite plate R = a. Once again it needs to be emphasised that this equation gives an estimate of C and to obtain an exact value a study of the actual cracked geometry in question is needed. 5.3 Creep solutions for short times The above solutions are for long times when the stress and strain rate elds are constant throughout the structure. The period between initial loading and nal steady 86 K 2 ref , ref

state is called the redistribution period. Figure 5.2 shows how stresses redistribute during creep from the initial elastic K eld at t = 0 to the steady state value at long times.

/
Figure 5.2 Redistribution of stresses during creep In the analysis shown in Fig. 5.2 the power law creep exponent, n was equal to 3. Therefore the steady state creep distribution is quite close to the elastic distribution (n = 1). The higher the value of n the lower the crack tip elds so for n = 10 more redistribution of stress will take place. During the redistribution period, the global elastic and creep strain rates are comparable. However, because of the power law nature of creep and the large stress gradients near the crack tip, there will still be a region in the vicinity of the crack where the creep strain rates are much higher than the elastic strain rates and the HRR-type solution discussed above applies. We write, ij /0 = c / 0 ij = C(t) 0 0 In r C(t) 0 0 In r
1/(n+1)

ij (; n)
n/(n+1)

ij (; n),

where the notation C(t) is used to emphasise that the amplitude of the stress depends on time. The zone of dominance of the HRR-type eld may be very smallit approaches zero at very short times (when the K eld dominates). The equation for C(t) is identical to that for C the only dierence being that C(t) is no longer a path independent integral and can only be dened asymptotically, i.e. the contour very close to the crack tip, 87

u W ()dy t ds. x 0 Figure 5.3 shows the variation of C(t) with time determined from a numerical C(t) = analysis. The decreasing magnitude indicates that the amplitude of the crack tip stress eld reduces with time due to creep redistribution (see also Fig. 5.2),

Figure 5.3 Variation of C(t) with time, (Webster and Ainsworth). After some time the value of C(t) approaches the steady state C value. The normalising time, tT , in Fig. 5.3 is an approximation to the time required to reach steady state and is called the transition time (there is also a quantity called the redistribution time which provides a slightly improved estimate of time to reach steady state). The equation for tT is K2 . (n + 1)E C For the case illustrated in Fig. 5.3 it may be seen from the numerical result, that tT = the time to reach steady state tss 3tT . The approximate solution to C(t) shown in the gure is given by the equation, C(t) = and it may be seen that C(tT ) = C . Note also that the above equation gives the physically unrealistic solution that C(t) = at t = 0. There are other more complicated equations to estimate C(t), but creep fracture mechanics is still primarily based on C . 88 tT t C = K2 , (n + 1)E t

Note that the stress and strain rate elds given above are strictly speaking applicable only for stationary cracks and under creep conditions the crack will be growing. However, because the crack is growing very slowly, there is sucient time for the stationary crack distributions to re-establish themselves after each increment of crack growth. It should also be noted that even under constant load the value of C will be increasing as the crack length will be increasing. So for example, in the reference stress estimate for C both ref and R will increase (slowly) with time. 5.4 Characterisation of creep crack initiation and growth Under a constant load and at high temperature, a pre-existing crack will grow slowly due to creep until nal failure. Generally, failure due to creep is by a stable ductile process, involving growth and coalescence of micro-voids. Since under steady state conditions C characterises the stress and strain rate at the crack tip, it is to be expected that C will also provide a good measure of the crack growth rate under creep conditions. A typical curve is shown in Fig. 5.4 where the rate of crack growth, a, is plotted against C for two specimen geometries of an alloy steel. It may be seen, that within the scatter of the data, the creep crack growth rate for the steel is characterised by C .

Figure 5.4 Crack growth rates, a, plotted against C for two specimen types Note that a single test can generate a full set of a vs. C data as the crack is growing during the test. Typically, in such a test, C is estimated using the load-line displacement rate and the rate of increase of crack length a is estimated using visual inspection, compliance methods or potential drop techniques, whereby a constant cur89

rent is applied across the crack plane and the potential drop is correlated with the increase in crack length. In the latter two methods, comparison with the nal crack length determined from heat tinting and breaking open of the specimen should be used to check the predicted amounts of crack growth. 5.4.1 Model for steady state creep crack growth We consider a crack which is assumed to be growing at a rate, a with a steady state creep eld characterised by C . It is assumed that the process of crack growth is a strain controlled process with material failure occurring at a point initially a distance rc ahead of the crack tip, when a strain
f

is reached (see Fig. 5.5).

Figure 5.5 Distribution of creep strain rate ahead of a crack in a creeping material Given the expression for the creep strain rate at a distance r from the crack tip, c = 0 ij C 0 0 In r
n/n+1)

ij (; n),

the creep strain accumulated over a time t is given by


c c t

=
0

c dt = 0 (0; n)

C 0 0 In

n/n+1 0

1 r(t)n/n+1

dt,

where

is the equivalent (von Mises) creep strain, and the dependence of distance, r,

on time, due to the movement of the crack, is emphasised. It is assumed in the above equation that creep crack growth occurs directly ahead of the crack tip ( = 0) but it is not dicult to generalise the equation to allow creep crack growth at any angle. Using a change of variables in the above,
c

= 0 (0; n)

C 0 0 In

n/(n+1)

0 rc

1 rn/n+1

dt dr. dr

Since

dr 1 dt = 1/ = , dr dt a 90

we can write
c

0 (0; n) = a = (n + 1)

C 0 0 In

n/(n+1)

rc 0 n/(n+1)

1 rn/n+1

dr

0 (0; n) a

C 0 0 In

1/(n+1) rc . f.

Crack growth occurs when this strain is equal to the material ductility, the above equation as an equation for crack growth rate, a, a = (n + 1) 0 (0; n)
f

Note that as

discussed previously, the ductility depends on the triaxiality, (m /e ). We may rewrite

C 0 0 In

n/(n+1) 1/(n+1) rc .

The distance rc is sometimes identied as the creep process zone size, i.e. the distance ahead of the current crack tip where creep damage is signicant. Note that because rc is raised to the power 1/(n + 1), the dependence of a on the value of rc is weak, so its value is not very signicant. However, the crack growth rate depends inversely on the ductility,
f.

The triaxiality, m /e , under plane strain conditions is


f

about a factor of three higher than plane stress, leading to a creep failure strain

about 30 times lower than in plane stress, using a void growth criterion, as discussed earlier. Thus crack growth rates are expected to be about 30 times higher under full plane strain conditions than under plane stress conditions at the same value of C . The important point is that there is a one-to-one relationship between a and C and that crack growth rate is given by an equation of the form, a = A(C ) . The above form has been conrmed by experiment and leads to a straight line on a log-log plot. A general equation has been proposed by Nikbin, Smith and Webster to cover a wide range of n values. They have shown that the theoretical equation derived earlier can be simplied by follows, a= 3(C )0.85
f

This equation has been shown to predict crack growth rates to within a factor of two for a wide range of materials. In the above equation,
f

is the appropriate ductility,

i.e. equal to the uniaxial creep ductility under plane stress (i.e. thin components) and equal to 1/30 times the uniaxial ductility under plane strain (thick components). A 91

comparison of these two crack growth rate equations with the earlier experimental data is shown in Fig. 5.6. The increased creep rates predicted by the plane strain solution may be seen. It appears that the data for the tests in Fig. 5.6 correspond more closely to plane stress conditions.

Figure 5.6 Comparison of theoretical creep crack growth equations with experimental data 5.4.2 Creep Initiation The equations derived earlier have given the rate of crack growth under secondary creep. However, recently there has been increased interest in predicting the incubation period, i.e. the amount of time before a crack starts to extend under creep conditions. This can be determined in a very similar manner to the creep crack growth rate. We assume again, that the crack tip elds are given by C and the HRR eld. Crack initiation occurs when the strain at some critical distance from the crack tip reaches the material ductility. This critical distance, may be a material distance like a grain size or simply the resolution of the crack growth detecting device. As before
c t

=
0

c dt = t c .

Note that the above assumes that even during incubation, most of the strain accumulated is during the steady state period, i.e. C(t) = C and therefore the strain rate is constant during incubation (there is no theoretical diculty in allowing C(t) to vary 92

with time and include this in the integration.) The analysis for initiation diers from the crack growth analysis in that in this case the crack tip is assumed stationary. Using the above we get,
c

= t 0 (0; n)

C 0 0 In rc

n/n+1

and ti =
f

0 0 In rc C

n/n+1

and again the dependence of incubation time on C is illustrated. The quantities in the above equation are not always easily obtainable. A good estimate of the incubation time has been shown to be given by the following, reference stress based, expression: ref tr ti = 0.0025 K2
0.85

where tr is the time to rupture in a uniaxial test carried out at = ref . In the above equation, units of stress are in MPa, time in hours and K in MPa m. 5.5 Elastic-plastic creep In the preceding analyses, it has been assumed that the plastic strains are negligible. (The distinction made here between rate independent plasticity and rate dependent creep is somewhat questionable, but it suces to explain most high temperature fracture phenomena.) Incorporating the eect of plasticity is not dicult: the initial stress eld is then given by the elastic-plastic HRR eld, rather than the K-eld and, as for the elastic-creep case, the stress elds will redistribute until steady state conditions are reached. Usually the redistribution times are shorter, as the elastic-plastic stress distribution are closer to the creep distributions than the elastic ones. The equations derived previously still hold with the requirement that for steady state, c > e + p , where e and p are elastic and plastic strain rates, respectively. In the derivation of creep crack growth laws, a vs C , and incubation times it is generally assumed that damage due to plastic strain is of a dierent type than that due to creep strain and hence the contribution from the plastic strain is not included in the total strain required to cause crack initiation or growth. 93

5.6 Micrographs of creep failure As discussed previously, creep failure occurs primarily in a ductile manner. Under creep conditions voids typically nucleate on grain boundaries or triple grain junctions as shown in Fig. 5.7(a). Final failure is then associated with linking up of microcracks or voids along grain boundariesintergranular fracture as shown in Fig. 5.7(b).

Figure 5.7 Damage observed in materials loaded under creep conditions

94

6. Appendices 6.1 Appendix A, Extracts from two key papers on non-linear fracture mechanics Fundamentals of the phenomenological theory of nonlinear fracture mechanics by J.W. Hutchinson, Journal of Applied Mechanics, Vol. 50, 1983. On macroscopic and microscopic analyses for crack initiation and crack growth toughness in ductile alloys by R.O. Ritchie and A.W. Thompson, Metallurgical Transactions A, Vol. 16A, 1985.

95

These pages are not made available on the college web page

96

6.2 Appendix B, List of important equations for Advanced Fracture Mechanics (Appendix B and C may be used freely in exams)
Strain energy density, W, for a linear elastic material under uniaxial stress, : W=

2
2E n n+1

Strain energy density, W, for a power law material under uniaxial stress, : W=

Definition of strain energy Ue: U e = WdV


V

Strain energy for a linear elastic material under an applied force, P, giving rise to a displacement, : Ue = P 2

Strain energy for a linear elastic material under an applied moment, M, giving rise to a curvature, : Ue = M 2

Strain energy for a power law material under an applied force, P, giving rise to a displacement, : Ue = n n +1 P

External work, Wext, for a moment, M, giving rise to a curvature, : Wext = M. External work, Wext, for a force, P, giving rise to a displacement, : Wext = P. Definition of energy release rate G, G= 1 U B a dC da

Energy release rate-compliance relation for a linear elastic material: G= 1 2B P2

Definition of K for a cracked geometry: K = Y a ; Center crack in an infinite plate under tension, Y = ; Edge crack in an infinite plate under tension, Y = 1.12. Page 1 of 4

97

Crack opening displacement for a Mode I crack in a linear elastic material:

u ( r ) = 8

E 2

; E = E for plane stress; E =

E 1 2

for plane strain.

Relationship between energy release rate and stress intensity factor:

G=

2 2 K I2 + K II K III + ; G = shear modulus 2G E

Phase angle, = tan -1

K II KI

; Mode mixity, M =

Branching angle, , for a crack under mixed mode conditions in a linear elastic material, using the maximum hoop stress criterion:

Plastic zone size, rp, ahead of a Mode I crack:


2

Plastic zone correction for K: K = Y (a e ) a e ; a e = a + rp / 2 ASTM size requirement for KIC testing: a, b, B > 2.5(KIC/y)2. Definition of J line integral:

J = Wdy t

HRR crack tip stress and strain distribution for a power law hardening material:
J
      

Relationship between crack opening displacement, , and J integral:

= dn

; d n 1 for plane stress; d n 0.5 for plane strain.



ij J = 0 0 0 I n r
 

rp =

; =

for plane stress; =

u x

ds

1 /( n +1)

~ ij ( ; n ); ij = 0

K II

1 KI tan ( 2 ) = 4 K II

KI

+8

1 for plane strain 3

rp

y x

n /( n +1)

0 0 I n r

~ ij ( ; n )

Page 2 of 4

98

Plane stress limit moment for a shallow cracked beam in bending:

ML = y

B (W a )2 4

Plane stress limit load for a shallow cracked beam in tension:


PL = y B (W a )

Evaluation of J using the factor:

J=

K2 E

B (W a )

A p ; 1 for tension loading; 2 for bend loading.

Evaluation of J using GE-EPRI equation:

P0

ASTM size requirement for JIC testing (for deep cracked bend specimen)

a, b, B > 25(JIC/y)
Rice and Tracey void growth relation:

r = 0.558sinh (1.5 m / y ) p r BS7910 Failure Assessment Diagram: Kr =


K P P ; Lr = ref = ; S r = ref = K IC y PL f PL ( f )

PL

Je

BS7910 Level 2A FAD, K r = 1 0.14 L2 0.3 + 0.7 exp 0.65 L6 r r BS7910 Level 2B FAD, K r =
 

Page 3 of 4

99

Lr y

2 E ref

E ref


3 1 Lr y

Failure condition: K r =

J


ref =

P


y ; f = 1.2 y or ( y + u ) / 2
1 / 2

J=

K2

+ 0 0 a

n +1

h (a / W ; n )

)(

))

1 / 2

Lr max =

f y

Definition of C* line integral:  u  C* = Wdy t ds x Evaluation of C* for a power law material using the factor: C* =

B (W a ) n + 1

 P

Evaluation of C* using GE-EPRI equation: P  C* = a 0 0 P 0


n +1

h (a / W ; n )

Equation for C* based on ref: K  C* = ref ref R ; R = ref


2

Creep crack growth rate equation:  a = A(C * )

Page 4 of 4

100

6.3 Appendix C, Linear Elastic K eld distributions (Appendix B and C may be used freely in exams)

101

102

103

S-ar putea să vă placă și