Vous êtes sur la page 1sur 37

Received: 17 April 2018 Revised: 28 July 2018 Accepted: 1 August 2018

DOI: 10.1111/ffe.12912

INVITED REVIEW ARTICLE

Computational fracture mechanics: An overview from early


efforts to recent achievements

Aleksandar Sedmak

Faculty of Mechanical Engineering,


Abstract
University of Belgrade, Belgrade, Serbia
An overview of the state of the art of computational fracture mechanics is pre-
Correspondence sented, starting from early efforts and going all the way up to recent achieve-
Aleksandar Sedmak, Faculty of
Mechanical Engineering, University of
ments. Some specific aspects of linear elastic and elastic plastic fracture
Belgrade, Belgrade, Serbia. mechanics problems have been discussed, including the fact that even for a
Email: asedmak@mas.bg.ac.rs static loading, numerical simulation is not a simple task because of complex
geometries, material nonlinearity, and heterogeneity, and especially if crack
growth is considered. Therefore, micromechanical modelling of elastic‐plastic
crack growth is presented as new and promising approach to overcome some
of the shortages of traditional approach. Besides static loading, couple of other
important practical problems are tackled, like fatigue crack growth, with
remaining life in the focus of investigation, using the empirical laws for crack
growth rates. Numerical simulation of fatigue crack growth is inherently com-
plex both because of complex material damage processes and lack of sound
theoretical basis to define them. Therefore, combination of theoretical, experi-
mental, and numerical approach is presented here to get reliable and efficient
estimation of life under fatigue loading.

Nomenclature: (R/R0)c, Critical value of void growth; 2a, Crack length; 2a0, Initial crack length; A, Secondary void initiation; AA, volume and surface
fraction of detected oxides; Ai, Area of detected inclusions; AT, Measurement field area; a/w, Crack length/specimen width ratio; a, b, Thomason
constants; b, Specimen semiwidth; B, Specimen nominal thickness; Bn, Ligament thickness; C, m, Paris equation coefficients; CMOD, Crack mouth
opening displacement; COD, Crack opening displacement; CTOA, Crack tip opening angle; CTOD, Crack tip opening displacement; D, Chu‐
Needleman model parameter; dεpeq , Equivalent plastic strain increment; E, Young modulus; f , Void volume fraction; F , Load; f *, Void volume
fraction function; f 0, Initial void volume fracture; f c, Critical value at which void coalescence occurs; f F , Void volume fracture at final failure;
F max, Maximum force; f N, Volume fraction of secondary‐void forming particles; f V, Volume fraction of nonmetallic inclusions; h, Specimen
height; Ji, J‐integral at crack initiation; K, Strength coefficient; KI, KII, KIII, Stress intensity factors for crack opening modes I, II, and III
respectively; Keff, Equivalent stress intensity factor; lc, Original element length; n, Hardening exponent; q1, q2, Twergaard parameters; r, Void space
ratio; R, Mean void radius; R0, Mean void radius initial value; Rm, Ultimate tensile strength; Rp0.2, Yield stress; SN, standard deviation; t, Specimen
thickness; U, External work area below F ‐VLL curve; VV, volume and surface fraction of detected sulphides; W‐a0, Ligament length; α, β, Rice‐
Tracey model parameters; δa, Crack length increment; Δa, Crack growth; ΔD, Cross‐sectional reduction (round tensile specimen); εe, Elastic strain;
ε_ pii , Strain rate tensor; εN, Mean strain at void nucleation; επ, Plastic strain; Φ, Plastic potential; λ, Mean free path; σ1, maximum principal stress; σ2,
0
Stress in the y direction; σeq, Equivalent stress; σ ij , Stress deviator; σm, Mean stress; σm/σeq, Stress tri‐axility; σT, True stress; σYS, Yield strength;
CFM, Computational fracture mechanics; LEFM, Linear elastic fracture mechanics; EPFM, Elastic plastic fracture mechanics; GTN, Gurtin‐
Tvergaard‐Needlman; CGM, Compete Gurson model; FEM, Finite element method; EFG, Element free Galerkin; C (T)OD, Crack (tip) opening
displacement
This paper is dedicated to Prof. Stojan Sedmak (1929‐2014), father of Fracture Mechanics in Southeast Europe, founder of Serbian Structural Integrity
and Life Society, named after him in 2014. His gigantic effort made Serbia and ex Yugoslavia important part of the Fracture Mechanics world map.

Fatigue Fract Eng Mater Struct. 2018;1–37. wileyonlinelibrary.com/journal/ffe © 2018 Wiley Publishing Ltd. 1
2 SEDMAK

1 | INTRODUCTION mechanics problems. Anyhow, the real breakthrough


was another pioneering work, by Prof. Mladen Berković
Computational fracture mechanics (CFM) is a broad term in 1980, with application of linear elastic FEM to a simple
encompassing many different numerical methods applied problem of central cracked plate in tension.24 This early
to solve fracture mechanics problems using computers. effort resulted in very accurate prediction, using special
Here, we consider the finite element method (FEM) as elements and uniform mesh refinement at the same time,
the most powerful tool for numerical simulation of frac- with stress intensity factors (SIF) calculated by extrapola-
ture mechanics problems. Therefore, it is only natural to tion or displacement and via J intergral, including 3D
start from the very beginning, with 2 pioneering papers, problems.25 Later on, nonuniform mesh refinement with
1 by Argyris1 and the other by Turner et al.2 More than extrapolation technique also provided excellent
a decade later, when FEM was already a well‐established results.26,27
numerical tool for solving many structural problems,3,4 Application of FEM to static crack growth was natural
difficulties of its application to crack problems emerged. next step, leading to micromechanical material model-
Namely, the first attempts to calculate stress intensity fac- ling, applied as the so‐called local approach to brittle
tors were not very successful because of relatively large and ductile fracture,28-31 although the basic idea can be
error, even with well‐refined meshes.5 Soon, first singular traced back to late 70s.32 Numerous published papers
elements for linear elastic fracture mechanics (LEFM) and several round robin projects lead to publishing of 2
were introduced by Tracey,6 followed by elastic‐plastic procedures of the European Structural Integrity Society
singular elements,7 as described in the most important (ESIS)33,34 and to further improvement of local approach,
paper on CFM in early 70s, written by Rice and Tracey.8 by introducing GTN model35,36 and complete Gurson
Later on, the so‐called 4‐point and 8‐point noded singular model (CGM),37 also supported by the conference focused
elements were introduced, or better to say, reproduced by exclusively on the local approach.38
Barsoum,9 Henshell and Shaw,10 and Shih et al.11 At the All these early efforts look obsolete nowadays com-
same time, another approach to crack problems was to pared to preprocessing and postprocessing capabilities of
use nonuniform mesh refinement,12-14 reducing the error softwares such as ABAQUS and ANSYS, just to mention
to acceptable level without singular elements. Neverthe- 2 of them, probably most frequently used to solve all
less, in both approaches, it was difficult to build a mesh, kinds of different types of fracture mechanics prob-
and also practically impossible to take into account crack lems.39,40 So, one can say that CFM has progressed signif-
growth. icantly over last few decades, or better to say, exploded
Anyhow, it was recognised that nonuniform mesh from small village of handful number of enthusiasts back
refinement and/or singular elements would do the job in 70s of the last century, to a megacity of researchers
for LEFM and elastic‐plastic fracture mechanics (EPFM), producing beautiful colourful images of fracture mechan-
and just few years later, thanks to fast improvement of ics problems, be it brittle, ductile, plastic, fatigue, or
both hardware and software facilities, even uniform creep; static, dynamic or impact; and metallic or nonme-
meshes, with very large number of elements, were tallic. We have witnessed dramatic change in our scien-
equally efficient. Especially, when EPFM problems were tific and engineering practice because of, so far,
tackled, this was true, because “normal” meshes were unstoppable increase in computer capacity and efficiency.
actually needed to follow crack growth. Once computers One can say that “difficult” problems, like those with sin-
and software are mentioned, couple of significant contri- gularities, inspired significant improvement in numerical
butions in software need to be referenced.15-17 analysis, which in turn has been providing solutions for
Many other researchers have contributed to these problems that we even could not dream of just few
early developments,18-20 including elastic‐plastic model- decades ago.
ling of 3‐dimensional (3D) crack problems, as well as pro- As an excellent example of this interaction between
viding solutions for J‐integral and crack tip opening FEM and FM, one can take modelling of crack growth,
displacement (CTOD), before ESIS TC8 committee, including fatigue crack growth, because the right time
devoted to numerical methods in fracture mechanics, for numerical simulation came only after meshless
published the “Recommendations for Use of FEM in method has been introduced by Belytschko,41-46 starting
Fracture Mechanics” in 1991.21 All relevant details about with element free Galerkin (EFG) method, to be com-
elastic‐plastic modelling of crack problems can be found pleted and named as extended finite element method
in recently published book by Brocks.22 (XFEM) by the end of the last century. It should be men-
In ex‐Yugoslavia, it was Prof. Stojan Sedmak, who tioned that the FEM was much exploited in crack growth
first mentioned the FEM (p. 41 in his PhD thesis, pub- modelling, ever since it was introduced into FM practice.
lished in 197623) as an efficient way to solve fracture Anyhow, restrictions in FEM‐based crack growth
SEDMAK 3

simulations discouraged researchers who tried numerous reference numbers are kept): “Three types of finite ele-
variations of FEM, until XFEM was developed for model- ments were used in the analysis. For elastic solutions,
ling discontinuities in 1D, 2D, and 3D domains. The the r−1/2 singular element25 was used nearest the crack
essence of the XFEM lies in subdividing a domain into tip with arbitrary quadrilateral 4‐node isoparametric ele-
2 distinct parts: mesh generation for the geometric ments over the remainder of the configuration. To study
domain (cracks not included) and enriching the finite ele- plastic effects at the crack tip, a new singular element
ment approximation by additional functions that model a was designed, similar to that of Levy et al,37 which has
crack, as introduced by Melenk and Babuska.47 The a 1/r shear strain singularity (with a bounded dilatational
essential feature was multiplication of the enrichment strain) and a uniform strain as admissible deformations.
functions by nodal shape functions, incorporating the The singularity elements have the shape of isosceles tri-
elastic asymptotic crack‐tip function of the Westergaard angles and are focused along radial lines into the crack
field form. Many improvements were made later on, like tip. However, they are treated as degenerate isosceles
the modification of enriched functions in an attempt to trapezoids in the sense that 4 nodes are assigned to the
apply XFEM to elastic/plastic crack growth.48 elements, 1 at each vertex, even though 2 of the nodes
In this paper, an overview of the state of the art of coincide at the crack tip. Levy et al37 introduced this coin-
computational fracture mechanics is presented, including cident node technique to study the crack tip displacement
LEFM, EPFM, micromechanical modelling, and fatigue variation. Contrary to their procedure, however, the coin-
crack growth with life assessment. Focus will often be cident nodes were here constrained to move as a single
on heterogeneous materials such as welded joints, point in obtaining the elastic response of the cracked
because “one should never assume there are no cracks body, because the nonunique crack tip displacement is a
in a welded joint”, as stated by Nichols.49 plasticity effect. The variation of stress and hence consti-
tutive relation in the plastic case within elements was
accounted for in the following approximate manner. Each
2 | N U M E R I C A L SI M U L A TI O N O F near‐tip element was viewed as the composite of 3
LINEAR ELASTIC A ND ELASTIC ‐ subelements, each extending one‐third of the height of
P L A S T I C FR A C T U R E ME C H A N I C S the element. The area average strain of an individual
subelement was used in evaluating the stress state and
Extensive overview of CFM, both in LE and EP, is given constitutive matrix representative of the subelement.
by Kuna50 and Brocks,51 so here it will be presented only The 3 subelement stiffnesses were then formed and added
briefly, starting with few remarks about the FEM, to to obtain the total element stiffness matrix. For the
emphasise often forgotten fact that we are dealing with adjoining isoparametric elements, the midpoint strain
mathematical method of solving partial differential equa- was judged adequate to calculate the stress representative
tion. Typically, static FM problems are defined as the of the entire element. To obtain elastic‐plastic solutions,
elliptical boundary value problem with prescribed con- the procedure was to specify the r−1/2 element just up to
tour conditions for displacements and/or forces. Its main the load necessary to yield 1 of the subelements. There-
features are discretisation (domain), approximation (con- upon the r−1 element was used with its associated non-
vergence), and interpolation (FE itself), transforming par- unique crack tip displacement capability. Clearly, the
tial differential equation into a set of linear algebraic elastic singularity implies yielding under infinitesimal
equations. This is the moment when computers started load so that there is some error involved in the plastic
to be important, at least from the point of view of efficient solution by specifying the r−1/2 near tip strain distribution
solvers of large number of equations,15 while preprocess- up to finite loads. Actually for the size element used at
ing and postprocessing came later to turn this beautiful the tip, this error should be very small”.
mathematical method into a kind of black box. Thus, it Keeping in mind the fact that all necessary equations
is never too many times to repeat how important it is to for both linear elastic and elastic‐plastic elements were
know basics of FEM to apply it correctly and use modern provided, including relevant examples, it is clear that 3
softwares appropriately. Finally, let us remind readers papers, referenced as Tracey,6 Levy et al,7 and Rice and
that from a mathematical point of view, crack in a body Tracey,8 established computational fracture mechanics,
is just a different boundary condition (free surfaces), long time ago, in 1973. Also, let us not forget that just
although it produces relatively large error, even with well 2 years later, McMeeking and Rice52 have shown that tri-
refined meshes,5 because of singularity. angular or prismatic collapsed 8‐node or 20‐node ele-
To start with CFM, let us quote few statements from ments respectively, with a 1/r singularity, are well
the brilliant paper, written by Rice and Tracey in early suited for elastic‐plastic 2D and 3D calculations. Having
70s,6 Chapter 4.2 on ELEMENT DESIGN (note: original in mind also pioneering work in the mechanics of crack
4 SEDMAK

tip deformation and extension by fatigue,53 it seems that * Phase 2: Finite element simulations of ductile crack
we are still just recycling gigantic efforts and phenomenal growth (task A) and cleavage fracture (task B) on CT
results provided by James Rice in late 60s and early 70s. specimens.77
Later on, Barsoum9 formulated (or better to say * Phase 3: Finite element simulations of CT specimens
reinvented) triangular or prismatic isoparametric ele- in the brittle to ductile transition curve, 2002 to 2005.
ments by collapsing 1 side to the point and shifting the
midside nodes to a quarter position. To obtain the 1/√r The results were then incorporated in the ESIS guide-
singularity, the tip nodes have to undergo the same dis- line document entitled “Guidance on local approach of
placement, whereas the 1/r singularity is produced with- rupture of metallic materials”.34
out midside nodes being shifted and without It is interesting to note that both singular elements for
displacements of the tip nodes being constrained. After stationary cracks and regular meshes for crack growth
completing EPFM solutions, a large‐strain FEM analysis simulation have been developed at a time when computer
was capable of simulating a crack tip blunting, both in capacities were still rather limited. Today, well‐
displacements, and principal stresses.13 established procedures and exponential increase in com-
It is now irresistible temptation for this author to pay puter capacities enabled wide range of applications,
a respect to early CFM efforts in ex Yugoslavia, by including modifications of J‐integral (heterogeneous
Mladen Berković, who was the first in the region to materials like welded joints, ie, multimaterial bodies,78-
84
completely understand, develop, and use the FEM. His 3‐dimensional problems,85,86 volume forces,87 dynamic
doctoral thesis on thin plate isoparametric finite ele- problems,88 thermal problems,89,90 nonflat surfaces like
ment54 booked him 1 of just 16 places in the First World thin shells.91-94
Congress on Computational Fracture Mechanics.55 By all This spectacular development both in software and
means, Mladen was the father of computational mechan- hardware provided sound basis for reliable structural
ics in exYu, including CFM.56-59 integrity assessment of almost any practical problem,
Anyhow, after all these efforts and developments, sin- however complex it may be. As an illustration, one can
gular elements gradually lost their utmost importance, think about complex 3D geometry, like cracked housing
mainly because of an alternative way to calculate fracture of HP turbine, 2 pipes, cracked at the intersection, or
mechanics parameters, eg, to use virtual crack extension complete aircraft structure with a crack under complex
for J‐integral calculations,60,61 providing useful results dynamic loading. Therefore, several examples with practi-
even with coarse meshes, as well as because they were cal application of fracture mechanics to structural integ-
unable to simulate crack growth. Another important rity assessment of different components are presented
alternative, using regular elements, is separated into 3 now.
groups52: Computation procedure for the structural integrity
• Node release techniques, controlled by fracture assessment of damaged housing of high‐pressure turbine
mechanics parameters (J, CTOD, CTOA).62-65 thermoenergetic power plan is presented in Jovocic
• Cohesive elements.66-70 et al.95 It can be seen from Figure 1, which is an excellent
• Constitutive equations based on damage mechanics example of modelling a complex structure, that the crack
concepts.71-76 of depth 30 mm and length up to 375 mm produced the
At this point, reader should not oversee 2 significant equivalent stress of relatively low level, 100 MPa, with
contributions, made by joint effort of ESIS Technical the corresponding stress intensity factor, Keq = 8 MPa m0.5,
Committee 8 (TC8) members. First, it was round robin significantly lower than the fracture toughness, leading to
on CT specimen modelling, resulting in the Recommen- the conclusion that the integrity of turbine housing was
dations for Use of FEM in Fracture Mechanics in not jeopardised, although the crack was detected during
1991,21 including mesh, solution procedure, material repair. More details, about this procedure and results
law, type of analysis, crack tip modelling, fracture obtained, are given in Jovocic et al.95
mechanics parameters, and reference solution, This was As another example of a complex structure modelling,
followed by the round robin on micromechanical model- we present here numerical simulation of cracks in coated
ling of fracture mechanics specimens as initiated by W. materials,96 describing the effect of multicracking under
Brocks, and performed in following 3 phases: the thin layer during the hot dip galvanised steel process
(Figure 2A). For the modelling of bimaterial steel‐zinc
* Phase 1: Determination of local approach parameters structure, ABAQUS 6.10.1 was used, providing results
for ductile tearing (task A) and for cleavage (task B) for different cases, ie, for 1 crack and for 3 cracks under
for a German 20MnNiM055 RPV reactor pressure ves- the loading 500 MPa, as shown in Figure 2B.96 One crack
sel steel, 1993 to 1995. model indicated SIF values 16.2 to 24 MPa m0.5 for crack
SEDMAK 5

FIGURE 1 The equivalent stress in


vicinity of the crack (375 × 30 mm)95
[Colour figure can be viewed at
wileyonlinelibrary.com]

(A) (B)

FIGURE 2 A, Structure of steel‐zinc bimaterial. B, FEM model, 3 cracks, loading 500 MPa96 [Colour figure can be viewed at
wileyonlinelibrary.com]

length 0.03 to 0.06 mm.96 Results for the model with 3 toughness for these alloys, eg, a SIF value for the steel
cracks (Figure 3) indicated increased SIF value because (80‐170 MPa m0.5). More details, especially about
of the cumulative effect of cracks. Anyhow, this increase debonding area and its interaction with cracks, are given
does not exceed a critical value of mode I fracture in Azari et al.96
Yet another practical application of the FEM is an
estimation of the elastic‐plastic semielliptical through‐
wall crack opening displacement (COD),97 with an aim
to analyse the leak‐before‐break of a pressuriser. The
COD value is calculated through thickness along the cen-
tre of crack, and used for the calculation of J‐integral by
the engineering approach taking into account the mate-
rial full stress‐strain data. Actually, this paper presents
an engineering procedure for determining critical crack
length by combining J‐integral values at the crack tip
and COD value, in respect to the fracture toughness of
the material. Conditions for leak before break (LBB) are
met if any value of the J‐integral overcomes the J‐value
at stable crack initiation. Under this condition, the orifice
size can be estimated by using calculated COD values in
FIGURE 3 SIF vs crack length for model with 3 cracks96 [Colour respect to the postulated crack length 2a. The calculation
figure can be viewed at wileyonlinelibrary.com] showed that steam mass flow at the critical crack length
6 SEDMAK

(4560) of 3D 20‐node elements, as shown in Figure 5 for


specimen with crack tip in FGHAZ (Figure 6). Material
behaviour was modelled by multilinear stress‐strain
curves, obtained for each region of welded joint (WM,
BM, FGHAZ, and CGHAZ) through a combination of
microhardness values and the Ramberg‐Osgood law, as
explained in Adžiev et al.99 Numerical results for F vs
FIGURE 4 A, Finite element model of pressuriser. B, von Mises crack mouth opening displacement (CMOD) are shown
stress distribution97 [Colour figure can be viewed at in Figure 7 for the specimen with crack tip in the
wileyonlinelibrary.com] FGHAZ, together with results of 2D FEM calculation
and the experimental results. Good agreement between
2a = 80 mm is 0.109 kg/second. This value is significant 3D FEM and experimental results is obvious, verifying
enough to recognise LBB effects before final rupture numerical simulation.
occurs. Figure 4 shows FE model of the pressuriser and The equivalent strain distribution around the crack
results for von Mises stress at the crack tips. tip in the FGHAZ, obtained by 3D FEM, is shown in
Now, we focus our attention to welded joints, treated Figure 8A. Significant influence of overmatching is obvi-
here primarily as multimaterial structure. To start with, ous, acting as barrier for strain development, pushing
one can mention the review paper by Zerbst et al,98 basi- the strain to the more ductile base material and
cally on mismatch, misalignment, and round robin on protecting the welded joint from sudden failure. This
residual stresses for welded joints, as an excellent over- was also the case for the crack tip in the CGHAZ
view of weldment fracture mechanics in general. (Figure 8B). More details are given in Adžiev et al,99
Two detailed analyses of crack effects in HAZ are now explaining also some subtle differences between 2 cases,
presented, starting with extensive research on weldment like decreasing of the fracture resistance of the welded
fracture mechanics,99 where 2 different crack tip loca- joint in the case of crack tip in FGHAZ, because of strain
tions were analysed: 1 in the fine‐grained (FG) HAZ constraint because of CGHAZ, and contrary to that its
and the other on in the coarse‐grained (CG) HAZ of increase in the case when the crack tip is located in the
slightly overmatched welded joint. The 3D FEM using CGHAZ, because the more ductile surrounding regions
ANSYS 5.7 was applied to simulate the experimental
research on tensile specimens,99 ie, to assess effects of
welded joint heterogeneity on the distribution of stress
and strain in front of the crack tip. Special care was taken
to model HAZ and crack region, requiring large number

FIGURE 5 The 3D model with enlarged crack tip region99 FIGURE 7 The F‐CMOD curves for crack in FGHAZ99 [Colour
[Colour figure can be viewed at wileyonlinelibrary.com] figure can be viewed at wileyonlinelibrary.com]

FIGURE 6 Location of crack in FG HAZ99 [Colour figure can be viewed at wileyonlinelibrary.com]


SEDMAK 7

(A) (B)
FIGURE 8 Equivalent strain distribution at remote stress σ = 291.7 MPa99 [Colour figure can be viewed at wileyonlinelibrary.com]

of the WM and FGHAZ (compared to CGHAZ) accom-


modates the strain.
Numerical simulation of cracked welded joints is also
reported in Doncheva et al,100,101 where an experimental
research on the so‐called large cracks (LC) and small
cracks (SC)102 has been modelled by the use of 3D FEM.
To verify procedure, first the BM was modelled, and then
the welded joint, with different tensile properties in 4 dif-
ferent regions (BM. WM, CGHAZ, FGHAZ). Verifying
procedure included an analysis of the influence of crack
geometry (LC, SC) on fracture parameters as shown in
Figure 9, for J‐CMOD plot in the case of the SC, indicat-
ing good agreement between numerical and experimental
results. Figure 10 shows enlarged FE mesh around the SC
FIGURE 9 The J‐CMOD curves for tensile panel with small and the stress distribution at the crack tip, obtained by
crack100 [Colour figure can be viewed at wileyonlinelibrary.com] using ABAQUS.

FIGURE 10 Results obtained from BMSC, with details on mesh and stress distribution100 [Colour figure can be viewed at
wileyonlinelibrary.com]
8 SEDMAK

predict the behaviour of components with different geom-


etry under different loading conditions. To overcome
these deficiencies, a local approach was introduced based
on a large number of models of microscopic damages, as
an effort to explain and predict crack growth and macro-
scopic failure.
Following differences in 2 basic fracture mechanisms
for metallic materials, namely cleavage and ductile frac-
ture, the local approach developed 2 different
micromechanical models. On one hand side, it was rela-
tively simple model for brittle fracture, ie, cleavage,28,29
whereas more complex mechanisms of ductile fracture
FIGURE 11 Finite element model for WMLC101 [Colour figure
can be viewed at wileyonlinelibrary.com]
were modelled as a process of void nucleation, growth,
and coalescence,30,31 which is the focus here. In this
Three‐dimensional FE models were also developed for chapter, numerical methods, used for simulation of duc-
2 tensile specimens, made of welded joint, with different tile fracture process, are presented as a compilation of
surface cracks, large and small. The zone ahead of the previously accomplished research,104-157 followed by an
crack front is modelled with minimum of 2 layers of ele- overview of more recent results.158-166
ments with a highly refined mesh stretch out across the
ligament, Figure 11, because of expected damage and
3.1 | Micromechanical modelling of
crack propagation in this region. Tensile properties of dif-
ductile fracture
ferent regions in the undermatched welded joint have
been obtained by using special iterative procedure.103 The first micromechanical investigations113-116 included
The results for J‐CMOD and J‐remote strain are given round tensile specimens and notched tensile specimens
in Figure 12. One should notice relatively large difference with different values of notch radius, enabling develop-
in the case of J‐CMOD for LC (Figure 12B) probably ment of ESIS procedure21. Further investigations117-120
because of constraint, produced by undermatching, espe- lead to the ESIS draft “Guidance on local approach of
cially for increased loading. More detailed analysis is rupture of metallic materials”.34 The more recent
given in Doncheva et al.101 methods and approaches are focused to specific problems,
including void coalescence criterion, detailed microstruc-
tural analysis, and the effect of size, shape, and distribu-
3 | MICROMECHANICAL M ODEL- tion of voids.121,122 These micromechanical models have
LING —LOCAL A PPROACH TO been successfully applied to several classes of problems,
FRACTURE including pressure vessels,123-125 pipelines,126,127 and
welded joints.128,129
Driving force for local approach development was the fact Modelling of ductile fracture includes void nucleation
that elastic‐plastic fracture mechanics parameters, like around inclusions and second‐phase particles, followed
COD and J‐integral, cannot completely describe and by their growth and coalescence (Figure 13). Proposed

(A) (B)

FIGURE 12 The J‐CMOD curves for 3D numerical model compared with experimental results: (A) WMSC and (B) WMLC101 [Colour
figure can be viewed at wileyonlinelibrary.com]
SEDMAK 9

FIGURE 13 Process of ductile fracture


on geometries: (A) without and (B) with
initial crack135 (A) (B)

   
models typically assume that voids nucleate when critical R σm
ln ¼ ∫α⋅ exp β⋅ ⋅dεpeq
stress within inclusion or at inclusion‐matrix interface R0 σ eq
has been reached.
Early attempts to model ductile fracture assumed that where R stands for the mean void radius, R0 is its initial
voids do not significantly affect material behaviour. These value, σm/σeq represents the stress triaxiality, and dεpeq is
so‐called uncoupled models30,31 typically used von Mises the equivalent plastic strain increment. The critical value
yield criterion and have not incorporated any damage of void growth ratio, (R/R0)c, is the value at the crack ini-
parameter into the constitutive equation. Soon after- tiation. Values of parameters α = 0.283 and β = 1.5 for
wards, the so‐called coupled models were developed,36 Rice‐Tracey model are given in the original reference,32
based on constitutive equations suggested by Gurson,35 as the constants for a given steel.
with damage parameter “built into” the numerical proce- The main advantages of this approach are simple
dure and estimated during the finite elements (FE) anal- numerical procedure and possibility to use the results of
ysis. Main variable was void volume fraction, which was a single FE calculation for many postprocessing routines.
directly incorporated into the flow criterion. Numerical It makes this approach favourable for engineering prac-
and experimental tests have shown that this model, com- tice, as it enables quick assessment of crack initiation
monly called Gurson‐Tvergaard‐Needleman (GTN) on tested material or plant component.142
model, can describe both damage development at micro-
scopic level and plastic strain, at macroscopic level of 3.3 | Coupled modelling
material behaviour. Some of the recent advances114-117
proved that GTN model is capable to predict the macro- The coupled modelling of ductile fracture initiation con-
scopic ductile toughness using microscopic parameters.136 sider material as a porous medium, where the influence
One of the most successful recent models is CGM,37 of nucleated voids on the stress‐strain state and plastic
developed by incorporating the Thomason's limit load cri- flow is taken into account, as defined by GTN model.
terion,132 into the GTN model. Here, the void volume fraction f is introduced into the
expression for plastic potential35,36,144:
0 0  
3σ ij σ ij 3q2 σ m  
3.2 | Uncoupled modelling φ¼ þ 2q1 f * cosh − 1 þ ð q1 f * Þ 2 ¼ 0
2σ 2YS 2σ YS
According to the model of Rice‐Tracey,32 improved by 0

Beremin,29 Huang,139 and Chaouadi et al,140 void growth where σYS is the yield stress, σ ij the stress deviator, σm is
is strongly dependent on stress‐field multiaxiality. All the mean stress, q1 and q2 parameters introduced
these models are uncoupled; hence, damage is calculated by Tvergaard,144 and f* is a function of the void volume
by postprocessing routines, after FE analysis of the stress fraction36:
and strain fields. 
According to Beremin,30 growth of an isolated void in * f for f ≤ f c
f ¼
a remote uniform von Mises stress field can be defined as: f c þ K ðf −f c Þ for f > f c
10 SEDMAK

where f c is the critical value at which void coalescence Instead of a constant value, Zhang et al37 proposed a lin-
occurs. For f* = 0, the plastic potential Φ is identical with ear dependence of a on hardening exponent n, which is
that of von Mises. The parameter K defines the slope of a applied in the CGM. Taking the void space ratio r as
sudden drop of the force on the force‐diameter reduction given in Zhang et al,37 the critical void volume fraction
diagram, thus defining the final stage of ductile fracture‐ f c can be calculated during the FEM, ie, it is not a mate-
void coalescence, which leads to complete loss of the rial constant in the CGM. This critical value depends on
material load carrying capacity. the strain field, so it does not necessarily have the same
In the initial phase of the ductile fracture, the voids value in all finite elements and even in all integration
nucleate mostly around the nonmetallic inclusions. To points within 1 element. This fact is especially important
quantify this micromechanism, the volume fraction of in joints, where the material heterogeneity causes severe
nonmetallic inclusions, f V, that can be determined using gradients in the stress and strain fields around the crack
light microscopy, should be known.145 Two phenomena tip before and after the crack growth initia-
contribute to the increase of the void volume fraction: the tion.130,133,134,137,138 If the GTN model is used, it is neces-
growth of existing voids and the nucleation of new voids: sary to determine this value (eg, by transferring its value
from a tensile specimen or from unit cell analyses) and
f_ ¼ f_ nucleation þ f_ growth ; f_ nucleation ¼ A_εpeq and f_ growth ¼ ð1− f Þ_εpii use it as material parameter for fracture assessment.

where ε_ pii is the plastic part of the strain rate tensor.


Growth of nucleated voids is strongly dependent on 3.4 | ESIS round robin—modelling of
stress and strain state. Most of experiments and analyses ductile fracture initiation
show an exponential increase with the stress triaxiality, The starting point in making an overview of
defined as the ratio of the mean stress σm and equivalent micromechanical modelling results is the ESIS round
stress σeq. robin on ductile fracture initiation. Low‐alloyed steel 22
Chu and Needleman146 proposed a model for initia- NiMoCr 3 7 (according to DIN), used mainly for pressure
tion of the secondary voids. According to this model, vessels, has been tested on the specimens in forged heat‐
the secondary void initiation can be described using nor- treated condition,147 and later used for ESIS TC8 round
mal distribution: robin project on micromechanical models, Phase II Task
"  # A1, through which the numerical part of the analysis is
fN 1 εeq −εN 2 performed. Tensile properties of steel were determined
A ¼ pffiffiffiffiffiffi exp −
sN 2π 2 sN on a smooth specimen at 0°C: yield strength
Rp0.2 = 476 MPa and ultimate tensile strength
where f N denotes the volume fraction of secondary‐void Rm = 620 MPa.
forming particles, εN is the mean strain at void nucle- Initial void volume fraction f 0 is determined by quan-
ation, and SN is the corresponding standard deviation. titative optical microscopy. Volume fraction of sulphides
All parameters described above can be divided into 3 and oxides in tested steel has been determined based on
groups: equality with surface fraction:
Ai
1. Constitutive parameters: q1, q2. V V ¼ AA ¼
AT
2. Initial damage parameters: f 0, f N, and εN (according
to Chu‐Needleman). where VV and AA are the volume and surface fraction of
3. Critical damage parameters: f c, f F , and f u. detected sulphides and oxides respectively, Ai is the area
of the detected inclusions, and AT is the measurement
The GTN model has undergone many modifications, field area. Volume fraction f 0 is determined as a mean
one of them proposed by Zhang et al,37 who applied the value of surface fraction of nonmetallic inclusions for all
Thomason's void coalescence criterion,132 to formulate of 100 measurement fields. In this way, the value of initial
the complete Gurson model (CGM). This model intro- volume fraction of nonmetallic inclusions f 0 = 0.00226
duces a criterion for the onset of void coalescence: has been determined. Considering the similar densities
of structural steel and Fe3C particles, it has been adopted
   
σ1 1 β  that the volume fraction of iron carbides is 0.03, used as
¼ α −1 þ pffiffi 1−πr 2
σ r r f N in Chu‐Needleman equation. Using procedure
described in Younise,134 the mean edge‐to‐edge distance
where r1 is the maximum principal stress and a and b are between inclusions, represented by the average value of
constants fitted by Thomason132 (a = 0.1 and b = 1.2). mean free path, λ = 219 μm, is determined.
SEDMAK 11

(A) (B)

FIGURE 14 Standard round tensile specimen: dimensions (A) and simulation by FE mesh (B)104 [Colour figure can be viewed at
wileyonlinelibrary.com]

Calculations on standard tension specimen calculation in which secondary void nucleation is taken
(Figure 14) are made, and criteria of initiation of ductile into account predicts significant deviation from experi-
fracture are established according to Rice‐Tracey model mental data after neck has been formed.
and GTN model. Because of the symmetry, one‐quarter Change of void volume fraction in radial direction for
of the specimen is used for FE model. Loading is intro- the last 3 tensile steps is shown in Figure 16A. The dam-
duced by prescribing displacements of specimen edge in age is most severe in the middle of the specimen, and this
several steps. Eight‐noded quadrilateral elements with is the location of crack initiation. Accordingly, the critical
reduced integration are used, and the results obtained values of model parameters had been determined for ele-
using 4‐noded elements can be found in Nichols.49 Size ment in the specimen centre. The last step of numerical
of elements in the centre of the specimen is calculation of tension corresponds to the specimen frac-
0.3 × 0.125 mm. Material nonlinearity is modelled using ture, which can also be seen in Figure 16A.
true stress‐true (logarithmic) strain curve, and large Critical void volume fraction, f c, is determined accord-
strain (updated Lagrangian) FE formulation. ing to Figure 11B, from the increase of void volume frac-
The GTN model is integrated into the FE software, tion in central element of the specimen, depending on
ABAQUS.39 Elastic‐plastic FE calculations are carried cross‐sectional reduction. For determination of f c value,
out with Tvergaard‐Needleman parameter value for the necking process experimental result,117
q1 = 1.5, according to Tvergaard,144 both with and with- ΔD ≈ 2.63 mm is used. This is the value at the point of sud-
out secondary void nucleation. When secondary void den drop of force caused by coalescence of voids. Obtained
nucleation is taken into account, parameter D according value is in good agreement with recommended
to the Chu‐Needleman model is determined for εN = 0.3, values114,117 for this steel ( f c = 0.05) and the steel that is
SN = 0.1, and f N = 0.04, as the reference values, most fre- most similar to it according to the American standard
quently used for various steels.136,150 A508Cl.2, f c = 0.045.151 The critical value (R/R0)c = 3.045
Figure 15 shows a diagram load, F , vs reduction of is obtained in the same way as shown in Rakin et al.119
cross section, ΔD, of a round specimen. Calculation with- The second part of the micromechanical analysis
out nucleation of secondary voids and calculation with includes determination of the onset of crack growth on
traditional von Mises criterion give approximately the standard CT25 specimen (Figure 17). Within the scope of
same results. Moreover, differences between the von the round robin project, Phase II, Task A2,117 6 specimens
Mises law and the GTN model are negligible almost to were tested, providing curves load‐load line displacement
the point of the initiation of damage. Finite element and J‐integral‐crack growth increment. One half of the
specimen is modelled using quadrilateral 8‐noded
isoparametric plain strain elements (Figure 17A), with
the same true stress‐true strain curve as for the tensile
smooth specimen calculation. Loading is modelled by pre-
scribing displacements of the load line, in several steps.
Crack tip is modelled using refined FE mesh without
singular elements (Figure 17B). According to numerous
researches,117,136 element size at the crack tip corre-
sponds to the mean free path between the nonmetallic
inclusions (λ = 219 μm), which is considered to be unaf-
fected by orientation. Therefore, the size of elements at
the crack tip was 0.2 × 0.2 mm, used for both calcula-
tions: Rice‐Tracey void growth model with von Misses
yield criterion and GTN model with the same initial void
FIGURE 15 Load, F , vs reduction of cross section, ΔD, of the volume fraction value as in the tensile smooth specimen.
round tensile specimen107 Subsequent void nucleation is not taken into account.
12 SEDMAK

(A) (B)

FIGURE 16 A, Distribution of void volume fraction, f . B, Critical void volume fraction, f c107

(A) (B)

FIGURE 17 A, CT specimen. B, FE mesh with detail near the crack tip107 [Colour figure can be viewed at wileyonlinelibrary.com]

Distribution of void volume fraction ( f ) near the below numerically determined curves, according to the
crack tip, at the crack growth initiation, is shown in expression152:
Figure 18A. Concentration of large values ( f > 0.06) very
close to the crack tip is obvious. One can also see a large ηU
a
0
Ji ¼ and η ¼ 2 þ 0:552 1− (10)
variation of f in the element in front of the crack tip. Dis- Bn ðW −a0 Þ W
tribution of the stress component, normal to the crack
direction, σ2, at the crack growth initiation is shown in where U is external work‐area below F ‐VLL curve,
Figure 18B. W − a0 is ligament length, and thickness Bn = 20 mm
Based on FE calculation, load F ‐load line displace- because of 20% side grooves (specimen nominal thickness
ment VLL curves are established. The value of J‐integral B = 25 mm). Load line displacement VLL corresponding
at crack initiation, Ji, is obtained using the area to the onset of crack growth is determined according to

(A) (B)

FIGURE 18 A, Distribution of f in front of crack tip. B, Distribution of σ2 at the crack growth initiation107 [Colour figure can be viewed at
wileyonlinelibrary.com]
SEDMAK 13

the values (R/R0)c = 3.045 and f c = 0.0611, determined for the element at the crack tip, until the onset of crack
for smooth specimen. Failure in front of the crack tip is growth, indicating significant differences.
defined by R/R0 ≥ (R/R0)c and f ≥ f c. One should notice that the recommendation to use
In this analysis, 2 failure criteria are possible, 1 at the the size of elements at the crack tip similar to the mean
Gauss point nearest to the crack tip (GP1) and the other free path between nonmetallic inclusions was correct in
one for the entire element. Having in mind that the goal both models, uncoupled and coupled, because the onset
is determination of Ji, and that its experimental verifica- of crack growth is successfully predicted. Anyhow, there
tion means determination of final stretch zone width, fail- are other expressions, explanations, and recommenda-
ure at the first Gauss point is more realistic, as also tions,153,154 regarding the size of FE at the crack tip,
proved by the following 2 values of Ji (experimental value which have proved to be correct for certain
was Ji = 230 N/mm117): materials.155,156
However, disregarding the applied method for deter-
a. Ji = 220.4 N/mm (GTN model), Ji = 249.2 N/mm mination of FE size at the crack tip, there is unanimous
(Rice‐Tracey model), failure at the GP1. opinion that FE size strongly effects prediction of the
b. Ji = 267.9 N/mm, failure of the entire element at the onset of crack growth, making micromechanical model-
crack tip. ling sensitive to FE mesh. This is not clearly visible in
Figure 20, where force F is plotted against CMOD until
Figure 19 shows increase of void volume fraction vs the failure at the GP1, for 3 different values of FE at the
load line displacement VLL, in Gauss points (GP1‐GP4) crack tip, 0.2 × 0.2, 0.4 × 0.4, and 0.1 × 0.1 mm, but def-
initely true if Ji values are compared (220.4 N/mm for
0.2 × 0.2 mm, 414.5 N/mm for 0.4 × 0.4 mm, and
127.2 N/mm for 0.1 × 0.1 mm, almost proportional to
the element size).

3.5 | Micromechanical modelling of crack


initiation and growth in welded joints
Weldment fracture mechanics has been in focus for long
time, because welding is, unfortunately, ideal process to
produce cracks.49 This is also the reason to apply
micromechanical modelling, presented here briefly, as a
fraction of extensive study on welded joints made of
NIOMOL 490, high‐strength‐low‐alloyed (HSLA) steel
used mostly for pressure vessels.134
To start with, the quantitative microstructural analy-
FIGURE 19 Void volume fraction in gauss points of the first FE sis was performed to estimate the micromechanical mate-
in the ligament107 rial parameters: volume fraction of nonmetallic

(A) (B)

FIGURE 20 A, Load F vs load line displacement VLL until the moment of crack growth initiation using f c as failure criterion. B, Part of
diagram corresponding to crack growth initiation107] [Colour figure can be viewed at wileyonlinelibrary.com]
14 SEDMAK

TABLE 1 Microstructural parameters of welded joint materials

Material fv fN λ, μm (m)

BM (NIOMOL 490K) 0.0094 0.014748 578


HAZ 0.0086 0.014748 497
WM 0.0194 0.010685 202

inclusions ( f v) and mean free path (λ) between the non-


metallic inclusions in the BM, HAZ, and WM.134 The ini-
tial porosity ( f 0) is assumed to be equal to the volume
fraction of nonmetallic inclusions ( f v). Volume fraction
of void nucleating particles ( f N) was calculated from the
content of carbon in tested materials, using the carbon
FIGURE 22 Finite element mesh for the region of measured
content in BM and filler metal for HAZ and WM respec-
strain134 [Colour figure can be viewed at wileyonlinelibrary.com]
tively, as presented in Table 1, together with volume
fraction of nonmetallic inclusions ( f v) and mean free
path (λ). and BM) using digital image correlation method with
Stress‐strain curves of various zones of welded joint, ARAMIS measuring system118 (Figure 21).
especially heat‐affected subzones (CGHAZ and FGHAZ), Then, the smooth tensile specimen was numerically
are difficult to be determined by conventional methods. modelled using ABAQUS 6.7 with 3‐dimensional 8‐node
Therefore, special iteration procedure based on compari- brick elements. Mesh has been refined in regions where
son of experimental and numerical results103 was applied strains were measured (Figure 22).
here for the first time to estimate true stress‐true strain Engineering remote stress‐true strain data for welded
curves. Toward this end, smooth tensile plate (without joint regions (BM, WM, and HAZ) was obtained from
cracks) was cut transversally from welded plate and lon- strains, measured by ARAMIS (Figure 21) using corre-
gitudinal strains at different load levels measured in var- sponding applied force. Strains in each region at corre-
ious regions of welded joints (WM, CGHAZ, FGHAZ, sponding forces were calculated as an average value

FIGURE 21 Geometry of smooth


tensile specimen with strains measured by
ARAMIS134 [Colour figure can be viewed
at wileyonlinelibrary.com]
SEDMAK 15

TABLE 2 Mechanical properties of the materials for numerical estimated from the obtained experimental true stress‐
iterations strain curves. The mechanical properties (E and σYS)
Material Iteration E (MPa) σYS (MPa) n K (MPa) and Hollomon parameters (K and n) are given in
Table 2 for 3 iterations as example for performing itera-
WM 1 169 320 459 0.1 944
tion procedure. The third iteration for Hollomon parame-
2 195 000 550 0.08 971
ters is represented in Figure 23A as a plot true stress‐true
3 200 000 530 0.21 1255
strain curves for all 4 regions of welded joint and in
CGHAZ 1 190 037 459 0.11 935
Figure 23B as a plot of strain vs distance along the weld-
2 200 000 530 0.09 947
3 203 000 550 0.17 968
ment, compared with the experimental results. Finally,
Figure 24 shows comparison between tested specimen
FGHAZ 1 190 037 459 0.11 935
and longitudinal true strain distribution in numerical
2 200 000 500 0.08 901
3 195 000 500 0.23 1217
models for 3 iterations, indicating good agreement after
third iteration.
BM 1 176 972 459 0.11 929
2 202 000 540 0.1 940
3 202 900 520 0.22 1157
3.6 | Modelling of crack initiation and
growth in welded joints
along measured line. Then, engineering remote stress (σ) Micromechanical modelling was first done on 2 SENB
was calculated using initial cross section of tested speci- specimens, 1 precracked in WM and the other 1 in
men and then converted to true stress (σT) using expres- FGHAZ. They were analysed as 2D model under plane‐
sion σT = σ (1 + ε) where ε is average strain.134 True strain conditions with isoparametric quadrangular 8‐
stress‐strain behaviour of materials was found to follow noded elements with 2 × 2 Gauss integration used for
Hollomon power law up to maximum load according to simulating crack initiation and 4‐noded elements with
expressions: full Gauss integration used for crack propagation. In front
of the crack tip, squared finite elements (0.2 × 0.2 mm for
ε ¼ εe þ εp; σ ¼ Eεe if σ ≤ σ YS; σ ¼ K εpn if σ > σ YS: specimen with a precrack in WM and 0.5 × 0.5 mm for
specimen with a precrack in FGHAZ) were used, in
where εe and εp are elastic and plastic strains respectively, accordance with the mean free path λ between nonmetal-
E is Young modulus, σYS is yield strength, K is strength lic inclusions in tested materials (see Table 1). For this
coefficient, and n is material hardening exponent in analysis, we present only specimen with precrack in WM
Hollomon equation. The proper combination of mechan- (Figure 25).
ical properties was determined by varying them until More detailed analysis is given in Younise,134 includ-
good agreement between numerical and experimental ing verification of FEM results by comparing with exper-
results at different loads was obtained. This procedure imental ones, using plots of CMOD vs force ( F ), as well
was introduced for engineering stress‐strain curves and as the effects of heterogeneity and constraint on ductile
applied to determine yield strength and hardening coeffi- crack initiation and propagation. Here, we focus our
cient in Sedmak et al.103 The initial iteration was attention only on crack growth, ie, J‐R curves. The crack

(A) (B)

FIGURE 23 A, True stress‐true strain curve. B, Strain vs distance along the weldment134 [Colour figure can be viewed at
wileyonlinelibrary.com]
16 SEDMAK

FIGURE 24 Comparison between tested specimen and longitudinal true strain distribution in numerical models for 3 iterations at the
same remote displacement134 [Colour figure can be viewed at wileyonlinelibrary.com]

(A)

(B)

FIGURE 25 A, Geometry of SENB specimen with precrack in WM. B, Finite element mesh134 [Colour figure can be viewed at
wileyonlinelibrary.com]

growth (Δa) has been simulated by tracing the path of final failure f F is reached according to the relation
completely damaged elements, shown in Figure 26. Crack f F = 0.15 + f 0. Then, the corresponding value of J‐inte-
growth has been estimated by multiplying the original gral is calculated to construct crack growth resistance
length of an element (lc) with the number of completely curves (Figure 27).
damaged elements. The element is assumed to be The effect of finite element size on the prediction of
completely damaged when the void volume fraction at crack initiation and crack resistance curves was analysed
SEDMAK 17

FIGURE 26 Distribution of void


volume fraction for SENB specimen with a
precrack in WM134 [Colour figure can be
viewed at wileyonlinelibrary.com]

HAZ was used.134 Three‐dimensional finite damage


models for welded tensile panels with surface crack were
developed. Ductile crack growth initiation values and J‐R
curves have been analysed numerically, including effects
of crack tip constraint, mechanical heterogeneity, and
loading condition. Finite element code ABAQUS with
material user subroutine UMAT was used to model the
tensile panels with a surface crack in FGHAZ (length
22 mm, depth 5 mm), with 3‐dimensional 8‐node ele-
ments (Figure 28).
Keeping in mind that most metals have very narrow
fracture process zone,129 3D modelling of the whole spec-
imen using a damage constitutive model is unnecessary
and inefficient. Therefore, a single layer of elements with
FIGURE 27 Experimental and numerical J‐R curves for SENB a damage constitutive model was used in front of the
specimen precracked in WM134 [Colour figure can be viewed at crack to simulate the ductile tearing, following procedure
wileyonlinelibrary.com] described in Rakin et al112 and Zrilic.115 In the tearing
zone, ahead the crack front, minimum 2 layers of ele-
by many researchers, including Sedmak and Rakin107 and ments with a highly refined mesh were used for CGM
Younise,134 and where it is found that the size of FE mesh modelling. The BM and WM have been modelled by
significantly affects the material resistance curves and using conventional J2 flow theory von Mises plasticity
crack initiation values for specimens with precracks in with large displacement analysis. The mesh size, lc, was
WM and HAZ. Here, this effect is clearly visible in chosen to approximate the mean free path between non-
Figure 27, where excellent agreement between experi- metallic inclusions.
mental and numerical results have been obtained with Comparison between numerical and experimental
lc equal to mean free path value, λ = 202 μm ≈ 0.2 mm, results of force ( F ) vs crack tip opening displacement
whereas strongly overestimated values for J were (δ5), force ( F ) vs CMOD, and CMOD vs δ5 has been given
obtained with lc = 0.5 mm. Similarly, J values for the in Younise,134 indicating good agreement, and, thus, ver-
crack in CGHAZ were strongly underestimated when lc ifying numerical procedure.
was taken as 0.25 mm instead of 0.5 mm.134 Here, we focus our attention to ductile crack growth
initiation, defined by J‐integral at initiation (Ji), modelled
using the critical void volume fraction criterion ( f c),
which represents the end of stable void growth and the
3.7 | Three‐dimensional finite element
start of void coalescence in the material. The values of Ji
analysis of tensile panels
have been estimated numerically at the middle of the
To study fracture behaviour of welded high‐strength specimen thickness in front of crack, where the highest
steels, tensile panel with semielliptical surface cracks in value of void volume fraction occurs. The value of
18 SEDMAK
Colour online, B&W in print

(A)

(B)

FIGURE 28 The 3D model for tensile panel with surface crack in FGHAZ: (A) 3D mesh for half of specimen with boundary conditions
and (B) detailed mesh for the region near the crack front134 [Colour figure can be viewed at wileyonlinelibrary.com]

Ji = 346.2 N/mm agrees well with the value obtained by including constraint levels analysed on basis of stress
experiment, using FSZ, Ji = 321.4 N/mm, as described triaxiality.
in details in Younise et al.134,149,157 Now, we focus on some of the most recent research,
Figure 29 shows the distribution of void volume frac- including ring pipe specimen to evaluate behaviour of
tion ( f ) at the instant of crack growth initiation, indicat- real structure, such as pressure vessel. In recently pub-
ing concentration of large values very close to the crack lished paper,158 the influence of specimen size and stress
front. The location of crack initiation is also shown in concentrator size/shape was analysed for a new type of
this figure, which corresponds to the location of maxi- specimen, the pipe ring notch bend specimen (PRNB).
mum stress triaxiality. Equivalent von Mises stress and In Musraty et al,158 fracture criteria of this specimen
plastic strain distribution at the onset of crack growth (Figure 33A made of HSLA steel, NIOMOL 490K, are
in FGHAZ is given in Figures 30 and 31. The effects of determined by micromechanical analysis. The specimens
applied stresses and strains on the distribution of void with radius 60 and 40 mm were precracked, while those
volume fraction at crack growth initiation can be seen with radius 85 mm were notched (tip radius 0.25 mm).
also in Figure 32. More details are given in Younise,134 Numerical analysis was performed in software package
SEDMAK 19

FIGURE 29 Distribution of void volume fraction ( f ) in HAZ at the onset of crack growth134 [Colour figure can be viewed at
wileyonlinelibrary.com]

FIGURE 30 Equivalent von Mises stress distribution at the onset of crack growth in FGHAZ134 [Colour figure can be viewed at
wileyonlinelibrary.com]

FIGURE 31 Equivalent plastic strain distribution at the onset of crack growth in FGHAZ [Colour figure can be viewed at
wileyonlinelibrary.com]
20 SEDMAK

FIGURE 32 Distribution of void volume fraction, which shows crack growth in FGHAZ134 [Colour figure can be viewed at
wileyonlinelibrary.com]

(A) (B)

FIGURE 33 A, Specimen dimensions and testing scheme. B, FE mesh158 [Colour figure can be viewed at wileyonlinelibrary.com]

Simulia ABAQUS (Figure 33B) using CGM with the ini- 24 mm (Figure 35B). One can also see that specimen
tial value of porosity f 0 set equal to the volume fraction radius, defect shape, and size do not have a pronounced
of nonmetallic inclusions, f v = 0.0094.158 influence on fracture behaviour.
Crack growth curves obtained experimentally are Another recently published paper163 on prediction of
shown in Figure 34A for precracked rings (R60 and ductile crack growth in laboratory samples (SENB, SENT,
R40) and R85 specimens, whereas their numerical coun- and CT), made of X65, welded grade pipe steel, used
terparts are shown in Figure 34B, indicating good agree- round notched bars (RNB) to identify damage parameters
ment for R40 and R60. Somewhat lower crack resistance (Figure 36) for both parent and weld materials. Toward
was obtained both experimentally and numerically for this end, FEM code MSC MARC was used with 3D 8‐
R85 specimen, because of its larger thickness. node brick elements specialised for ductile crack growth.
The influence of wall thickness is further examined Initial blunting (r = 0.15 mm, consistent with SZW
for R60 specimen, with initial thickness of 7.5 mm stretch zone width data) was used.
increased to 12 mm (Figure 35A) indicating lower resis- In Figure 37A, the comparison of the predicted
tance to crack growth (Figure 35B). On the other hand applied load vs crack mouth opening displacement, and
side, there is no effect if width is increased from 15 to stroke displacement (TRAV), with experimental data for
SEDMAK 21

(A) (B)

FIGURE 34 Crack growth curves: (A) experimental and (B) CGM158 [Colour figure can be viewed at wileyonlinelibrary.com]

(A) (B)

FIGURE 35 Crack growth for 2 models: (A) with different wall thickness and (B) wall thickness effect on crack growth curves158 [Colour
figure can be viewed at wileyonlinelibrary.com]

SENT base metal is shown. The overall agreement is very here, we just emphasise 1 important conclusion, that
good and confirmed by the qualitative comparison of the the CTOD at crack initiation is insensitive to the Lüders
predicted crack growth and crack front shape at test end plateau length.
(Figure 37B). It should be noted that finite element simu-
lation with damage is capable to reproduce all main fea-
tures of the deformation and crack growth process. 4 | A P P L I C A T I O N OF X F E M TO
Finally, we present just a fragment of an extensive FATIGUE CRACK GROWTH
study on the effect of Lüders plateau on crack resistance SIMULATION
curve with sent specimens.166 Material behaviour has
been modelled as shown in Figure 38, with hardening The FEM, as the most popular numerical method nowa-
exponent n = 0.05, damage parameter f 0 = 0.005, crack days, was also used for fatigue crack growth simula-
length a/w = 0.1, and crack length increment δa = 0.5 mm. tion,167-169 but restrictions in FEM crack growth
The stress triaxiality distribution ahead of crack tip for dif- simulations discouraged researches, until XFEM was
ferent levels of strain at Lüders plateau, εl, is shown in introduced and developed to solve different crack growth
Figure 38B, indicating that the stress triaxiality scales duc- problems.41-48 As explained in the introduction, the main
tile crack growth, depending on the Lüders plateau length. advantage of XFEM is to enable the crack growth without
More details are presented in Tu et al,166 including remeshing of the domain. The cracks are represented
effects of hardening exponent, n, damage parameter, f 0, with the help of 2 signed distance functions that are
crack length a/w, and crack length increment δa, while discretised on the same mesh as the displacement field
22 SEDMAK

FIGURE 36 Identification of damage parameters using only RNB traction data163 [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 37 A, Applied force vs TRAV. B, Crack growth comparison163 [Colour figure can be viewed at wileyonlinelibrary.com]

with first‐order shape functions. After each step of the different directions depending on loading and con-
propagation simulation, the SIFs are computed from the straints, eventually changing plane and direction of
numerical solution at several points along the crack growth, not to mention multiple cracking. This is difficult
fronts. Interaction integrals are used to extract the to simulate, and therefore, precise evaluation of fatigue
mixed‐mode SIFs with the help of auxiliary fields. After life of the damaged structure is not easy to achieve. The
that, fatigue crack growth model (like Paris‐Erdogan best estimates are made when stress intensity factors
law) can be used for fatigue life evaluation. (SIFs) are known for the real geometry, so only values
In the last couple of years, many papers dealt with the obtained for 3D models can be used for further calcula-
problems related to crack initiation and resulting damage tions without any modification.
growth under different loadings, but in the most cases, In Daux et al,173 authors used XFEM for analyzing
basic geometry was used in XFEM. Researchers carried models containing several cracks with voids, cracks with
out simulations either on 2‐dimensional specimen multiple branches, and cracks emanating from the holes.
models170,171 or 3D models of simplified structures.172 More recently, Kim et al174 investigated the effect of com-
Furthermore, even in 3D models, crack growth was sim- pression stresses, stress level, and stress order on fatigue
ulated in 1 plane, taking into consideration only 1 direc- crack growth of multiple site damage. In this investiga-
tion of propagation. In reality, cracks can grow in tion, the XFEM was applied to predict lifetime under
SEDMAK 23

(A) (B)

FIGURE 38 A, Lüders plateau. B, The stress triaxiality distribution ahead of crack tip166 [Colour figure can be viewed at
wileyonlinelibrary.com]

constant amplitude cyclic loading of fatigue tests on sev- 4.1 | Fatigue life estimation of CCT
eral multiple site damage specimens of Al 2024‐T3. Then, specimen
the multiple crack growths under service stress spectra
To demonstrate the capability and reliability of XFEM in
are calculated to investigate the effect of compressive
prediction of the crack propagation trajectory and the
stress, stress order, and the effect of sequence cyclic load-
evaluation of SIF values, specimen with a central crack
ing on stress level by using Forman and NASGROW
was analysed, as defined and described in details in
equations.174
Eldwaib et al.181 Dimensions of the specimen were
There are studies, such as Budyn et al,175 Loehnert
semiheight h = 40 mm, semiwidth b = 20 mm, and thick-
and Belytschko,176 and Singh et al,177 where XFEM was
ness t = 1 mm. Initial length of the crack was
used for analyzing various effects of multiple microcracks
2a0 = 4 mm, and a uniform tension of 25 MPa was
on a macrocrack. A homogenised XFEM was proposed in
applied.
Kumar et al178 for the evaluation of fatigue life of an edge
It is important to emphasise that 2D simulations of
crack plate in the presence of multiple discontinuities,
crack propagations are still dominant and—to the best
like holes, inclusions, and minor cracks. A detailed
authors' knowledge—simulations presented here were 1
numerical study is performed to decide the modelling
of the first attempts of XFEM application for 3D crack
region for the discontinuities. To improve the results, a
growth analysis. In 2D simulations, values for SIFs are
homogenisation scheme based on strain energy density
calculated in 1 point only—at the tip of the crack propa-
approach is used to determine the effective properties of
gating in plane, whereas, in 3D simulations, the values
heterogeneous region containing the holes, inclusions,
are calculated in several points/nodes along the crack
and both. A nonuniform meshing is used to discretise
front that propagates in space. In this way, it is possible
the entire domain. All mentioned studies and research
to determine the stress intensity factors for all 3 modes
activities, regarding XFEM, were concentrated on its
(which is important for fatigue life estimation in the case
implementation on relatively simple configurations, with
of mixed loading), while 2D analysis determines KI and
several cracks in 2D models of panels and sheets, where
KII only.
genuine multiple site damage (MSD) was not really an
issue. Also XFEM itself in these studies were carried out
through specifically made “custom code” numerical cal- TABLE 3 Comparison of KI (MPa mm0.5) for CCT, crack front 1
culations, which make their application hard and Step a (mm) a/b KI Morfeo KI ASGRO
reduced to a very small number of potential users. It
1 2.00 0.100 66.88 63.05
should be mentioned that there are also several studies
where XFEM was used for simulating 3D fatigue crack 5 3.98 0.199 95.04 90.63

propagations, but only with 1 crack.179,180 10 6.40 0.320 127.40 119.81


Now, we focus our attention on series of papers on 15 8.74 0.437 160.12 148.92
XFEM application published recently by Grbovic and 20 11.06 0.553 198.86 183.34
coworkers, which deserves to be named “Belgrade school
21 11.55 0.578 204.30 191.93
of XFEM simulation”,179-189 starting with an analysis of
22 12.00 0.600 211.41 200.24
fatigue crack growth in a CCT specimen.
24 SEDMAK

ABAQUS defines initial crack as a separate entity another 250 000 cycles. It will take additional 40 000 cycles
with no element mesh and the first step in 3D analysis to grow crack to final length a = 19 mm just before com-
of crack propagation is crack “opening”, followed by cal- plete failure.
culation of stresses.179 Morfeo/Crack for ABAQUS190 uses Comparing XFEM results with number of cycles
ABAQUS solutions to calculate stress intensity factors in obtained by NASGRO (Figure 39), it is evident that
nodes of the crack front and generates a file with KI, NASGRO predicts longer fatigue life, which can be
KII, and KIII results. Then, the equivalent stress intensity explained by the fact that NASGRO uses more sophisti-
factor Keq, which combines all 3 SIF modes, is calculated cated and empirically improved equation than Morfeo.
as well as the kink angle (crack propagation angle) which Namely, NASGRO software uses the multiparameter
defines the direction in which crack will be propagated in material equation for fatigue crack growth, contrary to
a next step. At the same time, Morfeo/Crack for ABAQUS the simple 2‐parameter Paris equation, used by
calculates number of loading cycles necessary to grow MORFEO. Anyhow, from an engineering point of view,
crack by a given length. results obtained by XFEM are acceptable, because they
Values for all 3 modes and equivalent SIF are shown are conservative.
for (x, y, z) coordinates of the nodes on the crack fronts
after each step of crack propagation in Eldwaib et al,181
including all nodes on the crack front of CCT after first
4.2 | Fatigue life estimation of multiple
and second steps of XFEM simulation. Values of SIFs
site damage
mode I for crack front 1 obtained by XFEM and NASGRO
software191 are presented in Table 3 for characteristic Two important investigations have been performed for
steps of crack growth. MSD effects on fatigue life of aluminium alloy Al‐
Comparison between XFEM and NASGRO results in 2024T3: 1 for tensile panel with 3 holes, each with 1 or
Table 3 indicates good agreement. In general, XFEM 2 cracks (Figure 40),186 and the other 1 practically the
gives somewhat greater SIF values, which can be same, but with 11 holes182,183 (Figure 41).
explained by the fact that thickness of analysed 3D model To predict crack growth rate of structure with MSD,
is relatively small. Once the SIF values are evaluated. it is accurate calculation of stress intensity factors (SIFs) is
possible to estimate fatigue life, ie, the number of loading needed. Therefore, the FE model of a panel
cycles which would propagate a crack to a complete fail- (L1 × L2 = 600 × 400 mm, Figure 40) with 3 holes (radii
ure. All relevant data, including material properties, are r = 2.4 mm at distances b = 25 mm) is subjected to uni-
given in Eldwaib et al.181 form uniaxial tensile stress (σ = 100 MPa). The middle
The values obtained by Morfeo/Crack for ABAQUS. hole has 2 radial cracks (nos. 1 and 2 in Figure 40), and
Figure 39 show that the crack front 1 will move by other 2 holes have 1 radial crack each (nos. 3 and 4 in
2 mm (from initial crack length a0 = 2‐4 mm) after cca Figure 40). The SIFs are calculated for each crack tip as
700 000 load cycles, while it will move next 2 mm (to total a function of crack length, and used to predict direction
crack length a = 6 mm) after approximately 250 000 cycles and relative rate of growth of each crack. The FEM natu-
(because of the symmetry, crack front 2 will behave in the rally includes the effects of crack interaction, enabling
same manner). After that, front 1 will start to move rap- easy comparisons of the rates of no‐interaction growth.
idly and crack length will be doubled (a = 12 mm) after Figure 41 shows that the length of each crack varied with
its stress intensity factor history. The lengths of middle
cracks (1 and 2) are longer than other 2 cracks (3 and
4), which has been also observed in experiments. The
analysis was arbitrarily stopped after the 62nd step,
because after that step the cracks would form 1 large.
The accuracy of the calculated SIFs is verified apply-
ing a simple approximation procedure, using an existing
solution for SIFs in the case of 2 unequal cracks in an
infinite plate subjected to remote uniform stress. The
solution for this configuration was used to obtain the
interaction effect coefficients which take into account
the increase of SIFs of analysed crack tip because of the
FIGURE 39 Estimated number of cycles vs crack length for CCT interaction with an existing adjacent crack tip. The com-
under tension [Colour figure can be viewed at wileyonlinelibrary. parison of the results has shown very good agreement
com] between solutions (Figure 42), verifying that XFEM can
SEDMAK 25

FIGURE 40 Analysed configuration with multiple cracks: (A) 3 holes186 and (B) 11 holes183

FIGURE 41 FEM model of the MSD


panel after 50 steps of crack
propagation186 [Colour figure can be
viewed at wileyonlinelibrary.com]

be successfully used for SIF calculations and fatigue life the solutions for cracks 1, 2, 11, 12, 21, and 22 only. These
predictions of MSD in aircraft structures. 6 cracks were selected because of their unique positions.
In the second case, plate with 11 holes and 22 crack That is, at these positions, the influences of adjacent
tips,182,183 the FE model of a panel (dimensions crack interaction are either minimal (at the 1st and the
L1 × L2 = 609.6 × 863.6 mm, thickness = 1.6 mm, 11th hole; cracks 1, 2, 21, and 22 respectively) or maximal
Figure 43, with 11 fastener holes (radii r = 3.23 mm at (the 5th hole; cracks 11 and 12).
distances b = 25.4 mm) is subjected to different levels of The maximum SIF values are obtained for cracks
uniform uniaxial tensile stresses (50, 100, and 200 MPa). emanating from the fifth hole, whereas hole nos. 1 and
All other details, including material properties, loading, 11 have the minimum ones (Figure 55). This becomes
and geometry, are given in Aldarwish et al.182,183 Results more obvious with crack growth, for all 3 load cases. This
are given in Figure 43 (step 40), as an example, where can be explained by the fact that the crack interaction is
more detailed presentation of results is given in stronger for the cracks at middle holes, than for the
Aldarwish et al.183 cracks at “edge” holes.
As mentioned before, SIFs were calculated for each To verify results obtained by XFEM, SIFs calculated
crack front and different crack sizes. Maximum values in Morfeo/Crack for ABAQUS are compared with results
of SIFs calculated along the crack fronts were used as a obtained using classical FEM (FRANC software192) and
reference. The SIF results shown in Figure 44 represent approximate method, as shown in Aldarwish et al.183

(A) (B)

FIGURE 42 Normalised SIFs: (A) crack tip 1 and (B) crack tip 4 [Colour figure can be viewed at wileyonlinelibrary.com]
26 SEDMAK

FIGURE 43 XFEM model of MSD


panel after 40 steps of cracks propagations
(σ = 200 MPa)182 [Colour figure can be
viewed at wileyonlinelibrary.com]

The results are presented through normalised stress get significant improvements in riveting technology
KI regarding the extension of the fuselage fatigue life. Devel-
intensity factors (geometry factor β ¼ pffiffiffiffiffiffi) for 2
σ πa opment of new welding processes, like laser beam (LBW)
selected cracks, nos. 11 and 12, ie, for the maximal influ- and friction stir welding (FSW), enabled production of
ences of cracks interaction Figure 45. integral skin‐stringer structures, but despite its numerous
Comparing the results obtained by XFEM (Morfeo/ advantages, welded joints are prone to cracking. Two
Crack for ABAQUS) and FEM (FRANC2D/L), it can be types of cracks, most frequently associated with the struc-
seen that SIF maximum differences occur for initial crack tural integrity of the fuselage, are longitudinal cracks
size, because of the proximity of the hole. For cracks at under hoop stresses and circumferential cracks under
fifth hole, this difference starts to decrease with crack stresses from vertical bending. Therefore, here we present
growth to 0.5% (crack 11), ie, 1% (crack 12), but with fur- results obtained by using XFEM for a complex 3D geom-
ther crack growth, it increases up to 6.3% (crack 11), ie, etry,180,183 compared with experimental investigation for
8.1% (crack 12). Increasing differences with crack growth verification.
may be explained by the different manner in which these The first step in numerical modelling was to test
2 methods implement mutual crack interaction. The dif- XFEM by making FE model of base metal plate with ini-
ference can probably be reduced by increasing the num- tial crack, simulating the real loads from experiment.180
ber of nodes in the crack regions in 3D model. Base metal plate was chosen as it had simple geometry,
Comparing the XFEM results with the results of approxi- and because the calculated values of SIF could be verified
mate method, better agreement can be seen, leading to using other methods. The second step was FE modelling
the conclusion that XFEM also takes crack interaction of 4‐stringer panel and determination of the number of
effect into consideration in a very good manner. load cycles that would grow crack to critical length that
will then be compared to number of cycles obtained in
experiment with real 4‐stringer structure. In both simula-
4.3 | Fatigue crack growth in integral tions, aluminium alloy AA6156 T6 was used (Young
skin‐stringer panels modulus E = 71 GPa, Poisson ratio ν = 0.33). The loads
used in simulations were equal to average values of max-
Fatigue crack initiation and growth, as well as fracture
imum tensile forces over time measured in experiments
resistance and corrosion issues associated with riveted
(obtained because of the courtesy of researchers from
structures, are well understood, and it seems difficult to
GKSS Research Center, Geesthacht180). For the base
metal plate, average maximum force was
F max = 112.954 kN, while the load ratio R = 0.146 was
determined on the basis of average minimum tensile
force measured. Coefficients for Paris equations were
obtained by tests with base metal plates184: m = 3.174
and C = 1.77195E−012 MPa mm1/2. Initial crack in the
first simulation was propagated to length 2a = 275 mm.
Figure 46 shows that in XFEM simulation, number of
cycles to critical crack length is less than that obtained
in experiment (169 076 cycles vs 189 514 cycles); however,
up to crack length 2a = 60 mm (almost linear growth),
the numbers of cycles differ insignificantly. This confirms
conclusion that in case of simple geometry, XFEM is a
FIGURE 44 SIF histories for selected cracks for σ = 200 MPa182 fairly reliable method for crack growth rate determina-
[Colour figure can be viewed at wileyonlinelibrary.com] tion, providing conservative values.
SEDMAK 27

FIGURE 45 Comparison of SIF values obtained by 3 different methods (σ = 200 MPa)183 [Colour figure can be viewed at
wileyonlinelibrary.com]

After successful numerical simulation of crack growth concentrator on the spar bottom cap, which can lead
in base metal plate, more complex geometry of 4‐stringer to catastrophic failure if not detected during service
plate has been analysed.184 The central crack of the and repaired. The most common used material in light
length a0 = 14 mm was initiated, and the load identical aircrafts, aluminium alloy 2024‐T3, has been tested
to that used for base metal plate was applied. The crack extensively in the form of standard specimens under
propagation was simulated in 173 steps, each being constant amplitude or spectral loading.185,188 However,
2 mm; after 68 steps, it reached the wall of the left complex geometry of a spar highly influences fatigue
stringer and started growing along it. At the same time, life, making predictions, based on specimens, obsolete.
crack continued to spread through the base metal plate, Anyhow, even the smallest change in the geometry or
reaching the wall of the right stringer after 78th step material necessarily would require new experiment,
(Figure 47A) and started growing along that stringer. making design process too expensive. Perfect alternative
During the 160th step, complete failure of the left stringer is numerical modelling, because any change is easy to
occurred (Figure 47B), after which the crack continued to make, providing new estimations of fatigue life quickly
spread along the right stringer and through the base and at low costs. Nevertheless, numerical model has to
metal plate. be verified experimentally, to have any relevance for real
application.
Therefore, here the XFEM for crack growth simula-
4.4 | Integral (welded) vs riveted joints tion and fatigue life estimation of the integral AA2024‐
T3 wing spar was performed to evaluate possible replace-
The main load‐carrying member of the wing is the spar ment of differential spar of light aircraft UTVA 75 (shown
(Figure 48A), usually composed of thin shear panel in Figure 48). Toward this end, numerical models of both
(web) and flanges (caps) at the top and bottom AA2024‐T3 integral and differential spars were developed
(Figure 48B). Under service loading, a fatigue crack and number of loading cycles (to propagate crack to a
may initiate and grow from the most severe stress given length) was compared with the number of cycles
obtained by testing the differential spar, using the equip-
ment shown in Figure 49, as explained in detail in
Petrašinović et al.185
During the experiment with narrow band random
loading, 2 cracks appeared. Load used (minimum value
+391.2 N, maximum value +2028.0 N, frequency
12.5 Hz) produced strong tensile stress in caps; conse-
quently, visible crack appeared on the left cap below the
support, after 8542 cycles. The crack began to spread rap-
idly towards the spar web, then changed its direction and
continued along the cap at an angle of 90° with respect to
the original direction. After 39 450 cycles, another crack
FIGURE 46 Numbers of cycles obtained in experiment and was spotted on the right spar cap, later determined to
XFEM simulation (base metal T6)180 [Colour figure can be viewed have occurred on the fastener hole. Cracks then contin-
at wileyonlinelibrary.com] ued to grow below the strengthening washers but were
28 SEDMAK

(A) (B)

FIGURE 47 Crack growth (A) after 78 steps and (B) after 160 steps184 [Colour figure can be viewed at wileyonlinelibrary.com]

(A) (B)

FIGURE 48 A, Wing root assembly of light aircraft UTVA 75. B, Spar and supporting elements [Colour figure can be viewed at
wileyonlinelibrary.com]

not visible. After 58 520 cycles, the test was stopped, and Figure 51A, whereas the crack growth after 45th step is
the spar was examined to determine what was going on shown in Figure 51B. Number of cycles for this simula-
below the strengthening washers. Damaged spar is shown tion, 58 687, was extraordinary close to the experimental
in Figure 50, not indicating any visible damage on the one, 58 520. Anyhow, before running into conclusion,
spar web. one should pay attention to the fact that crack growth
The experiment with riveted joints was modelled by path was different, ie, numerical simulation predicted dif-
the XFEM, as explained in detail in Petrašinović et al.185 ferent change of direction of the initial crack, compare
The finite element mesh with an edge crack is shown in Figures 50A and 51B). Nevertheless, this is still valuable

(A) (B)

FIGURE 49 Fatigue testing system: (A) photo and (B) schematic presentation185 [Colour figure can be viewed at wileyonlinelibrary.com]
SEDMAK 29

(A) (B)

FIGURE 50 Damaged spar after experiment: (A) general view and (B) details188 [Colour figure can be viewed at wileyonlinelibrary.com]

(A) (B)

FIGURE 51 A, FE mesh with the initial crack. B, Crack after a 45th step (view from below)185 [Colour figure can be viewed at
wileyonlinelibrary.com]

result, proving great potential of numerical simulation of C3D8R) and an average element size of 0.5 mm (average
3D crack growth. element size around the crack tip was 0.2 mm). Initial
To simulate fatigue crack propagation in the integral crack lengths were 1 mm, and cracks did not penetrate
spar, the FE model with dimensions equivalent to the dif- the entire thickness of the spar. The value of displace-
ferential spar was made (Figure 52). It was decided to ment used in the simulation was 3 mm (Figure 52).
simulate growth of 2 penny‐shaped cracks in the left After opening, cracks were propagated in steps of
and right edge of spar cap (Figure 52) next to the approximately 1 mm. Growth was not restricted to a sin-
constraining zone, where cracks in the experiments gle plane; instead, the crack could grow in the direction
always appeared. Finite element mesh was significantly defined by kink angle, calculated within Morfeo/Crack
refined in the areas where cracks were expected to spread for ABAQUS. After 22 steps of propagation, the first crack
for getting more precise values of SIFs at nodes on crack reached the vertical wall of the integral spar (Figure 53B).
fronts. The mesh consisted of 278 616 3‐dimensional 8‐ Its path was not straight, but curved, like in the experi-
node linear brick elements with reduced integration (type ment with differential spar, proving that the numerical

FIGURE 52 Integral spar geometry and FE mesh with initial penny‐shaped cracks (blue circles)188 [Colour figure can be viewed at
wileyonlinelibrary.com]
30 SEDMAK

(A) (B)

FIGURE 53 Cracks after the 8th step (A) and 1st crack after 22nd step of propagation (B)188 [Colour figure can be viewed at
wileyonlinelibrary.com]

model simulated spar behaviour adequately. The second The number of output values for each propagation step
crack grew until eighth step (Figure 53A), when it was lit- might be large and depends on the number of points on
erally closed by the second crack, ie, spar deformation. the crack front, which, again, results from the density of
One should notice that if crack growth is restricted to a the FE mesh in propagation areas; therefore, values
single plane, both grow at the same rate and second crack obtained during simulation had to be processed. Equiva-
does not stop, but this situation is supposed to be less lent SIF values obtained for the first crack, along with
realistic. estimated number of cycles for some characteristic steps,
First crack then split, and 2 new fronts continued to are shown in Table 4. Number of cycles was calculated
propagate simultaneously in the horizontal wall of the using modified Paris‐Erdogan law (the stress ratio
spar (first front) and the vertical wall of the spar (second R = 0.15 used in calculations was identical to ratio kept
front). After the 35th step of propagation (counting from in experiment with differential spar).
the crack opening), the simulation stopped because the Table 4 indicates that the value of equivalent SIF
1st front left the area with refined mesh (Figure 54A) increases until the 22nd step, with the mean value
and Morfeo/Crack for ABAQUS could not proceed with Keq=792.263 MPa mm0.5. After that step, it starts to
calculations (as an order of magnitude, it is recom- decline, which coincides with the first crack reaching
mended to generate elements 5 to 10 times smaller than the vertical wall of the spar. The equivalent SIF values
the characteristic lengths of the initial crack both in the continue to decline until the 27th step when crack leaves
initial crack area and in the area where it is expected to the area between horizontal and vertical wall (which is
propagate). Position of the second front after 35 steps of thicker than other areas), and then starts to grow again
propagation is given in Figure 54B. until the last step. The number of cycles obtained for first
Stress intensity factor modes I, II, and III were calcu- crack on integral spar (Table 4 and Figure 55) shows that F55
lated for each step; then, the equivalent stress intensity the initial crack (of length 1 mm) will extend to 2 mm
factor Keq was calculated for each node on crack fronts. after approximately 198 000 cycles of applied

(A) (B)

FIGURE 54 A, First front of the crack. B, Second front of the crack after 35 steps of propagation188 [Colour figure can be viewed at
wileyonlinelibrary.com]
SEDMAK 31

TABLE 4 Values of equivalent SIFs obtained by XFEM simulation of crack growth in integral spar

Equivalent SIF (Keq) (MPa mm>0,5)


Number of Cycles
Step Max. Value Min. Value Mean value for Each Step

1 93.826 88.801 91.719 0


11 449.562 448.396 449.129 2601.38
21 782.222 771.832 779.340 401.018
22 793.353 789.838 792.263 382.973
29 823.883 653.519 688.550 627.077
30 724.467 686.779 706.439 502.862
31 733.845 716.974 726.257 483.992
32 786.265 754.693 766.747 363.528
33 839.233 800.253 816.207 295.627
34 908.389 869.451 883.331 282.955
35 920.845 906.537 914.249 235.858
36 964.471 944.922 951.777 209.427

The main goal of presented analysis was to compare


residual life of damaged integral spar with the life of dam-
aged differential spar obtained in experiment. Two cracks
propagated on spar caps riveted to spar web until com-
plete failure, while web remained undamaged during test.
Contrary to the experiment, one crack on FE model of
integral spar stopped after few steps of propagation,
whilst other continued to grow and produced consider-
able damage on vertical wall (which has the same func-
tion as web of differential spar). Comparison of
numerical results and experimental values revealed that
FIGURE 55 Estimated number of cycles vs crack length for significant increase in fatigue life can be expected, justify-
integral spar [Colour figure can be viewed at wileyonlinelibrary. ing the idea of replacing the old concept with new one.
com]

5 | C ON C L U D I NG R E M AR K S
displacement, while in the experimentally tested differen-
tial spar, the initial crack extended to 2 mm after approx- It was indeed a long way from early efforts in solving rel-
imately 27 000 cycles (for the same applied displacement). atively simple problems of 2D cracked solids using man-
Also, it grew up to 3 mm after another 52 230 cycles com- ual input data to feed computers with kilobytes of
pared to just 5850 cycles in differential spar. The crack memory, to nowadays sofisticated automatic mesh gener-
entered the area between the horizontal and vertical wall ators capable of modelling 3D geometries, being as com-
of the spar after 341 251 cycles compared to 45 000 cycles plex as real solids, running on computers with
in differential spar. Finally, propagation ended after terabytes... and still growing! Anyhow, one should keep
345, 795 cycles compared to 50 743 cycles obtained in in mind that it takes more than just software and hard-
experiment. It must be noted that final crack lengths were ware to simulate real life. We still struggle with correct
not the same: Crack in experiment riches 45 mm, while prediction of crack growth path in 3D structures, just to
length of crack in simulation is 36 mm; nevertheless, mention one of many problems yet to be solved.
fatigue life of integral spar is obviously much longer. Esti- An attempt was made here to make an overview of
mated number of cycles for integral spar is approximately CFM not only from the task posed by the title itself, ie,
7 times greater than for differential spar under the same from early effort to recent achievements but also as a
applied displacement. kind of cross section through arbitrary chosen topics,
32 SEDMAK

covering mainly elastic‐plastic crack analysis and growth, 9. Barsoum RS. Triangular quarterpoint elements as elastic and
as well as fatigue crack growth, but leaving many other perfectly‐plastic crack tip elements. Int J Numer Methods Eng.
important topics untouched because of limited space 1977;11(1):85‐98.

and time. Let me just mention emerging topics, like bio- 10. Henshell RD, Shaw KG. Crack tip finite elements are unneces-
materials, which deserve equal treatment, especially if sary. Int J Numer Methods Eng. 1975;9(3):495‐507.
one considers their importance. 11. Shih CF, de Lorenzi HG, German MD. Crack extension model-
Finally, let me try to answer to the question, naturally ing with singular quadratic isoparametric elements. Int J Fract.
arising at the end, ie, “what has been achieved”? From 1976;12(4):647‐651.

the point of view of the time when early attempts were 12. Fried I. Discretization and round‐off error in the finite element
made, simple answer would be “fantastic, almost unbe- analysis of elliptic boundary value and eigenvalue problems,
Ph.D. thesis, Massachusetts Institute of Technology (June 1971)
lievable results have been achieved”, whereas from the
point of view of future developments, it could be that this 13. Rice JR, Johnson MA. The role of large crack tip geometry
is just a beginning. Anyhow, let us state and underline changes in plane strain fracture. In: Kanninen MF, Adler W,
Rosenfield A, Jaffe R, eds. Elastic behaviour of solids. McGraw
the main achievements, at least for topics covered in this
Hill; 1977.
paper:
14. Banks‐Sills L, Bortman Y. Reappraisal of the quarter‐point
quadrilateral element in linear elastic fracture mechanics. Int
• Modelling of singularity, both in linear elastic and
J Fracture. 1984;25(3):169‐180.
elastic‐plastic FM problems.
15. Irons BM. A frontal solution program for finite element analy-
• Successful simulation of crack growth in plasticity,
sis. Int J Numer Methods Eng. 1970;2(1):5‐32.
especially micromechanical modelling, based on
16. Luxmoore AR, Owen DRJ. Proc. 1st Int. Conf., Numerical
material microstructure, including heterogeneous
Methods in Fracture Mechanics, Pineridge Press, Swansea, 1978
materials like welded joints.
• Modelling of 3D crack problems, in relatively simple 17. Luxmoore AR, Owen DRJ. Proc. 2nd Int. Conf., Numerical
Methods in Fracture Mechanics, Pineridge Press, Swansea,
manner, by using sophisticated, commercially avail-
1980
able pre‐processors.
18. Yuan H, Brocks W. On the J‐integral concept for elastic‐plastic
• Reasonably successful simulation of fatigue crack
crack extension. Nucl Eng Design. 1991;131(2):157‐173.
growth in complex 3D welded structures, enabling
reliable prediction of fatigue life. 19. Brocks W, Olschewski J. On J‐dominance of crack‐tip fields in
largely yielded 3D structures. Int J Solids Structures.
1986;22(7):693‐708.
20. Brocks W, Noack HD. J‐integral and stresses at an inner surface
R EF E RE N C E S flaw in a pressure vessel. Int J Pres Ves Piping. 1988;31(3):187‐203.

1. Argyris JH. Energy Theorems and Structural Analysis, a series 21. European Structural Integrity Society, ESIS TC8 the recom-
of articles on in Aircraft Engineering, Oct., Nov., 1954; Feb., mendations for use of FEM in fracture mechanics, Newsletter
March, April, May, 1955. Nr 15, 1991
22. Brocks W.. Plasticity and fracture series: solid mechanics and
2. Turner MJ, Clough RW, Martin HC, Topp LC. Stiffness and
its applications, Vol. 244
deflection analysis of complex structures. J Aeronaut Sci.
1956;23(9):805‐823, 854. 23. Sedmak, S. The effects of notches and cracks on fracture with
elastic and plastic deformation (in Serbian), D.Sc. thesis, Uni-
3. Oden JT. Finite Elements of Non‐linear Continua. McGraw‐ versity of Belgrade, 1976
Hill; 1972.
24. Berkovic M. Determination of stress intensity factors using
4. Zienkiewiitz O. The Finite Element Method in Engineering Sci- finite element method, Structural Integrity and Life 4, 2004,
ence. McGraw‐Hill; 1971. pp. 57–62, based on lecture held in Int. Fracture Mechanics
Summer School, Smederevska Palanka, 1980
5. Chan SK, Tuba IS, Wilson WK. On the finite element method
m fracture mechanics. Eng Fracture Mech. 1970;2(1):1‐17. 25. Sedmak S, Berković M. Jarić, J., Fracture mechanics problems
(in Serbian), 15. Congress of Rational and Applied Mechanics,
6. Tracey DM. Finite elements for determination of crack tip elas-
Kupari, 1981.
tic stress factors. Eng Fracture Mech. 1971;3(3):255‐265.
26. Sedmak A, Berković M. Rapid mesh refinement as a simple
7. Levy N, Marcal PV, Ostergren WJ, Rice JR. Small scale yielding technique for SIF evaluation, Proc. 3rd Int. Conf., Numerical
near a crack in plane strain: a finite element analysis. Int J Methods in Fracture Mechanics, Pineridge Press, Swansea,
Fracture. 1971;7:143‐156. 1984
8. Rice JR, Tracey DM. In: Fenves SJ et al., eds. Computational 27. Sedmak A. Finite element evaluation of FM parameters using
Fracture Mechanics, in Numerical and Computer Methods in rapid mesh refinement, proc. Int. Conf. Fracture 6, Pergamon
Structural Mechanics. N.Y: Academic Press; 1973:585‐623. press, New Delhi, 1984, Vol. 2, pp. 1095–1101
SEDMAK 33

28. Pineau A. Review of fracture micromechanisms and a local 47. Melenk JM, Babuska I. The partition of unity finite element
approach to predicting crack resistance in low strength steels method: basic theory and applications. Comp Meth App Mech
advances in fracture research, Proc. 5th Intl. Conf. on Fracture, Eng. 1996;39:289‐314.
Cannes, Ed. D Francois et al, Pergamon (1981) pp 553–580
48. Jovičić G, Živković M, Sedmak A, Jovičić N, Milovanović D.
29. Beremin FM. A local criterion for cleavage fracture of a nuclear Improvement of algorithm for numerical crack modeling.
pressure vessel. Met Trans. 1983;14a:2277‐2287. Archives Civil Mech Eng. 2010;10(3):19‐35.

30. Beremin FM. Experimental and numerical study of the differ- 49. Nichols RW. The use of fracture mechanics as an engineering
ent stages in ductile rupture: application to crack initiation tool, Proc. Int. Conf. Fracture 6, Pergamon Press, New Delhi,
and stable crack growth. In: Nemat‐Nasser S, ed. Three‐dimen- Fracture 84, 1984, pp. 3717–3749
sional constitutive relations and ductile fracture. North‐Holland 50. Kuna M. Finite Elements in Fracture Mechanics, Theory‐
Publ; 1981:185‐205. Numerics‐Applications. Springer; 2013.
31. Mudry F. A local approach to fracture. Nucl Engn and Design. 51. Brocks W, Cornec A, Scheider I, Computational aspects of
1987;105(1):65‐76. nonlinear fracture mechanics, Technical Note GKSS/WMS/
32. Rice JR, Tracey DM. On the ductile enlargement of voids in tri- 02/05, 2002
axial stress fields. J Mech Phys Solids. 1969;17(3):201‐217. 52. McMeeking RM, Rice JR. Finite‐element formulations for
33. European Structural Integrity Society, ESIS Publication P6–98: problems of large elastic‐plastic formulations. Int J Solids
Procedure to measure and calculate material parameters for Struct. 1975;11:601‐616.
the local approach to fracture using notched tensile specimens, 53. Rice JR. The mechanics of crack tip deformation and extension
1998 by fatigue, in fatigue crack propagation, Special Technical Pub-
34. European Structural Integrity Society, ESIS Publication—Draft lication 415, ASTM, Philadelphia, 1967, pp. 247–311.
P9‐02D: Guidance on local approach of rupture of metallic 54. Berković M. Membrane finite elements (in Serbian), D.Sc. the-
materials, 2002 sis, University of Belgrade, 1978
35. Gurson AL. Continuum theory of ductile rupture by void nucle- 55. Berković M. General mеmbгanе isoparametric elements, I
ation and growth: part I—yield criteria and flow rules for World congress оn finite element methods in structural
porous ductile media. J Engng Mater Technol. 1977;99(1):2‐15. mесhanics, Bournemouth, 1975
36. Tvergaard V, Needleman A. Analysis of the cup‐cone fracture 56. Berkovic M. Numerical methods in fracture mechanics. Struc-
in a round tensile bar. Acta Metall. 1984;32(1):157‐169. tural Integrity Life. 2004;4:63‐66.

37. Zhang ZL, Thaulow C, Ødegard J. A complete gurson model 57. Berkovic M. Problems of plane and triaxial stress states in pres-
approach for ductile fracture. Eng Fract Mech. 2000;67:155‐168. sure vessels and pipelines. Structural Integrity Life.
2004;4:67‐74.
38. Euromech ‐ Mecamat. In: Besson J, Moinerau D, Steglich D,
eds. 9th European Mechanics of Materials Conference, Local 58. Berkovic M, Sedmak A. The application of finite element
Approach to Fracture. Moret‐Sur‐loing; 2006:9‐12 May 2006. method in calculation of thin shell J integral. Structural Integ-
rity Life. 2004;4:85‐90.
39. ABAQUS. User's Manual, version 6.12. RI, USA: Dassault
Systèmes Simulia Corp, providence; 2014. 59. Sedmak A, Berkovic M, Savovic N. Numerical analysis of sur-
face crack problems in pressure vessels. Structural Integrity
40. www.ansys.com Life. 2004;4:91‐100.
41. Belytschko T, Lu YY, Gu L. Element‐free Galerkin methods. 60. Hellen TK. On the method of virtual crack extension. Int J
Int J Numer Methods Eng. 1994;37(2):229‐256. Numer Methods Eng. 1975;9(1):187‐207.
42. Belytschko T, Gu L, Lu YY. Fracture and crack growth by ele- 61. Parks DM. The virtual crack extension method for nonlinear
ment free Galerkin methods. Model Simul Mater Sci. Eng. material behavior. Comput Methods Appl Mech Eng.
1994;2(3A):519‐534. 1977;12(3):353‐364.
43. Lu YY, Belytschko T, Gu L. A new implementation of the ele- 62. Siegele D, Schmitt W. Determination and simulation of stable
ment free Galerkin method. Comput Methods Appl Mech Eng. crack growth in ADINA. Comput Struct. 1983;17(5‐6):697‐703.
1994;113(3‐4):397‐414.
63. Brocks W, Eberle A, Fricke W, Veith H. Large stable crack
44. Moes N, Dolbow J, Belytschko T. A finite element method for growth in fracture mechanics specimens. Nucl Eng Design.
crack growth without remeshing. Int J Numer Methods Eng. 1994;151(2‐3):387‐400.
1999;46(1):131‐150.
64. Brocks W, Yuan H. Numerical investigations on the signifi-
45. Belytschko T, Black T. Elastic crack growth in finite elements cance of J for large stable crack growth. Eng Fract Mech.
with minimal remeshing. Int J Numer Methods Eng. 1989;32(3):459‐468.
1999;45(5):601‐620.
65. Gullerud AS, Dodds RH, Hampton RW, Dawicke DS. Three‐
46. Sukumar N, Moës N, Moran B, Belytschko T. Extended finite dimensional modeling of ductile crack growth in thin sheet
element method for three‐dimensional crack modeling. Int J metals, computational aspects and validation. Eng Fract Mech.
Numer Methods Eng. 2000;48(11):1549‐1570. 1999;63(4):347‐374.
34 SEDMAK

66. Needleman A. An analysis of tensile decohesion along an inter- 84. Kikuchi M, Miyamoto H. Evaluation of Jk integrals for a crack
face. J Mech Phys Solids. 1990;38(3):289‐324. in multiphase materials, Recent Research on Mechanical
67. Needleman A. An analysis of decohesion along an imperfect Behavior of Materials, Bulletin of Fracture Mechanics Labora-
interface. Int J Fract. 1990;42(1):21‐40. tory, Vol. 1, Science University of Tokyo, 1982.
85. Kikuchi M, Miyamoto H. Sakaguchi Y, Evaluation of three‐
68. Lin G, Kim Y‐J, Cornec A, Schwalbe K‐H. Fracture toughness of
dimensional J‐integral of semi‐elliptical surface crack in pres-
a constrained metal layer. Comput Mater Sci. 1997;9(1‐2):36‐47.
sure vessel, in “Trans. 5th Int. Conf. Structural Mechanics in
69. Siegmund Th, Bernauer GW, Brocks W.. Two models of ductile Reactor Technology (5th SMiRT),” paper G7/2, Berlin, 1979.
fracture in contest, porous metal plasticity and cohesive ele-
86. Gurtin ME. On the path‐independent integral for
ments, Proc. ECF 12, Fracture from Defects, Engineering
elastodynamics. Int J Fracture. 1976;12(4):643‐644.
Materials Advisory Services Ltd, Ort, 1998, pp. 933–938.
87. de Lorenzi HG. On the energy release rate and the J‐integral
70. Needleman A, Tvergaard V. An analysis of ductile rupture at a
for 3D crack configurations. Int J Fract. 1982;19:183‐193.
crack tip. J Mech Phys Solids. 1987;35(2):151‐183.
88. de Lorenzi HG. Energy release rate calculations by the finite
71. Rousselier G, JC Devaux, G Mottet, Devesa G. A methodology
element method, General Electric Technical Infomation Series,
for ductile fracture analysis based on damage mechanics, An
Report No. 82CRD205, 1982
illustration of a local approach of fracture, Nonlinear Fracture
Mechanics, Volume II ‐ Elastic‐Plastic Fracture, ASTM STP 89. Bui H, Jarić J, Radenković D.. Loi de Conservation en
995, Philadelphia, 1989, pp. 332–354. Termoelasticite Lineare, Ecole Polytechnique, Lab. de
Mecanique des Solides, Rapport interne No 3, 1977
72. Sun D‐Z, Siegele D, Voss B, Schmitt W. Application of local
damage models to the numerical analysis of ductile rupture. 90. Sedmak A, Pavišić M. Conservation law of J integral type for
Fatigue Fract Eng Mater Struct. 1988;12:201‐212. non‐stationary time dependent fracture mechanics. Int J Frac-
ture. 1994/95;69(2):R41‐R43.
73. Brocks, W., D Klingbeil, G Künecke, DZ Sun. Application of
the Gurson model to ductile tearing resistance, Second Sympo- 91. Vukobrat M, Sedmak A. Conservation law of J integral type for
sium on Constraint Effects, ASTM STP 1224, 1995, pp. multilayered shells, ECF9, Reliability and structural integrity
232–252. of advanced materials, EMAS, Warley, U. K, 1992, p. 865

74. Xia L, Shih FC. Ductile crack growth—I. a numerical study 92. Sedmak A.. Berković M, Jarić J.. Finite element method of thin
using computational cells with microstructurally‐based length shell J integral evaluation, Proc. of ICF7, Houston, Pergamon
scales. J Mech Phys Solids. 1995;43:223‐259. Press, 1989

75. Xia L, Shih C, Hutchinson FW. A computational approach to 93. Sedmak A, Berković M, Jarić J. J integral for thin shells, Defect
ductile crack growth under large scale yielding. J Mech Phys Assessment in Components—Fundamentals and Applications,
Solids. 1995;43(3):389‐413. in ESIS/EGF9. London: MEP; 1991:45‐53.
94. Jarić J, Sedmak A, Berković M. On the problem of path depen-
76. Schmitt W, Sun D‐Z, Blauel JG. Damage mechanics analysis
dency of J integral for thin shells, Proc. of ECF7, Budapest,
(Gurson model) and experimental verification of the behaviour
EMAS, 1988
of a crack in a weld‐cladded component. Nucl Eng Des.
1997;174(3):237‐246. 95. Jovicic G, Zivkovic M., Kojic M. Numerical procedure for cal-
culation of stress intensity factors and its use for life
77. Bernauer G, Brocks W. Micro‐mechanical modeling of ductile
assessment of steam turbine housing of thermal power plant,
damage and tearing ‐ results of a European numerical round
Proc. Int. Fract. Mech. Summer School 8, Belgrade, 2003, p.
robin. Fatigue Fract Eng Mater Struct. 2002;25(4):363‐384.
321–332
78. Smelser RE, Gurtin M. On the J integral for bi‐material body.
96. Azari Z, Casavola C, Pappalettere C, Pruncu CI. Numerical
Int J of Fracture. 1977;13:382‐384.
simulation in coated materials: model of crack propagation
79. Bleackley MH, Jones RD, Luxmoore AR. Path dependency of bi‐material. Structural Integrity Life. 2012;12:125‐129.
the Rice J integral in weld geometries. In: Fracture Control of
97. Gubeljak N, Predan J, Kozak D. Leak‐before‐break analysis of
Engineering Structures, Proceedings, European Conference on
a pressurizer—estimation of the elastic‐plastic semi‐elliptical
Fracture 6. Amsterdam, Vol. I: EMAS; 1986:643‐654.
through‐wall crack opening displacement. Structural Integrity
80. Sedmak, A. The role of weldment interfaces in fracture Life. 2012;12:31‐34.
mechanics parameters evaluation, Proc. of Int. Conf. Fracture
98. Zerbst U, Ainsworth RA, Beier H‐TH, et al. Review on fracture
9, Vol. 5, p. 2345–2356, Sydney, Pergamon Press, 1997
and crack propagation in weldments – A fracture mechanics
81. Sedmak A. Sedmak, S., Weldment fracture mechanics parame- perspective. Eng Fract Mech. 2014;132:200‐276.
ters evaluation, Proc. of XXII Congress of TAM, Vrnjačka
99. Adžiev G, Sedmak A, Adžiev T. Numerical analysis of tensile
Banja, pp.73–83, 1997
specimen fracture with crack in HAZ. Structural Integrity Life.
82. Savović N, Sedmak A. Numerical simulation of weldment het- 2008;8:107‐113.
erogeneity. Zavarivač. 1994;39:185‐190. 100. Doncheva E, Medjo B, Sedmak A. Finite element analysis of
83. Sedmak S, Adžiev T, Sedmak A, Gočev J. Experimental and fracture resistance parameters for stationary semi‐elliptical sur-
numerical analysis of cracked welded joints, Proc. of ICM 7, face cracks in high strength steel. Structural Integrity Life.
Hague, 1995, pp. 291, 1995 2015;15:131‐134.
SEDMAK 35

101. Doncheva E, Medjo B, Adziev G, Sedmak S. Elastic‐plastic temperatures), Ph.D. Thesis, Faculty of Technollogy and Metal-
numerical analysis of tensile specimens with surface center‐ lurgy, Belgrade University, (in Serbian), 2004
cracked asymmetric welded x‐joints, Proc. 7th International Sci- 116. Rakin M. Application of the local approach in analysis of frac-
entific and Expert Conference TEAM 2015, Belgrade, pp. 1–5
ture of steel, M.Sc. Thesis, Faculty of Technollogy and
102. Sedmak S, Petrovski B, Sedmak A. The resistance to crack Metallurgy, Belgrade University, (in Serbian), 1996
growth of different regions of weldments in a real structure.
117. Bernauer G, Brocks W. Numerical round robin on micro‐
Int J Pres Ves & Piping. 1992;52:313‐335s.
mechanical models—results, ESIS TC8, GKSS Research Cen-
103. Sedmak S, Sedmak A, Vukomanović N. Theoretical, numerical ter, Geesthacht, 2000
and experimental analysis of cracked welded tensile panel,
118. Rakin M. The analysis of ductile fracture initiation in struc-
Proc. of ECF8 Fracture Behavior and Design of Materials and
tural steel using micromechanical models, Ph.D. Thesis,
Structures, pp. 1596–1599, Torino, 1990
Faculty of Technollogy and Metallurgy, Belgrade University,
104. Rakin M. Numerical analysis in local approach. In: Sedmak S, (in Serbian), 2003
Sedmak A, eds. Experimental and numerical methods of frac-
119. Rakin M, Cvijovic Z, Grabulov V, Putic S, Sedmak A. Predic-
ture mechanics in structural integrity assessment. Belgrade, (in
tion of ductile fracture initiation using micromechanical
Serbian): TMF‐GOSA; 1997:345‐352.
analysis. Eng Fract Mech. 2004;71(4‐6):813‐827.
105. Zrilić M, Rakin M, Sedmak A, Sedmak S. Experience in frac-
120. Besson J (Ed). Local Approach to Fracture. Paris: Les Presses
ture mechanics local approach application, proceedings of the
de l'Ecole des Mines; 2004.
6th international scientific conference on achievements in the
mechanical materials engineering ‐ AMME97, Miskolc, 71‐75, 121. Rakin M, Cvijovic Z, Grabulov V, Gubeljak N, Sedmak A. The
1997 influence of microstructure on the ductile fracture initiation in
low‐alloyed steel. Mater Sci Forum. 2004;453‐454:175‐180.
106. Zrilić M, Rakin M, Sedmak A, Sedmak S. Measurement tech-
niques in local approach to fracture, Proceedings of the 122. Needleman A, Tvergaard V. Numerical modeling of the duc-
Fourth International Symposium on Measurement Technology tile‐brittle transition. Int J Fract. 2000;101(1/2):73‐97.
and Inteligent Instruments, Miskolc, 358‐361, 1998 123. Rakin M, Cvijović Z, Grabulov V, Zrilić M, Sedmak A. Deter-
107. Sedmak A, Rakin M. Application of fracture mechanics in mination of micromechanical ductile fracture parameters of
assessment of structural integrity. In: Sedmak S, Radaković Z, pressure vessel steel, Proceeding of The 7th European Confer-
eds. From Fracture Mechanics to Structural Integrity Assess- ence on Advanced Materials and Processes ‐ Euromat,
ment. DIVK‐TMF; 2004:373‐386. Published on CD, Rimini, 2001

108. Zrilić M, Rakin M, Cvijović Z, Sedmak A, Sedmak S, Structural 124. Rakin M, Sedmak A, Cvijović Z, Zrilić M, Sedmak S.
integrity assessment by local approach to fracture, Proc. of 16th Micromechanical coupled study of crack growth initiation cri-
European Conference of Fracture: Failure Analysis of Nano terion in pressure vessel steel. Strength of Materials.
and Engineering Materials and Structures, Ed. Gdoutos, E.E., 2004;36(1):33‐36.
Alexandroupolis, Greece, 2006, published on CD by Springer 125. Zrilić M, Rakin M, Sedmak A, Aleksić R, Cvijović Z, Arsić M.
109. Zrilic M, Rakin M, Gubeljak N, Dobrojevic M, Krunich N, Ductile fracture prediction of steam pipeline steel. Mater Sci
Sedmak A. Ductile fracture modeling of advanced metallic Forum. 2006;518:537‐542.
materials. EMMC9 ‐ MECAMAT, 9th European Congress on 126. Zrilić, M., Rakin, M., Sedmak, A., Sedmak, S., Assessment of
Mechanics of Materials, facilities of EDF "Les Renardières" in‐service degradation of steam pipeline steel by local approach
Moret Sur Loing, France, 247–253, 2006 method, Proceedings of the Conference: Life Assessment and
110. Guidance on Local Approach Methods, Appendix 17 of R/H/ Management for Sructural Components Kiev, 247‐253, 2000
R6 ‐ Rev. 3, BEGL Confidental, British Energy, 1998 127. Zrilić M, Rakin M, Milović LJ, Burzić Z, Grabulov V. Experi-
111. Rakin M, Cvijovic Z, Grabulov V, Sedmak A. Crack initiation mental and numerical evaluation of a steamline behaviour
analysis in structural steel by application of micromechanical using local approach. Metalurgija. 2007;46/2:87‐92.
approach. Polish Mater Sci Eng. 2001;22:741‐744. 128. Rakin M, Gubeljak N, Dobrojević M, Sedmak A. Modeling of
112. Rakin M, Sedmak A, Cvijović Z, Zrilić M, Sedmak S. ductile fracture initiation in strength mismatched welded joint.
Micromechanical approach ‐ transferability of ductile fracture Eng Fract Mech. 2008;75(11):3499‐3510.
parameters, proceeding of the 14th European conference on 129. Dobrojević M, Rakin M, Gubeljak N, et al. Micromechanical
fracture ‐ Vol III, EMAS Publishing, Krakow, 27‐34, 2002 analysis of constraint effect on fracture initiation in strength
113. Mudry F, Di Fant M.A round robin on the measurement of mismatched welded joints. Mater Sci Forum. 2007;555:571‐576.
local criteria. Rapport Abrege N_ RE 93.319, IRSID, St. 130. Burstow MC, Howard LC, Ainsworth RA. The influence of
Germain, 1993 constraint on crack tip stress fields in strength mismatched
114. Brocks W. Numerical round robin on micromechanical models welded joints. J Mech Phys Solids. 1998;46(5):845‐872.
—results, IWM‐Bericht T 8/95, Fraunhofer Institut fuer 131. Penuelas I, Betegon C, Rodriguez C. A ductile failure model
Werkstoffmechanik (IWM), Freiburg, 1995 applied to the determination of the fracture toughness of
115. Zrilic M. The application of local approach to residual life welded joints. Numerical simulation and experimental valida-
assessment of equipment components at elevated tion. Eng Fract Mech. 2006;73(18):2756‐2773.
36 SEDMAK

132. Thomason PF. Ductile Fracture of Metals. Oxford: Pergamon 150. Zhang ZL, Niemi E. A new failure criterion for the Gurson‐
Press; 1990. Tvergaard dilational constitutive model. Int J Fract.
133. Betegón C, Peñuelas I. A constraint based parameter for quan- 1995;70(4):321‐334.
tifying the crack tip stress fields in welded joints. Eng Fract 151. Sun DZ, Kienzler R, Voss B, Schmitt W. Fracture mechanics—
Mech. 2006;73(13):1865‐1877. twenty‐second symposium. In: Atluri SN, Newman JC, Raju I
134. Younise B. Micromechanical fracture analysis of high strength Jr, Epstein JS, eds. ASTM STP 1131. Vol.II Philadelphia:
steel weldments, D.Sc. thesis, University of Belgrade, 2013 American Society for Testing and Materials; 1992:368‐378.

135. Anderson TL. Fracture Mechanics. London: CRC Press; 1995. 152. European Structural Integrity Society: ESIS procedure for
determining the fracture behaviour of materials. ESIS P2–92,
136. Sun DZ, Siegele D, Voss B, Schmitt W. Application of local
1992
damage models to the numerical analysis of ductile rupture.
Fatigue Fract Eng Mater Struct. 1989;12(3):201‐212. 153. Bilby BA, Howard IC, Li ZH. Prediction of the first spinning
cylinder test using ductile damage theory. Fatigue Fract Eng
137. Marsavina L, Nurse AD, Braescu L, Craciun EM. Stress singu-
Mater Struct. 1993;16(1):1‐20.
larity of symmetric free‐edge joints with elasto‐plastic
behaviour. Compos Mater Sci. 2012;52(1):282‐286. 154. Marini B, Mudry F, Pineau A. Experimental study of cavity
138. Marsavina L, Nurse AD. The asymptotic structure of small‐ growth in ductile rupture. Eng Fract Mech. 1985;22(6):989‐996.
scale yielding interfacial free‐edge joint and crack‐tip fields. 155. Gubeljak N, Legat J, Kocak M. Effect of fracture path on the
Acta Mech. 2007;190(1‐4):115‐131. toughness of weld metal. Int J Fract. 2002;115(4):343‐359.
139. Huang Y. Accurate dilatation rates for spherical voids in triax- 156. Las V, Ocenasek J, Vacek V. Numerical determination of J‐R
ial stress fields. Trans ASME. 1991;58(4):1084‐1086. curve using void model. Mater Technol. 2005;39:83‐88.
140. Chaouadi R, De Meester P, Vandermeulen W. Damage work as 157. Younise B, Rakin M, Medjo B, Gubeljak N, Kozak D, Sedmak
ductile criterion. Int J Fracture. 1994;66(2):155‐164. A. Numerical analysis of constraint effect on ductile tearing
141. Younise B, Rakin M, Gubeljak N, Medjo B, Sedmak A. Numer- in strength mismatched welded CCT specimens using
ical simulation for studying constraint effect on ductile fracture micromechanical approach. Technical Gazette J. 2011;18:
initiation using complete Gurson model. FME Trans J. 333‐340.
2010;38:197‐202. 158. Musraty W, Međo B, Gubeljak N, et al. Ductile fracture of pipe‐
142. Rakin M, Sedmak A, Matejic P, Zrilic M, Sedmak S. Numerical ring notched bend specimens—micromechanical analysis. Eng
simulation of ductile fracture initiation by application of Rice– Fract Mech. 2017;175:247‐261.
Tracey void growth model. In: Fuentes M, Elices M, Martin‐
159. Younise B, Rakin M, Gubeljak N, et al. Micromechanical anal-
Meizoso A, Martinez‐Esnaola JM, eds. Proceedings of the ECF
ysis of mechanical heterogeneity effect on the ductile tearing of
13 Fracture Mechanics: Applications and Challenges. San
weldments. Mater Des. 2012;37:193‐201.
Sebastian: ESIS Publication‐Elsevier Science Ltd, On CD; 2000.
160. Younise B, Rakin M, Gubeljak N, Međo B, Sedmak A. Effect of
143. Schmitt W, Keim E, Nagel G, Sun DZ. Engineering applica-
material heterogeneity and constraint conditions on ductile
tions of the local approach. In: Blauel JG, ed. Failure
fracture resistance of welded joint zones—micromechanical
assessment concepts and applications. Fraunhofer‐Institut fuer
assessment. Eng Failure Anal. 2017;82:435‐445.
Werkstoffmechanik (IWM); X‐1390; 1997.
161. Rakin M, Gubeljak N, Međo B, Sedmak A. Micromechanical
144. Tvergaard V. Influence of voids on shear band instabilities
assessment of mismatch effects on fracture of high‐strength
under plane strain conditions. Int J Fract. 1981;17(4):389‐407.
low alloyed steel welded joints. Eng Fract Mech. 2013;109:
145. Rakin M, Sedmak A, Grabulov V, Gubeljak N, Cvijović Z. 221‐235.
Determination of initial damage parameters in Gurson‐
Tvergaard‐Needleman model, 9th International Conference 162. Međo B, Rakin M, Gubeljak N, Arsić M, Šarkočević Z, Sedmak
on Mechanical Behaviour of Materials, Published on CD, A. Failure resistance of drilling ring casting pipes with an axial
Geneva, 2003 crack. Eng Failure Anal. 2015;58:429‐440.

146. Chu CC, Needleman A. Void nucleation effects in biaxially 163. Bonora N. A nonlinear CDM model for ductile failure. Eng
stretched sheets. J Eng Mater Technol. 1980;102(3):249‐256. Fract Mech. 1997;58(1):11‐28.

147. Heerens J, Hellmann D, Fracture toughness of steel in the duc- 164. Bonora N, Ruggiero A, Esposito L, Gentile D. CDM modeling
tile to brittle transition regime. Final Report of the EU‐Project of ductile failure in ferritic steels: assessment of the geometry
MAT1‐CT‐940080, GKSS Research Center, Geesthacht, 1999 transferability of model parameters. Int J Plasticity. 2006;
22(11):2015‐2047.
148. Younise B, Rakin M, Gubeljak N, Medjo B, Sedmak A. Numer-
ical simulation of constraint effect on fracture initiation in 165. Testa G, Bonora N, Gentile D, Carlucci A, Madi Y. Ductile frac-
welded specimens using a local damage model. Struct Integrity ture assessment of X65 steel using damage mechanics. Procedia
Life J. 2011;11:51‐56. Structural Integrity. 2017;3:508‐516.
149. Younise B, Rakin M, Меdjo B, Sedmak A. Local approach for 166. Tu, S., Ren, X., He, J., Zhang, Z., Numerical study on the effect
prediction of ductile fracture initiation in welded specimens. of Lüders plateau on the ductile crack growth resistance of
Welding & Material Testing J. 2011;1:31‐35. SENT specimens, submitted for Int. J. Fracture, 7.4.2018
SEDMAK 37

167. Souiyah M, Muchtar A, Alshoaibi A, Ariffin AK. Finite ele- 180. Sghayer A, Grbović A, Sedmak A, Dinulović M, Doncheva E,
ment analysis of the crack propagation for solid materials. Petrovski B. Fatigue life analysis of the integral skin‐stringer
American J Appl Sci. 2009;6(7):1396‐1402. panel using XFEM. Structural Integrity Life. 2017;17(1):7‐10.
168. Citarella R, Cricrì G. A two‐parameter model for crack growth 181. Eldwaib KA, Grbovic A, Kastratovic G. Fatigue life estimation
simulation by combined FEM–DBEM approach. Adv Eng of CCT specimen using XFEM and Paris law. Structural Integ-
Softw. 2009;40(5):363‐377. rity Life. 2017;17(2):117‐124.
182. Aldarwish M, Grbović A, Kastratović G, Sedmak A, Vidanović
169. Yang ZJ, Wang XF, Yin DS, Zhang C. A non‐matching finite
N. Numerical assessment of stress intensity factors at tips of
element‐scaled boundary finite element coupled method for
multi‐site cracks in unstiffened panel. Structural Integrity Life.
linear elastic crack propagation modeling. Comput Struct.
2017;17(1):11‐14.
2015;153:126‐136.
183. Aldarwish A, Grbović A, Kastratović G, Sedmak A, Lazić M.
170. Nicak T, Mutz A, Keim E, Meier G. Application of XFEM to
Stress intensity factors evaluation at tips of multi‐site cracks
model crack initiation and propagation during a PTS event.
in unstiffened 2024‐T3 aluminum panel using XFEM, accepted
Proceedings of the ASME 2015 Pressure Vessels and Piping
for publication in Technical Gazzete, 2018
Conference PVP2015 July 19–23, 2015, Boston, Massachusetts,
USA, 1–11. 184. Sghayer A, Grbović A, Sedmak A, Dinulović M, Grozdanovic I,
Sedmak S, Petrovski B.. Experimental and numerical analysis
171. Huffman PJ, Ferreira J, Correia J. Fatigue crack propagation of fatigue crack growth in integral skin‐stringer panels,
prediction of a pressure vessel mild steel based on a strain accepted for publication in Technical Gazzete, 2018
energy density model. Frattura Ed Integrità Strutturale.
185. Petrašinović D, Rašuo B, Petrašinović N. Extended finite ele-
2017;42:74‐84.
ment method (XFEM) applied to aircraft duralumin spar
172. Lee S, Martin D. Application of XFEM to model stationary fatigue life estimation. Tehnički vjesnik. 2012;19(3):557‐562.
crack and crack propagation for pressure containing subsea 186. Rakipovski E, Grbović A, Kastratović G, Vidanović N. Applica-
equipment. Proceedings of the ASME 2016 Pressure Vessels tion of Extended Finite Element Method for Fatigue Life
and Piping Conference PVP2016 July 17–21, 2016, Vancouver, Predictions of Multiple Site Damage in Aircraft Structure.
British Columbia, Canada, 1–7. Structural Integrity Life. 2015;15(1):3‐6.
173. Daux C, Moes N, Dolbow J, Sukumar N, Belytschko T. Arbi- 187. Kraedegh A, Sedmak A, Grbovic A, Sedmak S. Stringer effect
trary branched and intersecting cracks with the extended on fatigue crack propagation in A2024‐T351 aluminum alloy
finite element method. Int J Numer Methods Eng. welded joint. Int J Fatigue. 2017;105:276‐282.
2000;48(12):1741‐1760.
188. Eldwaib KA, Grbovic A, Sedmak A, Kastratovic G, Petrasinovic
174. Kim JH, Chau‐Dinh T, Zi G, Kong JS. The effect of compres- D. fatigue life estimation of damaged integral wing spar Using
sion stresses, stress level and stress order on fatigue crack XFEM, accepted for publication in Technical Gazzete, 2018
growth of multiple site damage. Fatigue Fract Eng Mater Struct. 189. Živojinović D, Arsić M, Sedmak A, Kirin S, Tomić R. Practical
2012;10:1‐15. aspests of fail‐safe design‐calculation of fatigue life of cracked
175. Budyn É, Zi G, Moës N, Belytschko T. A method for multiple thin‐walled structures. Technical Gazette. 2011;18(4):609‐617.
crack growth in brittle materials without remeshing. Int J 190. Cenaero. Morfeo/crack for ABAQUS, from http://cneaero.be
Numer Methods Eng. 2004;61(10):1741‐1770. 191. Fatigue Crack Growth Computer program “NASGRO”, Refer-
176. Loehnert S, Belytschko T. Crack shielding and amplification ence Manual, National Aeronautics and Space Administration,
due to multiple microcracks interacting with a macrocrack. London B. Johnson Space Center, Houston, Texsas 77058–3696,
Int J Fract. 2007;145(1):1‐8. August 2002.

177. Singh IV, Mishra BK, Bhattacharya S, Patil RU. The numerical 192. FRANC 3D, Menu & Dialog Manual, July 1998.
simulation of fatigue crack growth using extended finite ele-
ment method. Int J Fatigue. 2012;36(1):109‐119.
178. Kumar S, Singh IV, Mishra BK. A homogenized XFEM How to cite this article: Sedmak A.
approach to simulate fatigue crack growth problems. Comput Computational fracture mechanics: An overview
Struct. 2015;150:1‐22. from early efforts to recent achievements. Fatigue
179. Krаedegh A, Li W, Sedmak A, Grbovic A, Trišović N, Kirin S. Fract Eng Mater Struct. 2018;1–37. https://doi.org/
Simulation of fatigue crack growth in A2024‐T351 “T” welded 10.1111/ffe.12912
joint. Structural Integrity Life. 2017;17(1):3‐6.

Vous aimerez peut-être aussi