Engineering Failure Analysis 16 (2009) 2541–2558

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Application of the cohesive model for predicting the residual strength

of a large scale fuselage structure with a two-bay crack
A. Cornec *, W. Schönfeld, K.-H. Schwalbe, I. Scheider
Institute of Materials Research, GKSS Research Centre – Member of the Helmholtz Association, Max-Planck-Strasse 1, D-21502 Geesthacht, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The residual strength of a curved and stiffened panel containing a two-bay crack was
Received 10 October 2008 assessed using the cohesive model. This panel represents a section of a wide-body aero-
Accepted 16 October 2008 plane fuselage. The tests were conducted at IMA GmbH Dresden in cooperation with Airbus
Available online 5 November 2008
Industries Germany. The structural panel was modelled using 3D finite elements and a
layer of cohesive elements ahead of each crack tip allowing for 70 mm crack extension.
Keywords: Identification of the cohesive parameters was done on small laboratory test pieces. Special
Residual strength
effort was made for the transfer of these parameters to the structure. Reasonably conser-
Riveted fuselage panel
Ductile crack extension
vative predictions of the residual strength of the panel were achieved. The boundary con-
Finite element analysis ditions of the loading devices of the test rig are shown to have substantial influence on the
Cohesive model predictions.
Ó 2009 Published by Elsevier Ltd.

1. Introduction

The requirements for high loading capacity, weight reduction and low production cost of engineering structures, com-
bined with high structural safety, particularly in transportation industry, have to be met by new design concepts, materials,
production technologies, and assessment methods. When safety of a new airplane has to be demonstrated during the devel-
opment stage, extremely expensive test programs are required. The residual strength of the fuselage, for instance, has to sat-
isfy aerospace certification requirements. A longitudinal crack of twice the distance between two stiffeners is inserted into a
full-scale barrel section representing the fuselage design which is then subjected to internal pressure [1]. The required resid-
ual strength of that cracked structure has to be at least 1.15 times the internal cabin pressure.
In order to reduce testing costs, Airbus in Germany in cooperation with IMA1 have developed a new, cost-saving, test pro-
cedure for fuselages. Instead of a complete barrel, only a section of it is loaded by internal pressure and biaxial tension. The inner
side of the IMA panel is sealed against a chamber so that internal pressure can be built-up. Tests can be done under monoton-
ically increasing and cyclic loads [2]. Although this test rig is quite complex, it allows testing a range of design solutions with
much less efforts than the classical barrel test. Furthermore, simulations applying advanced models of fracture and damage
mechanics would reduce test efforts even more, since as soon as the finite element model of a panel has been made up param-
eter variations with respect to geometrical parameters, boundary conditions and material properties can be effectively studied.
Thin-walled light-weight structures such as the fuselage of an aeroplane may exhibit a large amount of crack extension
before final failure ensues. The ability of treating large amounts of ductile crack extension can in general be met by a number
of assessment procedures and numerical models. It is a common practice to treat ductile crack extension in the form of a
crack extension resistance curve (R-curve), with crack extension expressed as a function of either of crack tip parameters

* Corresponding author. Tel.: +49 4152 87 2535; fax: +49 4152 87 2534.
E-mail address: [email protected] (A. Cornec).
1
IMA GmbH Dresden in Germany (Materialforschung und Anwendungstechnik GmbH).

1350-6307/$ - see front matter Ó 2009 Published by Elsevier Ltd.

doi:10.1016/j.engfailanal.2008.10.014
2542 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

Nomenclature

Variables
ai half crack length
a0 initial half crack length
pi internal pressure
w0 initial width of flat bar specimen
wnec necking width of flat bar specimen during tension loading
t thickness of the skin
t0 initial thickness of flat bar specimen
tnec necking thickness of flat bar specimen during tension loading
CMOD crack mouth opening displacement
CTOD crack tip opening displacement
F force
Ji fracture toughness for crack initiation in terms of the J-integral
Rp0.2 yield strength at 0.2% plastic strain
R0 outer radius of the fuselage
SZW stretch zone width
T(d) traction-separation law (TSL)
T0N cohesive strength for normal fracture
T0S maximum cohesive traction for slant fracture
d separation
d0 critical cohesive separation
d1, d2 shape parameters of TSL
d5 crack tip opening displacement across initial crack tip with 5 mm gauge length
Da crack extension
C0N cohesive energy for normal fracture
C0S cohesive energy for slant fracture

Acronyms
C(T) compact tension specimen
M(T) middle cracked tensile panel
TSL traction-separation law

like the stress intensity factor, K, the J integral, or crack tip opening displacement (CTOD) [3]. Among CTOD the d5 crack tip
opening displacement method with a fixed gauge length at the crack tip is extensively validated for dealing with large
amounts of crack extensions [4–8] and is now standardised by ASTM [9] and ISO [10].
Besides these classical fracture mechanics methods, numerical damage mechanics models are being increasingly used.
Their advantage over the classical fracture mechanics methods is that they avoid an intrinsic problem of the latter: The
essential transferability problem inherent in fracture mechanics can be better handled by damage mechanics models.
Numerical damage models such as the cohesive model, see e.g. general overviews in [11–14] and its application to thin-
walled structures in [15–20] or porous plasticity models, e.g. [21–24], have already been extensively validated by means of
numerous tests on laboratory test pieces, e.g. [15–17,25–27]. However, only few investigations can be found on the appli-
cation to complex large scale structural components. e.g. [28]. Earlier investigations carried out on a realistic fuselage com-
ponent are based on the crack tip opening angle (CTOA) [29].
In the work described in the present paper, the cohesive model was chosen for simulating the residual strength of a
cracked large scale and stiffened structural component. This choice was motivated by the suitability of this model for ana-
lysing large amounts of crack extension, the current lack of structural applications, and by the experience of our group with
this model. In order to test the versatility of the cohesive model for structural behaviour, the results of a panel test performed
by IMA were made available to the authors. This panel is shown in Fig. 1 and will henceforth be called ‘‘IMA panel”.
The paper starts with a description of the IMA panel test, as the details of the test setup are needed for the boundary con-
ditions of the numerical model. Then the cohesive model applied in this paper is introduced in detail. A description of the
determination of the various material parameters follows. The following section deals with the finite element model of
the panel and the results of its deformation behaviour. The paper concludes with the lessons learned during this exercise.
The following step-by-step procedure was set up:

(i) The tensile properties of all materials of the panel were experimentally determined.
(ii) Fracture mechanics tests were performed on modified Kahn specimens made of the skin material taken from the
tested IMA panel. The test results – together with the tensile properties – served for determining the cohesive
parameters.
(iii) With the thus generated cohesive parameters the load carrying behaviour of the IMA panel was analysed.
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2543

Fig. 1. Inside view of the IMA fuselage panel after the test, last frame part C7 was detached after the test.

2. IMA panel test setup

2.1. Panel design

The IMA panel is a part of a wide-body fuselage with an outer diameter of 5640 mm, Fig. 2a. Geometry and dimensions of
the panel are shown in Fig. 2b. The 1.8 mm thick skin is stiffened by seven frames (C1–C7) and eight stringers (P1–P8). The
details of the riveted design near the crack tip in frame C3 are displayed in Fig. 3. The frame assembly consisted of several
components. The stringers are adhesively bonded to the skin, representing a specific feature of this fuselage design. The other
parts are connected by rivets.
The boundary conditions in the panel test approximate the behaviour of a closed fuselage. The fuselage panel was bi-axi-
ally loaded, i.e. in circumferential and longitudinal directions, each with special load transfer by hydraulic actuators at the
panel edges, which will be displayed later together with the FE model.
A two-bay crack in the skin was made by fatigue loading with a high maximum load representing the cabin pressure.
Frame C4 was manually cut prior to fatiguing the defect. At the end of fatigue crack propagation, the crack tips were located
in the rivet axis of C3 and C5. Due to the stiffness of the bay between stringer and frame the fatigue crack path was slightly
curved.

2.2. Loading of the panel

2.2.1. Internal pressure

The internal pressure, pi, was realised by a pressure chamber mounted underneath the inner side of the panel, whose out-
er edges are indicated in Fig. 2b. The crack was sealed using a poly-urethane foam.

2.2.2. Longitudinal loading2

The panel (Fig. 2b) was mounted vertically in the IMA panel test rig. At the bottom edge of the skin a centred single force
was introduced by a rigid crosshead. A number of straps were attached to the circumference of the skin, which were con-
nected to each crosshead by freely rotating bolts. One edge of the IMA panel was attached to a fixed crosshead, the opposite
edge to the crosshead mounted on a hydraulic actuator (see Fig. 11, which will be discussed in more detail below). The
stringers are smoothly tapered towards the skin edge. The longitudinal force in the panel test is the equilibrium force ana-
lytically determined for an internally pressurized closed cylinder.

2
The terms ‘‘longitudinal” and ‘‘circumferential” refer to the direction related to the fuselage.
2544 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

a IMA panel

Wide-body
fuselage
45

Pressure pi

R0 = 2820

C1 C2 C3 C4 C5 C6 C7
b
(y)
(z)
(z)
P8
P7
P6
P5
2240

P4
C3 (x) C5
P3
202.9

P2
P1
(z)
(z)
(y)
533.4
3865
(x) Two-bay crack, (y) outer edges of the IMA panel,
(z) pressure area inside

Fig. 2. IMA fuselage panel: (a) panel section as part of the wide-body complete fuselage; and (b) fuselage panel dimension with frames, stringers, biaxial
loading, pressure border, and two-bay crack.

2.2.3. Circumferential loading

The load transfer to the skin and frames in circumferential direction of the IMA fuselage panel was performed by flexible
straps at the skin edge and a pin-joint at the end of the frames, each attached to freely rotating bolts mounted on hydraulic
actuators as shown in Fig. 11. At the skin edge the hydraulic actuators were fixed during the panel test to generate reaction
forces. All other forces applied in circumferential as well as longitudinal directions applied by hydraulic actuators increase
proportionally with the internal pressure. The forces were calculated from the tangential stresses in a closed un-cracked ves-
sel with frames under internal pressure. This means, that all frames (C1–C7) were loaded with equal forces.

3. Stress–strain data

The fuselage panel consisted of components made of five materials as specified in Fig. 4. The stress–strain curves were
determined on the small flat tensile specimens, depicted in Fig. 4. Elongation was measured with a clip of 7.0 mm initial
gauge length. The tensile specimens were taken from all components of the IMA panel.
Three of the stress–strain curves belong to one group, whose individual curves are similar, namely skin, clip and lower
frame. The stringer and the upper frame profiles have significantly higher strengths. For all materials the elastic modulus
was taken as E = 68000 MPa and the Poisson’ ratio, m, was 0.3. The yield strengths are listed in Table 1.
For the skin material the true stress–strain curve was determined even beyond ultimate strength until unstable failure.
This was done using optical instrumentation acquiring digital images focused the necking region. For the determination of
the displacement in that region the tensile specimen was coated by a titanium oxide spray in order to obtain a surface with
distinctive points. The local strain, eloc, in loading direction was determined from the displacements of two points with an
initial base length, L0,loc = 0.25 mm. In addition, the area of the specimen in the necking region, Anec, was calculated based on
the width reduction, wnec, determined from the images and the observation of almost equal necking ratios for width and
thickness reduction: wnec/w0 = tnec/t0. This is valid for the used nearly square tensile bar section. For the tensile test used,
the initial width was w0 = 2.02 mm and the thickness was t0 = 1.78 mm. From this test data, the true stress and strain
was calculated by
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2545

Fig. 3. Detail of the riveted built-up parts from different aluminium alloys and the rivet arrangements. The numbers are the used notation for the different
construction parts: 1 = skin, 2 = stringer, 3 = clip, 4 = upper frame, and 5 = lower frame.

wnec
tnec ¼ t0
w0

Anec ¼ wnec tnec

DLloc
eloc ¼ ð1Þ
L0;loc

etrue ¼ ln ð1 þ eloc Þ
rtrue ¼ F=Anec

4. Cohesive parameters

4.1. Traction-separation law

Cohesive models describe various kinds of decohesion processes by its constitutive behaviour, which is a relation be-
tween surface tractions, generally having one normal and two tangential components, and the corresponding material sep-
arations. In an FE model, the cohesive surface is introduced by interface elements at the boundaries of solid elements along a
pre-defined crack path. Cohesive elements are surface elements in 3D structures and line elements in 2D structures. They do
not have an initial thickness but nevertheless an upper and a lower surface with duplicated nodes, which can separate during
loading.
Fig. 5a shows schematically the mode I fracture process in a ductile material and its idealization by a cohesive model. It is
assumed that all micro-structural mechanisms of the fracture process can be captured by two cohesive parameters, the max-
imum traction or cohesive strength, T0, and a critical separation, d0, beyond which the respective cohesive element has lost
its stress carrying capacity, and the crack extends by one element length. Alternatively to d0, the separation energy, C0, being
the integral under the T(d) curve, see Fig. 5b, can be used as a cohesive parameter, as was done in the present study.
Various functions for the traction-separation law (TSL) have been proposed in the literature. Though the cohesive model
is phenomenological, the TSL has to represent the micromechanical mechanisms of material separation. In the present anal-
ysis, a function as proposed in [30] is applied
2546 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

700
Elastic, E = 68 GPa
2 c)
ε = ln(1+εeng) 6
600 4 true

3
5 1
500
b)

True stress, MPa

1 a)
400

a)b)c)
1 Skin
300
2 Sringer
Slant
3 Clip 40 7 fracture
4 Upper-frame
200 5 Lower-frame
6 FE input 11
100 a) Standard test analysis without necking
b) Optical strain analysis within necking
c) Slant fracture
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
True strain, -
Fig. 4. Stress–strain relations of the different built-up parts from flat bar tensile tests; the true stress–strain curve of the skin is also determined locally until
unstable failure.

Table 1
Parameters of the components of the panel.

Part Material Thickness (mm) Yield stress Rp0.2 (MPa)

Skin 2524-T351 1.8 276.0
Stringer 7149-T76511 1.6 540.6
Clip 7475-T762 1.8 287.0
Upper frame 7075-T79511 2.5 475.0
Lower frame 7475-T762 2.0 296.2

8    
2
>
>
>
> 2 dd10  dd10 for d < d1
<
T ¼ T0  1 for d1 6 d < d2 ð2Þ
>
>  3  2
>
>
:2 dd 2 dd
 3 d0 d2 þ 1 for d2 6 d < d0
2
d0 d2

Besides the cohesive parameters T0 and d0, this TSL contains two additional shape parameters, d1 and d2, allowing for vary-
ing the shape of the curve. With this function, the cohesive energy, C0, is related to T0 and d0 by
Z d0  
1 2 d1 d2
C0 ¼ TðdÞdd ¼ T 0 d0 1  þ ð3Þ
0 2 3 d0 d0
Fig. 5b shows the TSL for d1 = 0.01d0 and d1 = 0.5d0, which has proven to provide reasonable results for ductile crack exten-
sion [14,18,20].

4.2. Determination of the cohesive parameters

As already mentioned at the end of the introduction, some mechanical fracture mechanics tests are needed for determin-
ing the cohesive parameters. However, for the skin material no flat sheet of the same charge as that of the fuselage was avail-
able for this purpose. Therefore, two modified and fatigue pre-cracked Kahn specimens [31] as shown in Fig. 6 were
fabricated from the skin material. Since the geometrical offset in this specimen is around 50 lm, It is believed that the cur-
vature of the fuselage skin (R0 = 2820 mm) can be tolerated in this case. During the test, the applied force, F, and the crack
mouth opening displacement (CMOD), were measured.
In thin-walled structures, a crack usually starts its extension in a flat fracture mode, followed by slant fracture mode
where the crack is inclined at 45° to the surface of the component, see Fig. 7d. Since the initial flat mode is very short com-
pared to the amount of crack extension achieved prior to instability of the component, it can be ignored in a structural
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2547

a Undamaged
(bluck)
Damaged zone
(void growth in
Real ductile materials)

b Deformation:
Stress-strain curves
Crack σ
E

Deformation and Fracture σ0, ε0

ε
Δa Fracture:
δ Traction-separation law
Idealization
T
δ1 δ 2
Continuum Cohesive traction T0
Cohesive Solids
T(δ ) elements
Γ0
(Interfaces) δ0
0
0 δ
Crack δ 1 = 0.01δ 0
Separation δ 2 = 0.50δ0
of cohesive
Broken elementes
cohesive
elements δ δ0
δ > δ0

Fig. 5. General aspects of the cohesive model: (a) real ductile fracture and idealization with the cohesive model; and (b) deformation and fracture with solid
and cohesive elements, respectively, and the specified traction-separation law used in this study.

Notched flat bar Smooth flat bar Modified Kahn specimen

Normal and slant fractures
Normal fracture Slant fracture (A+B) T0N, Γ0N
(A) (B)
A

Deep
40 A+B 60 T0S, Γ0S
notch
B

11

Specimens with sheet thickness! 35

Fig. 6. Specimen used for identification the cohesive parameters for normal and slant fracture.

analysis. Crack initiation will then occur slightly earlier. Thus, consideration of this initial flat portion for the determination
of homogeneous cohesive parameters might not be necessary. Therefore, this is the usual way of performing cohesive model
simulations in thin-walled structures, where slant fracture occurs [16,17,19].
However, it turned out that in the Kahn specimen the transition from flat to slant fracture was fully developed past max-
imum force. Thus, the extraction of the cohesive parameters from the Kahn specimen required also consideration of the ini-
tial flat mode to fit the load level due to the flat fracture part. Hence, four cohesive parameters had to be identified: T0N, U0N,
T0S, U0S, with N for normal (mode I) and S for slant fracture (mixed mode I and III).

4.2.1. Mode I fracture

For mode I fracture, a procedure for the determination of the cohesive parameters has been proposed and validated in
[14] which was also used in the present study. The cohesive strength for normal fracture, T0N, can be identified as the max-
imum stress at failure of a deeply notched flat bar under tension (Fig. 6), which is determined from an elastic–plastic FE
2548 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

a Notched tensile specimen b Experimental & Analytical

2000 80
3 Γ0N = 40 N/mm
1800 70 Analytical
1 blunting
1600
2 60 line

J-integral, N/mm
1400 T0N 700 MPa
50 Expected

Force F, N
1200 R-curve
1000 40 Γ0N Initiation
Notch
800 4
30
600 1 Notched tensile tests (2x) 20 SZW
400 2 Load drop
3 FEM, elastic-plastic 10 SZWc = 25 µm
200 4 High tensile stress area
0 0
0 100 200 300 400 0 10 20 30 40 50
Clip displacement, µm Initial crack extension, µm

c d Combination of
Smooth bar test with slant fracture cxperiment & simulation
700
3 T0S Γ0S for slant fracture
2 from simulation
600

500
2 2
Stress, MPa

400
1 T0S 610 MPa 1
300 3
4 4
200 1 Engineering σ−ε
2 True σ−ε 1 mm
100 3 Failure point, T 0S
4 Rp0.2 = 276 MPa 1 Normal fracture 3 Fatigued crack
2 Slant fracture 4 Skin thickness
0
0 0.2 0.4 0.6 0.8
Strain, -

Fig. 7. Determination of the cohesive parameter for the 1.8 mm thick skin material 2524-T351: (a) T0N from deeply notched flat bar tensile specimen; (b)
C0N from a standardised procedure using analytical formulation and stretch zone width from fracture surface; (c) T0S experimentally determined value from
smooth flat bar tensile specimen including necking effect; (d) C0S from cohesive simulation of a precracked Kahn specimen together with the otherwise
determined parameters.

analysis, Fig. 7a, yielding T0N = 700 MPa at failure load. It has to be noted that the notched tensile specimen did not fail in a
pure mode I fashion; a shear mode contribution was also present. Therefore, fine tuning of T0N was carried out as described
below in the section on slant fracture.
In [14] it is also stated that the cohesive energy for normal fracture, U0N, is equal to the J-integral at physical crack initi-
ation in mode I, Ji, which is usually identified from a JR-curve as the intersection point between the critical stretch zone
width, SZWc, determined from the fracture surface and the JR-curve. As in this project no JR-curve was available, an alterna-
tive procedure was applied: Ji was taken from the intersection of the critical stretch zone, SZWc, with the blunting line, which
according to [9,32] is given by a validated analytical solution. The principle is shown in Fig. 7b and yielded U0N = 40 N/mm for
an average value of SZWc = 25 . This SZWc was determined from scanning electron microscope images of the initial fracture
surface of the tested Kahn specimen.

4.2.2. Slant fracture

In the present study, slant fracture is treated like a mode I fracture (normal separation) with the fracture plane normal to
the applied load and appropriate ‘‘effective” cohesive parameters. The cohesive traction for slant fracture, T0S, was deter-
mined experimentally from the flat, smooth, tensile bar shown in Fig. 6. At the end of the test sudden failure occurred as
slant fracture across the thickness. The true stress at failure as calculated by the force at failure divided by the actual cross
section of the specimen was then set equal to the cohesive traction, resulting in T0S = 610 MPa.
In contrast to normal fracture, no independent method for determining the cohesive energy for slant fracture is available.
Therefore, the cohesive energy, C0S, was determined by a finite element crack extension analysis of the two Kahn specimens
tested, based on the cohesive model using the three previously determined parameters T0N, C0N, and T0S. The 3D FE mesh
consisted of linear hexahedral elements. The cohesive elements along the ligament had a width of 0.125 mm and a thickness
of 0.3 mm, thus the element size was in the order of the maximum separation d0. The region, where the parameters for mode
I fracture were applied, is shown in the small sketch inserted in Fig. 8, in which this simulation is denoted as ‘‘nor-
mal + slant”. Comparing the thus simulated force-crack mouth opening displacement (F-CMOD) curve with the experimental
ones yielded U0S = 25 N/mm, see Fig. 8.
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2549

7000

Elast El-pl F
6000 CMOD N S t
1
3
N: Normal fracture
5000 2 S: Slant fracture
Initiation Cohesive simulations
range 1 Normal + Slant
4000
Force F, N
2 Slant 1 (homogeneous)
3 Slant 2 (homogeneous)

3000

2000

1000 Test 2 Test 1

0
0 1 2 3 4
CMOD, mm

Fig. 8. Cohesive parameters optimized by accurate fitting the experimental load–displacement curve (F-CMOD) of the precracked Kahn specimen with large
stable crack extension.

Table 2
Cohesive model parameters identified for the skin material.

Simulation Normal fracture Slant fracture

2
T0N MPa C0N kJ/m T0S MPa C0S kJ/m2
Normal + slant 660 40 610 25
Slant 1 – – 610 25
Slant 2 – – 518 55

10000

9000 Elast
1
8000

7000 2
3

6000
Force F, N

5000 F

CMOD N S t
4000
C(T)

3000 N: Normal fracture

S: Slant fracture
Cohesive simulations
2000
1 Normal + Slant
2 Slant 1 (homogeneous)
1000 Slant 2 (homogeous)
3

0
0 5 10 15 20 25 30
CMOD, mm

Fig. 9. Cohesive simulations providing F-CMOD curves on large C(T) specimen (W = 1000 mm, B = 1.8 mm, a0 = W/2) for three different cohesive parameter
sets.
2550 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

The test served also for fine tuning the cohesive strength for normal fracture, resulting in a final value of T0N = 660 MPa
which is quite close to the original approximated value of 700 MPa anticipated by the notched tensile bar.

4.2.3. Justification of the transition region for the parameter identification

In Fig. 8a further analysis is depicted: since – as mentioned above – in a large structure the normal fracture mode plays
only a marginal role and is usually neglected, the behaviour of the Kahn specimens was also homogeneously simulated using
only the parameters for slant fracture determined above (T0S = 610 MPa and U0S = 25 kJ/m2). It is seen that this simulation, in
Fig. 8 called ‘‘slant 1”, provides a curve substantially below the first one above.
In addition, a second run for identifying homogeneous cohesive parameters ignoring mode I fracture was performed: The
complete experimental F-CMOD curve was fitted and yielded T0 = 518 MPa and U0 = 55 kJ/m2. This simulation shown in Fig. 8
is labelled ‘‘slant 2” and is very close to the experimental curves as well as the first one (‘‘normal + slant”). Table 2 summa-
rizes all cohesive parameters for the three simulations.
Furthermore, in a pure numerical study, the transferability problem from a small specimen to a large structure is illus-
trated by an additional 3D FE exercise on a large fracture specimen, namely a C(T) specimen of 1000 mm width and
1.8 mm skin thickness. Of course, it can be assumed that the region a few millimetres ahead of the initial crack tip, i.e.
the mode I fracture part, has a negligible influence on the global response, here expressed by F-CMOD curves.
A comparison between a simulation with the original heterogeneous parameters T0N, C0N, T0S, C0S (‘‘normal + slant”) and
homogeneous ‘‘slant 1” parameters yields negligible difference, see the F-CMOD curve in Fig. 9. However, if the optimized
homogeneous parameters for ‘‘slant 2” are used, which yielded equal F-CMOD response compared to the heterogeneous sim-
ulation ‘‘normal + slant” in case of the Kahn specimen (Fig. 8), but one can see now, that the F-CMOD curve for ‘‘slant 2” is far
away from the other simulations. Thus, homogeneous cohesive parameters determined on small specimens are not applica-
ble for larger structures.

5. Finite element model of the IMA panel

5.1. FE mesh

A 3D CAD was generated and meshed using the programme IDEAS.3 All construction profiles are connected by rivet blocks,
see Fig. 10a. The two-bay crack was manually inserted. The FE model consisted of 77,208 solid elements with 551,086 nodes.
The elements had 20 nodes and eight integration points (ABAQUS type C3D20R). A refined crack tip mesh block was placed at
each of the two crack tips.
Cohesive elements were inserted ahead of both crack tips, in axial direction with an element width of 0.2 mm and a
length equal to the skin thickness, see Fig. 10b. An element set for normal fracture is considered at the crack front similar
as done for the Kahn specimen as shown in Fig. 8. Each crack tip region contained 702 quadratic cohesive elements with
16 nodes over a distance of about 70 mm ahead of the crack tips.
The finite element analyses were run on a HP workstation X4000. The runs for the panel model took up to 4 weeks.

5.2. Loading and boundary conditions

Loading of the panel was performed numerically as follows:

(i) The internal pressure, pi, was equally distributed over the skin surface and normal to the surface, Fig. 11a. The pres-
surized area is indicated in Fig. 2. The pressure was applied linearly increasing over time until the simulation failed to
converge numerically. Such instability occurs in all load driven tests, whereas stability is easier to achieve in a dis-
placement controlled test. A method to circumvent this problem is the modified RIKS algorithm implemented in
the finite element system ABAQUS/StandardÒ, in which the applied stress and the corresponding displacement at
the boundary are solved simultaneously by an arc length method. However, this algorithm is not applicable to such
complex structures for numerical reasons and thus the maximum pressure can only be determined indirectly by loss
of convergence.
(ii) The equilibrium force of the closed, un-cracked, fuselage was applied as a concentrated force at the left edge of the
panel, with an additional constraint condition such that constant displacement along the edge was achieved, see
Fig. 11a. This point force was proportional to the internal pressure. The right edge of the panel was kept fixed in lon-
gitudinal direction, however, free to move in radial and circumferential directions. For kinematic reasons the central
node was fixed in circumferential direction.
(iii) Further boundary conditions at the skin and at the frame edges were applied in circumferential direction. At the skin
edge a 380 mm long rigid rod was attached to the upper nodes of the solid elements, which rotate around the fixed
support. The boundary conditions induce reaction forces along the skin edge. For frame loading two limit cases,
denoted case A and B, were considered.Case A: The frame edge is loaded in the same way as the skin edge with a slid-

3
Product of Unigraphics solutions.
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2551

a P7 P6 P5 P4 P3
P8 P2
P1

Repeating Skin
unit
Frame

Rivet block inserted

in the global mesh
Repeating unit

b Rivet
axis C3

Two-bay
crack

(1)

Mesh around initial crack-tip

(1) Cohesive elements
Fig. 10. 3D solid mesh of the IMA fuselage panel: (a) frame mesh consisting of repeating units, all with inserted rivet blocks; and (b) mesh detail at the crack
tips together with the cohesive elements ahead into the skin.

ing support around the centre point of a 400 mm long rotating rigid rod, Fig. 11b. Reaction forces will occur at the
frame edges.Case B: Instead of applying boundary conditions at the frame, concentrated forces are applied to all frame
edges, Fig. 11c. This case was realised in the test. The force increases with the internal pressure and results from the
equilibrium force of the closed un-cracked fuselage. However, it should be recognized that the forces in the frames of
the cracked fuselage will be different and particularly vary at different positions in longitudinal direction.

5.3. Material properties in the FE model

In the FE model, the stress–strain curve of the skin identified in Section 3 was used for the whole assembly. As the finite
element analysis has demonstrated, this simplification does not affect the deformation behaviour of the fuselage panel be-
cause the equivalent stresses in the stringers and upper frame profiles adjacent to the crack tips do not significantly exceed
the yield strength of the skin, see later in Fig. 13. The rivets were also assigned the properties of the skin material.

6. Results and discussion

6.1. Deformation and plastic zone extension

First, the global deformation and the distribution of plastic zones are presented, as determined at the failure pressure of
the test and for the initial crack length. The largest deformation appears along the free edges of the two-bay crack as depicted
2552 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

a x-x

Pressure pi
Active force
at the Two-bay crack
skin edge

x-x
C1 C2 C3 C4 C5 C6 C7
x-x: pressure area

b Reaction
1.8 force
Skin edge
element
Pressure pi

End of pres-
sure Fixed
Rigid rod actuator
Rod
Case A R1
R2
Reaction
R1 = 380 mm forces at
R2 = 400 mm fixed actuator

c Pressure pi

Rigid rod

Case B R1
Equal forces from the
un-cracked fuselage

Fig. 11. Boundary conditions of the IMA fuselage panel: (a) pressure pi applied on the outside of the panel; (b) boundaries of case A with reaction forces on
frame and skin edges; and (c) mixed boundaries of case B with equal forces at all frame edges and reaction forces at the skin edge as used in the test.

in Fig. 12a, with two maxima in the middle of the two bays. Most important for the understanding of the deformation behav-
iour and the evaluation of the boundary conditions is the radial deformation along frame C4, shown in Fig. 12b for case A and
in Fig. 12c for case B. All deformations are magnified by a factor of 10.
In case A the radial deformation in the centre of the two-bay crack is significantly lower than in case B. Correspondingly,
the reaction force at the end of frame C4 in case A is by a factor of 1.85 higher compared to case B with equally distributed
forces at the edges of the frames according to an un-cracked fuselage. In case A only the forces at the frame edges at the end
of the panel, C1 and C7, approach the values as actively applied to all frames in case B.
In contrast to case B, the radial deformation in case A approaches gradually that of a complete fuselage. Therefore, the
deformation pattern in case A appears to be more representative of a closed (full circumference) cracked fuselage.
Nevertheless, the boundary conditions of case A do not perfectly represent the complete cracked fuselage, since no cir-
cumferential displacements are possible at the end of the frames. However, this constraint condition yields the highest forces
in the frames (as well at the skin edges), which accounts for the lowest radial displacements (see Fig. 12b). Lower radial dis-
placements are related also to lower crack opening displacements. Thus, case A approaches an upper residual strength level.
On the other hand, case B (also equivalent to the panel test arrangement) is expected to yield a result which is conser-
vative with respect to a complete two-bay cracked fuselage, since the equal forces for all frames yield the largest radial dis-
placements (see Fig. 12c) which are related to a lower residual strength level. Thus case A and B enclose in principle the
margin of the residual strength.
From this study one can see that the way of conducting the panel test does not fully represent the conditions in the com-
plete fuselage, since the length of the circumference is too short and the effect of the crack has not yet vanished in the 45°
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2553

C7
C6
C5
C4

C3
P8 P7 P6
P5 C2
P4
Case A P3
P2
P1 C1

b
P4 P3
P5 P2
P6
P7 P1
P8

End of pi
Crack
Undeformed
Case A
Reaction
forces

Reaction
Crack forces

Frame C4
Case B Equal forces from the
Load level: pi = pi,max,test uncracked fuselage
(test condition)
Fig. 12. Deformation pattern of the IMA fuselage panel at the reference load pi/pi,max,test = 1 and for a stationary crack: (a) overview of the outside
deformation of case A; (b) deformation of frame C4 for case A; and (c) deformation of frame C4 for case B (all deformations are 10-fold enlarged).

section (Fig. 2a), if one takes the analogy of an M(T) panel, where the length of the panel must be at least 1.5 times the width
of the panel in order to obtain results independent of the specimen size.
The plastic zones at frame C3 between stringers P3 and P4 are presented in Fig. 13 as outside and inside views for the
initial crack length and at the load level pi = pi,test,max. The highest iso-line value shows the plastic zone extension, i.e. the
equivalent stress for the yield strength, Rp0.2 = 276 MPa. One can see that plasticity occurs mainly in the skin, but also in
the stringer, the clip foot and the rivets that join the clip to the skin. However, it should be noted that all parts have been
simulated with the stress–strain curve of the skin. While in the real panel the plastic properties of the clip are very similar
to that of the skin (see Fig. 4), the yield strength of the stringer is much higher, and this part only would actually be elastic. As
the region of the plastic zone in the stringer is so small, that any error due to this inaccuracy can be neglected.
Significant plastification occurs as well across the clip rivets, particularly pronounced at the rivet adjacent to the crack
tips, close to stringer P4, since the plastic properties of the rivets could not be tested and thus are not known, again the
2554 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

Fig. 13. Equivalent stress distribution and the corresponding plastic zones at the reference load pi/pi,max,test = 1 at frame C3: (a) outside view; and (b) inside
view. Elastic–plastic FE results of a stationary crack (without cohesive elements).

properties of the skin have been applied. It can be assumed, however, that the real strength of the rivets is higher than that of
the skin, so that the crack shielding effect associated with the rivets is in principle underestimated.

1.2

1.1

1.0 5 pi,test,max
C3 C5 4
Pressure ratio: pi,prediction/p i,test,max

0.9 4
0.88
0.8 1

0.7
2
0.6

0.5

Case A
0.4 3(C3)
3(C5) 1 Simulation at C3
0.3 2 Simulation at C5
3 Crack initiation
0.2
4 Unstable failure, simulation
0.1 5 Unstable failure, test

0
0 5 10 15 20 25 30 35 40 45
Crack extension Δa, mm
Fig. 14. Cohesive simulation for case A: normalized pressure versus crack extension at frames C3 and C5.
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2555

6.2. Residual strength

The residual strength (here expressed by the maximum internal pressure) as predicted with the cohesive simulation for
case A is shown in Fig. 14. Instead of providing only a residual strength value, the complete pressure versus crack extension
curves (pi–Da) are presented for the crack tips in frame C3 and C5, respectively.
The two crack tips propagate slightly differently due to the unsymmetrical composition of the frames below the crack tips
(Fig. 3). The final point of the curve in Fig. 14 is the state where no numerical convergence could be achieved anymore, even
though the time increments are reduced significantly. This end point indicates unstable failure of the structure, i.e. sudden
collapse of a significant portion of the structure. The numerical loss of convergence is mechanically interpreted as equivalent
to the physical instability of the panel test. This was validated on small specimens at maximum load level under load con-
trolled condition. Similar experience regarding simulation of unstable failure is also described in [33].
The pressure–crack extension (pi–Da) behaviour for case B, which is the relevant case for the panel test, is shown in
Fig. 15. From this figure, one can see the distinct effect of the boundary conditions of case A and B. The maximum pressure
prediction is lower than for case A (about 9%) and shows a difference of 21% to the panel test. Crack initiation occurs at an
earlier stage as well. This residual strength reduction is as well consistent with the interpretation of the radial deformations
(Fig. 12b and c) as described above.
The reason for the distinct difference to the panel test can be contributed by a number of parameters within meshing,
boundary conditions, and material properties, idealization of the cohesive model, preloading history, and failure type. Among
these basic items, some contributions in the panel test may be particularly involved, which are not contained in the
simulations:

1. The accuracy of the meshing could have an influence on the structural response. Even though the effect of the length of
cohesive elements in crack extension direction has already been proven to be small if the elements are small enough to
capture the stress gradients in that direction; however, one element in thickness direction seems relatively coarse to cap-
ture the stress state in the out-of-plane direction accurately. It can be shown that a refined modelling change the local
stiffness of the structure, which may alter the crack growth rate slightly, as the cohesive strength follows the stress gra-
dient more gradually and the cohesive elements fails differently, either slower or faster. On the other hand, the increase of
the number of elements along both ligaments would significantly increase the model size and thus computational run
time, which could not be handled during this project.
2. Even though the boundary conditions have been modelled carefully, the connection of the actuators to the structure is
almost perfectly represented: The flexible straps are not modelled and the skin edge is directly connected to rigid rotating
rods. A possible minor influence of the additional compliance is not known.

1.2

1.1

1.0 5 pi,test,max
A4
Pressure ratio: pi,prediction/p i,test,max

0.9
0.88
0.8 1
0.79
B
0.7
2
0.6

0.5

0.4 At frame C3
4-A
1 Simulation, case A
0.3 4-B 2 Simulation, case B
0.2 3 Crack initiation
4 Unstable failure, simulation
0.1 5 Unstable failure, test
0
0 5 10 15 20 25 30 35
Crack extension Δa, mm

Fig. 15. Cohesive simulation for case A and B: normalized pressure versus crack extension at frame C3.
2556 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

1.2
a Residual b
strength 1.1
test pi,test,max
Fatigue flight 1.0 6
cycles for the

Pressure ratio: pi,prediction/p i,test,max

Preload 0.94
two-bay crack 0.9 2 5 1 0.88
*)
5
0.8
4
Internal pressure 0.7 W/o preload

0.6
0.5
(y)
Frame C5: Case A
0.4
1 Simulation, w/o preload
3
0.3 2 Simulation, with preload
3 Crack initiation w/o preload
(x) 0.2 4 Crack initiation with preload
5 Unstable failure, simulation
0.1 6 Unstable failure, test
0
-8 0 10 20 30 40
Time
Crack extension Δ a, mm

*) Cohesive simulation: (x) Preloading path until fatigue overload

One preloading cycle before static (y) Unloading and reloading path
loading; crack-tip starts 8 mm before
the frame axes (C3 and C5).

Fig. 16. Effect of a preloading on residual strength: (a) test situation with fatigue preloading together with a single overload considered in the cohesive
simulation starting at a crack tip 8 mm before the frame axes C3 and C5, respectively; and (b) comparison of cohesive simulations for case A with and
without preloading.

3. The rivets have a significant shielding effect towards the crack opening profile. The use of the single stress–stress curve
from the skin instead of an elastic or even rigid rivet behaviour provides obviously significant conservative predictions.
From additional elastic–plastic FE calculations of the IMA panel (without stable crack extension) it was found that rigid
rivets enhance the load capacity of about 10% compared to the skin–rivet behaviour, determined at the same CTOD val-
ues. Simular, for a stationary crack (at equal CTOD values) the load for case A is about 10% higher than for case B and the
residual strength form the simulation with crack extension is as well of the same order higher (Fig. 15). Therefore it can
be expected, that rigid rivet behaviour (in particular the rivets close to the crack tips) will significantly increase the resid-
ual strength of the panel of approximately 10%.
4. Cyclic preloading, prior to the residual strength test, have also an important effect on the evolution of the crack extension,
which was not yet considered in the simulations above. As mentioned already, from a saw cut the crack was extended by
cyclic loading between zero and cabin pressure. This preloading leads to plastic deformation ahead of the crack tips, and
the crack closure effect after unloading is beneficial for the subsequent strength of the fuselage panel. This aspect has
been investigated by an additional simulation with one preloading step (ramp loading up to cabin pressure and unloading
to zero pressure) starting with a crack tip at about 8 mm before the rivet axes in C3 and C5, respectively. During the first
loading up to cabin pressure the crack extends and reaches the rivet axes and after unloading the initial situation of the
residual strength test is equivalently conducted in the simulation. The subsequent reloading runs till unstable failure.
Fig. 16a shows the load sequence for the panel test together with the single preloading step in the simulation. The com-
plete pressure–crack extension (pi–Da) curves with and without preloading is depicted in Fig. 16b. Even though the curve
is shown for case A only, which is not the one with the boundary condition relevant for the panel test,4 one can see that
preloading in principle increases the pressure at failure at least by 6%. This is mainly due to the fact that crack extension
starts at a much higher internal pressure compared to the simulation without preloading. One can also see that crack exten-
sion before unstable failure is much lower for the preloaded panel.
5. The idealization of the cohesive model does not consider the real slant fracture in a thin-wall structure (under 45° across
the thickness). Crack extension is modelled in the projection plane equivalent to normal fracture but with cohesive
parameters consistently determined as well in the projection plane. However, the real slant crack extension may be
affected differently, particularly in combination with bulging of the crack flanks, compared to the projection plane of
the cohesive model.

4
The investigation has been performed before it was found out that case B reproduces the testing conditions better, but afterwards no more time for the long
simulation was left.
 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558 2557

7. Conclusions

A residual strength analysis of a curved fuselage panel with riveted-on frames containing a two-bay crack has been suc-
cessfully conducted using a 3D FE analysis in combination with the cohesive model. The main conclusions from this study
are as follows:

 A hybrid method combining experiments and simulations for determining the cohesive parameters of a thin-walled Alu-
minium sheet with combined normal and slant fracture modes is verified and successfully applied to the fuselage panel.
 It has been shown in a numerical study that for small specimens the first part, where the crack extends in pure mode I
conditions before it turns into the slant fracture mode, cannot be neglected for small specimens, but does not significantly
affect the response of the large structures. A homogenous fit to small specimen response (load-displacement) are not
applicable cohesive parameters for large crack extension with slant fracture.
 The 3D crack extension simulation of a complex fuselage panel using the cohesive model is robust and able to predict
crack extension of several centimetres (leading to more than 300 failed cohesive elements).
 The residual strength (pressure) was too conservative with respect to the panel test. Five effects are pointed out to under-
stand the predicted conservatism.
 High preloading has a dominant effect on the structural response: Residual strength increases while crack extension at
unstable failure decreases compared to non-preloading. For reasonable predictions this must be taken into account.
 The influence of the boundary conditions has been studied by two different loading types at the frames edges. It has been
shown, that the IMA panel is not sufficiently extended in circumferential direction to achieve the realistic boundary con-
ditions of a complete cracked fuselage barrel.
 The simulation with the cohesive model provides the onset of unstable failure by loss of convergence. This is an important
aspect for large structures, which typically fail unstable. Failure of a rivet panel can also be activated by rivet failure (here
in particular at the crack tips in the frame axes C3 and C5). With careful rivet meshing the presented cohesive model could
also be applied.
 The simulation can be used to substitute the very expensive panel tests for design studies, or to improve the design of this
kind of panel tests.
 The main goal of the presentations should demonstrate the applicability of the cohesive model for predicting large and
complex structural behaviour including unstable failure. The conservative assumptions made should provide the margin
of under-estimation of residual strength prediction. Form the experience made above, it is expected that fine tuning of the
relevant parameters the simulations will realistically approach the fuselage behaviour.

Acknowledgments

Authors gratefully acknowledge Dr. H.-J. Schmidt and Dr. Assler with Airbus Deutschland in Hamburg as well Dr. T. Flei-
scher and Dipl.-Ing. M. Semsch with IMA Materialforschungs und Anwendungstechnik GmbH in Dresden for their kind coop-
eration and support of this study with valuable information on the fuselage test setup. Our colleague Dipl.-Ing. V. Heitmann
is gratefully appreciated for supporting experiments on small specimens.

References

[1] Schmidt H.-J. Damage tolerance technology for current and future aircraft structure. In: Dalle-Donne C, editor. Proceedings of the 23rd ICAF symposium
of the international committee on aeronautical fatigue, 8–10 June 2005, Hamburg, vol. I. p. 1–41
[2] Fleischer T, Sachse M, Semsch M. Engineering approach and results of pressurized fuselage panel testing. In: Dalle-Donne C, editor. Proceedings of the
23rd ICAF symposium of the international committee on aeronautical fatigue, 8–10 June 2005, Hamburg, vol. I, p. 219–27.
[3] Zerst U, Heinimann M, Dalle-Donne C, Steglich D. Fracture and damage mechanics modelling of thin-walled structures – an overwiew. Eng Fract Mech
2007;12:11. doi:10.1016/j.engfracmech.2007.10.00.
[4] Schwalbe K-H, Zerbst U, Brocks W, Cornec A, Heerens J, Amstutz H. The ETM method for assessing the significance of crack-like defects in engineering
structures. Fatigue Fractigue Eng Mater Struct 1998;21:1215–31.
[5] Schödel M, Zerbst U. Application of the European flaw assessment procedure SINTAP to thin wall structures: analytical assessment levels. Eng Fract
Mech 2004;71:1035–58.
[6] Zerbst U, Schödel M, Webster S, Ainsworth RA. Fitness-for-service fracture assessment of structures containing cracks – a workbook based on the
European SINTAP/FITNET Procedure. Amsterdam: Elsevier; 2007.
[7] Schwalbe K-H. Heerens J, Zerbst U, Pisarski H, Koçak M. EFAM GTP 02 – the GKSS test procedure for determining the fracture behaviour of materials.
GKSS Report 2002/24, GKSS Research Center Geesthacht, 2002.
[8] Zerbst U, Brocks W, Heerens J, Schödel M, Scheider I, Steglich D, et al. Failure assessment concepts for thin-walled structures. In: Dalle-Donne C, editor.
Proceedings of the 23rd ICAF symposium of the international committee on aeronautical fatigue, 2005, Hamburg, vol. I; 2005. p. 161–75.
[9] ASTM E 2472-06. Test method for determination of resistance to stable crack extension under low-constraint conditions. Annual book of ASTM
standards, vol. 03.01. West Conshohocken, (PA), USA: ASTM International; 2007.
[10] ISO 22889:2007. Metallic materials – method of test for the determination of resistance to stable crack extension using specimens of low constraint.
Geneva, Switzerland: International Institute of Standardisation; 2007.
[11] Borst R. Numerical aspects of cohesive-zone models. Eng Fract Mech 2003;70:1743–57.
[12] Brocks W, Cornec A, Scheider I. Computational aspect of nonlinear fracture mechanics. In: Milne I, Ritchie RO, Karihaloo B, editors. Comprehensive
structural integrity. vol 3. Amsterdam: Elsevier Science Publication; 2003. p. 127–209.
[13] Elices M, Guinea GV, Gòmez J, Planas J. The cohesive zone model: advantages, limitations and challenges. Eng Fract Mech 2002;69:137–63.
2558 A. Cornec et al. / Engineering Failure Analysis 16 (2009) 2541–2558

[14] Cornec A, Scheider I, Schwalbe K-H. On the practical application of the cohesive model. Eng Fract Mech 2003;70:1963–87.
[15] Roy YA, Dodds Jr RH. Simulation of ductile crack growth in thin aluminum panels using 3-D surface cohesive elements. Int J Fract 2001;110:21–45.
[16] Li W, Siegmund T. An analysis of crack growth in thin-sheet metal via a cohesive zone model. Eng Fract Mech 2002;69:2073–93.
[17] Chabanet O, Steglich D, Besson J, Heitmann V, Hellmann D, Brocks W. Predicting crack growth resistance of aluminium sheets. Comput Mater Sci
2003;26:1–12.
[18] Scheider I, Schödel M, Brocks W, Schönfeld W. Crack propagation analyses with CTOA and cohesive model: comparison and experimental validation.
Eng Fract Mech 2006;73:252–63.
[19] Scheider I, Brocks W. Cohesive elements for thin-walled structures. Comput Mater Sci 2006;37:101–9.
[20] Brocks W, Scheider I, Schödel M. Simulation of crack extension in shell structures and prediction of residual strength. Arch Appl Mech 2006;76:655–65.
[21] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in around tensile bar. Acta Metall 1984;32:157–69.
[22] Rousselier G. Ductile fracture models and their potential in local approach of fracture. Nucl Eng Des 1987;105:97–111.
[23] Gologanu M. Recent extensions of Gurson’s model for porous ductile material. In: Continuum micromechanics. Udine: CISM; 1996.
[24] Reusch F, Svendsen B, Klingbeil D. A non-local extension of Gurson-based ductile damage modeling. Comput Mater Sci 2003;26:219–29.
[25] Lin G, Cornec A, Schwalbe K-H. Three-dimensional finite element simulation of crack extension in aluminium alloy 2024-FC. Fatigue Fract Eng Mater
Struct 1998;21:1159–73.
[26] Nègre P, Steglich D, Brocks W. Crack extension in aluminium welds: a numerical approach using the Gurson–Tvergaard–Needleman model. Eng Fract
Mech 2004;71:2365–83.
[27] Jinkook Kim, Guihua Zhang, Xiaosheng Gao. Modeling of ductile fracture: application of the mechanism-based concepts. Int J Solids Struct
2007;44:1844–62.
[28] Scheider I. Residual stength prediction of a complex structure using crack extension analyses. Eng Fract Mech 2008;75:4001–17.
[29] Chen C-S, Wawrzynek PA, Ingraffea AR. A methodology for fatigue crack growth and residual strength prediction with applications to aircraft fuselages.
Comput Mech 1997;19:527–32.
[30] Scheider I, Brocks W. Simulation of cup-cone fracture using the cohesive model. Eng Fract Mech 2003;70:1943–61.
[31] Annual Book of ASTM Standards: ASTM B 871-96. Standard test method for tear testing of aluminum alloy products. West Conshohocken, USA: The
American Society for Testing and Materials; 1996.
[32] Schwalbe K-H, Hayes B, Baustian K, Cornec A, Gordon R, Homayun M, et al. Validation of the fracture mechanics test method EGF P1-87D (ESIS P1-90/
ESIS P1-92). Fatigue Fract Eng Mater Struct 1993;16:1231–84.
[33] Kabir R, Cornec A, Brocks W. Simulation of quasi-brittle fracture of lamellar gTiAl using the cohesive model and a stochastic approach. Comput Mater
Sci 2007;39:75–84.