You are on page 1of 21

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/222557331

Stress based fracture envelope for damage


plastic solids

Article in Engineering Fracture Mechanics · February 2009


DOI: 10.1016/j.engfracmech.2008.11.010

CITATIONS READS

42 138

1 author:

Liang Xue
Thinkviewer, LLC
27 PUBLICATIONS 999 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sandia fracture challenge View project

Ductile fracture modeling and formulation of damage plasticity theory View project

All content following this page was uploaded by Liang Xue on 05 December 2017.

The user has requested enhancement of the downloaded file.


Engineering Fracture Mechanics 76 (2009) 419–438

Contents lists available at ScienceDirect

Engineering Fracture Mechanics


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

Stress based fracture envelope for damage plastic solids


Liang Xue *
Department of Mechanical Engineering, Northwestern University, 2145 Sheridan Road, M-110, Evanston, IL 60208, USA

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

Article history: Recent development of damage plasticity theory shows the critical plastic strain at fracture
Received 5 August 2008 for ductile solids depends on the pressure and the Lode angle on the octahedral plane along
Received in revised form 20 November 2008 the loading path. The determination of the fracture strain envelope is usually a difficult and
Accepted 24 November 2008
time consuming process. This is due to the experimental difficulties in maintaining a con-
Available online 6 December 2008
stant pressure and Lode angle at the fracture site, which is further complicated by the cou-
pled nature of the parameters to be calibrated and the geometrical localization of the
Keywords:
deformation. The fracture strain envelope is one of the key ingredients of the damage plas-
Ductile fracture
Damage plasticity
ticity theory and relates to the accuracy of predicted results. In the present paper, the Lode
Lode angle dependence angle dependence and the pressure sensitivity functions for the fracture strain envelope
Pressure sensitivity are derived from the hardening rule of the matrix using Tresca type fracture condition
Maximum shear stress and Drucker–Prager formula, respectively. Quantitative analyses of Clausing’s and Bridg-
man’s test data are presented. Then a pressure modified maximum shear stress condition
is adopted as fracture initiation condition to examine their joint effects on the fracture
strain envelope. The relationship of the strain hardening, the pressure sensitivity and the
Lode angle dependence are examined and verified by existing experimental results. We
show that within the moderate range of stress triaxiality, the pressure modified maximum
shear condition can be used as the fracture stress envelope for ductile metals within the
framework of damage plasticity. The present method reduces significantly the amount of
work to calibrate the material parameters for ductile fracture.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction

From a microscopic viewpoint, ductile fracture is the ultimate result of the damaging process which is generally speaking
controlled by the nucleation and evolution of micro voids and micro cracks and their coalescence. In this regard, theoretical
analyses and experimental investigations have been focused on the pressure (or mean stress) effect and the relationship be-
tween the plastic distortion and the void nucleation and coalescence, e.g. Refs. [1–5,26]. Numerical simulations show that
the void coalescence also depends on the Lode angle of the principal stresses [6,7]. The previously missing portion of damage
associated with the void shearing effect was recently introduced to the Gurson-type model such that ductile fracture at low
stress triaxiality can be predicted [8,9]. Macroscopically, the damage process can be described in terms of the local stress and
strain components. This type of continuum damage mechanics models have been derived from thermodynamic point of view
[10–13]. Recent development in ductile fracture modeling has extended the conventional I1, J2 type of continuum damage
plasticity theory in the stress invariant space to a full three-dimensional description of plastic damage which incorporates all
three stress invariants [14]. In this formulation of damage plasticity model, reference loading paths of proportional loadings
under constant pressure and constant Lode angle are adopted. The joint effects of the pressure and the Lode angle on the
ductile fracture initiation are represented by a fracture strain envelope. The damage accumulation is thus expressed in terms

* Tel.: +1 847 491 3046.


E-mail address: 1-xue@northwestern.edu

0013-7944/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2008.11.010
420 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

of the ratio of the current plastic strain to the fracture strain and is assumed to be self-similar. Under this framework, the
damage plasticity model was adopted and calibrated for aluminum alloy 2024-T351 from a series of experiments of different
stress states [15]. Numerical simulations using this model have illustrated the capability of capturing several features ob-
served in ductile fracture experiments, such as the formation of shear lips in tensile round bars [14] and in three point bend-
ing of rectangular bars [16] and the mode transition in compact tension specimens [17].
The determination of the fracture envelope in the principal strain and the mean stress space is one of the key elements for
the damage plasticity theory. However, the direct calibration of the fracture strain envelope is not an easy task for ductile
materials. First of all, the fracture strain depends on the entire loading history, i.e. the pressure and the Lode angle, which
are difficult to maintain at a constant level. Up to date, there is no experimental method available for the determination
of the ductile fracture envelope at constant pressures and constant Lode angles. Secondly, the ductile fracture problem is
highly non-linear – the deformed geometry is not uniform and the matrix stress–strain curve is non-linear either. The sit-
uation is worsened further by the coupled nature of the material parameters and the localization due to the weakening ef-
fect. The local configuration at the fracture site can change not only by a considerable size, but also in its shape. Therefore,
parallel numerical simulations have to be performed usually in order to gain a satisfactory confidence level on the calibrated
material constants.
In the present paper, an alternative approach to determine the fracture strain envelope is developed from describing the
fracture stress envelope in the first place and then derive inversely the fracture strain envelope from the matrix stress–strain
relationship. Due to the highly non-linear nature of the pressure dependence curve on the fracture strain and the assumed
shape for the Lode angle dependence function in the previous work [14], more material parameters are necessary to capture
the shape of the fracture strain envelope. Here, the pressure and the Lode angle dependence functions for the fracture stress
envelope is assumed to be linear in the principal stress space, hence the total number of material constants to be calibrated is
reduced to only two. This is a significant saving in the calibration work. The applicability of the present approach is further
discussed and the accuracy is assessed by reviewing existing experimental results. Numerical simulations of four experi-
ments in the range of the Lode angles and the mean stresses common to practical problems are also presented. Good agree-
ment is achieved in the simulation results compared with the experimental results in terms of the fracture pattern and the
load–displacement curves.

2. Continuum theory of damage plasticity

In an early paper, a continuum theory of damage plasticity is proposed by introducing the so-called ‘‘cylindrical decom-
position” for damage evaluation for ductile materials [14]. The fundamental hypothesis adopted is that ‘‘the damaging process
is self-similar on any deviatorically proportional loading path at any given pressure”. The essential parts of the damage plasticity
theory are a pressure sensitivity function, a Lode angle dependent function, a damage evolution law and a weakening func-
tion to characterize the material deterioration over the entire range of the plastic flow process until the onset of fracture.
Two kinds of Lode angle dependence function are proposed based on the experimental observation of less fracture plastic
strain in generalized shear condition (the Lode angle hL ¼ 0) than that in generalized tension condition (hL ¼ p=6), e.g.
for a number of metals reported in Ref. [18]. A logarithmic pressure sensitivity function for the fracture envelope is also de-
rived from tensile test results under superimposed hydrostatic pressure [14]. The damage plasticity model can be summa-
rized in the following set of equations:
8
>
> rM ¼ r^ M ðep Þ; ðmatrix stress—strain curveÞ
>
>
> b
< req ¼ ð1  D ÞrM ; ðdamage coupled Mises yield conditionÞ
 ðm1Þ ð1Þ
> e e_ p
>
> D_ ¼ m ep ef ; ðdamage evolution functionÞ
>
> f
:
ef ¼ ef0 lph ðp; hL Þ ¼ ef0 lp ðpÞlh ðhL Þ; ðductile fracture strain envelopeÞ

where
8  
> p
> lp ðpÞ ¼ 1  q log 1  plim ;
>
>
ðpressure sensitivity functionÞ
>
> pffiffiffi
< l ðh Þ ¼ c cos h0 ; where tan h0 ¼ 3  2c; ð1st kind of Lode angle dependence functionÞ
h L cosðh0 þjhL jÞ ð2Þ
>
> or
>
>  k
>
>
: lh ðhL Þ ¼ c þ ð1  cÞ jhL j
: ð2nd kind of Lode angle dependence functionÞ
p=6

where r^ M ðep Þ is the intrinsic matrix stress–strain property, ep is the current plastic strain, ef defines the fracture envelope,
the Lode angle hL denotes the deviatoric stress state, req is the equivalent stress, ð1  Db Þ is the weakening function to ac-
count for the material degradation, D_ is the damage rate, lph denotes the joint effects of the pressure and the Lode angle
to the fracture strain envelope and is assumed to have a multiplicative form of two scalar quantities – lp and lh , which
are the pressure and the Lode angle dependence functions, respectively. The Lode angle is defined by the principal stress
components
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 421

  !
1 2r2  r1  r3 1 1 27 J 3
hL ¼ tan1 pffiffiffi or hL ¼  sin ; ð3Þ
3 r1  r3 3 2 r2eq

where r1 ; r2 and r3 are the ordered principal stress components and J 3 ¼ s1 s2 s3 is the third stress invariant where s1 ; s2 and
s3 are the ordered principal stress deviator components. For arbitrary stress state, the Lode angle is in the range of [p/6,
p/6], where hL = p/6 denotes the generalized tension condition, hL = 0 denotes the generalized shear condition and
hL = p/6 denotes the generalized compression condition.
The material parameters are the damage exponent m, the weakening exponent b, the reference fracture strain ef0 , the
shape parameter q and the limiting pressure plim in the pressure dependence function and the fracture strain ratio c and
the Lode angle dependence exponent k. When the damage reaches a critical value of unity, the material loses its load carrying
capacity, i.e. fracture occurs. Due to the phenomenological nature of the present approach, these functions are not unique
and other functional forms are possible without altering the spirit of the method.
Previously the fracture envelope is defined and calibrated directly using the plastic strain at fracture. In the present paper,
we adopt a different approach to determine the fracture envelope. The fracture envelope is first expressed in terms of the
equivalent stress of the matrix material and then the fracture strain is inversely obtained from the hardening function.
The Lode angle dependence and the pressure sensitivity are then discussed. We firstly adopt a Tresca condition (maximum
shear stress) for the fracture envelope. This allows us to show the relationship of the Lode angle dependence function with
the matrix stress–strain curve. Secondly, we adopt a Drucker–Prager type of linear pressure sensitivity relationship for the
fracture envelope in the principal stress space. We also derive the pressure sensitivity function in the plastic strain–pressure
space. Lastly, we combine the above two fracture conditions for the matrix and show their joint effects to the fracture strain
envelope. The Coulomb-Mohr type stress envelope is also discussed. Applicability to polycrystalline metals is shown by com-
paring the derived formulas with experimental data. The proposed method is also introduced to finite element code and ver-
ified by a series of numerical simulations of lab tests.

3. Lode angle dependence of ductile fracture

We start by showing a simplified Lode angle dependence function can be determined from the maximum shear stress
condition. At least for a number of materials, the maximum shear stress has been found to give reasonable prediction for
the fracture loci of ductile materials [19]. The pressure dependence of the fracture envelope is neglected for the moment
and will be discussed later.

3.1. Determination of the Lode angle dependence function

Assuming a matrix material obeys von Mises yield condition and a maximum shear stress fracture condition, the plastic
strain at fracture can be determined for the matrix from the hardening rule. The hypothesis of the equivalence of the strain
tensor with the matrix strain tensor is adopted [20].
Let the hardening law be a Swift type hardening law, i.e.
 
ep n
rM ¼ ry0 1 þ ; ð4Þ
e0
where ry0 is the initial yield stress, n is the hardening exponent and e0 is a reference strain. Let the fracture stress at general-
ized tension be rf0 . The fracture surface in the principal stress space can be expressed by
h p pi
rf ¼ rf0 yh ðhL Þ; 8 hL 2  ; ; ð5Þ
6 6
where hL is the Lode angle and yh ðhL Þ is the Lode angle dependence stress function for the maximum shear stress condition,
i.e.
pffiffiffi
3
yh ðhL Þ ¼ : ð6Þ
2 cos hL
Substituting Eq. (6) into Eq. (5), one can obtain the matrix plastic strain at fracture from Eq. (4),
2 pffiffiffi !1=n 3
rf0 3
ef ¼ e0 4  15: ð7Þ
ry0 2 cos hL
  
1=n
At hL ¼ p=6, the fracture plastic strain is eu ¼ e0 rry0
f0
 1 . This eu denotes the fracture strain under generalized ten-
sion condition. Normalizing Eq. (7) with respect to eu , the Lode angle dependence function for fracture envelope is then
 pffiffi 1=n
rf0 3
ry0 2 cos hL 1
lh ¼  1=n : ð8Þ
rf0
ry0 1
422 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

The right hand side of Eq. (8) is defined by three material constants, namely the initial yield stress ry0 , the fracture stress
at generalize tension condition rf0 and the strain hardening exponent n, and the current Lode angle hL , which is defined
purely by the deviatoric stress state.
The ratio of plastic strain at fracture for hL ¼ 0 and hL ¼ p=6 is defined as a material parameter c in Ref. [14], which can
be determined as
 pffiffi1=n
rf0 3
ry0 2 1
c ¼  1=n : ð9Þ
rf0
ry0 1

By assuming the shape of the fracture envelope, we not only establish a relationship between the stress–strain curve to
the fracture parameter c, but also show the Lode angle dependence of the fracture surface. For metals with n = 0.1–0.3 and
rf0 =ry0 ¼ 1:5—5:0, the value of c varies between about 0.25–0.60 (see Figs. 1 and 2).
From the Swift relationship, we have ðrf0 =ry0 Þ1=n ¼ 1 þ eu =e0 . For ductile materials, the fracture plastic strain is usually
much greater than the reference strain, i.e. eu  e0 . Thus, ðrf0 =ry0 Þ1=n > 1. When the condition ðrf0 =ry0 Þ1=n  1 is satisfied,
an approximation solution to Eq. (8) can be obtained, i.e.
pffiffiffi !1=n
3
lh ¼ : ð10Þ
2 cos hL
pffiffiffi
Note, the Lode angle is defined for ½p=6; p=6, we have 3=2 6 cos hL 6 1 in the above approximation, therefore, y1=n
h is
not negligible. The material parameter c is approximately
pffiffiffi!1=n
3
c¼ : ð11Þ
2

The exact solution (Eq. 8) and the approximate solution (Eq. 10) for the Lode angle dependence function derived from
Tresca fracture condition are plotted in Fig. 1 for several n values for two materials having rf0 =ry0 ¼ 2:0 (Fig. 1 left) and
rf0 =ry0 ¼ 4:0 (Fig. 1 right), respectively.
Fig. 1 can be compared with the first and the second kind of Lode angle dependence functions proposed in Ref. [14]. These
curves in the polar coordinate systems resemble each other to a certain degree. However, the present shape belongs to nei-
ther of the two kinds. Rather than depends on a predefined shape function, the shape of the present Lode angle dependence
function depends on the magnitude of the fracture strain and the shape of the stress–strain curve of the matrix. It is note-
worthy that the function is smooth at hL ¼ 0 where in the stress representation is smooth and has a cusp at hL ¼ p=6 and
p=6 where the stress fracture envelope has point vertices.
For power law hardening matrix materials rM ¼ Aenp , the Lode angle dependence function can be derived as the same
equation as the approximate solution, i.e. Eq. (10). It can be seen from Eq. (10) that the magnitude of the stress–strain curve

σ /σ =2.0 ε σ /σ =4.0 ε
f0 y0 2 f0 y0 2
90 90
1 1
n=1.0 120 60 n=1.0 120 60
0.8 0.8
n=0.4 n=0.4
0.6 0.6
150 30 150 30
0.4 0.4
n=0.2 n=0.2
0.2 0.2

n=0.1 n=0.1
180 0 180 0

210 330 210 330


ε3 ε1 ε3 ε1

240 300 240 300


exact exact
270 approximate 270 approximate

Fig. 1. The Lode angle dependence functions derived from Tresca condition: exact solutions (solid lines) and approximate solutions (dash-dot lines). (left:
rf0 =ry0 ¼ 2:0; right: rf0 =ry0 ¼ 4:0).
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 423

0.9
σf0/σy0=∞
0.8
σf0/σy0=5.0

Fracture Strain Ratio – γ


0.7
σ /σ =2.0
f0 y0
0.6
σ /σ =1.5
0.5 f0 y0

0.4
σf0/σy0=1.33
0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
Hardening Exponent – n
Fig. 2. The relationship between the fracture strain ratio c and the strain hardening exponent n for several ratios of fracture to yield stress.

(A, having same unit as stress) and the matrix fracture stress at generalized tension condition rf0 does not affect the Lode
angle function.
It is trivial to derive for the Ludwig type of hardening matrix rM ¼ A þ Benp (ry0 ¼ A and B have the same unit as stress),
which is used by Johnson and Cook [21]:
0 pffiffi 1 1=n
rf0 3
1
lh ¼ @ry0 r2 cos hL A : ð12Þ
ry0  1
f0

From Eq. (12), the material parameter B does not affect the Lode angle dependence function. When n ¼ 1, both Swift type
and Ludwig type of hardening rules degenerate to the same lh .
In the present model, by assuming a Tresca type of fracture stress surface, the fracture strain ratio c is no longer a material
parameter to be calibrated. It is implied by the fracture stress loci and the strain hardening properties. However, it can serve
as an additional check point to verify the accuracy of the model if c is known. A fast calibration methods for c is proposed in
Xue [14] by utilizing the conventional method in calibrating the stress triaxiality dependence of fracture of ductile materials.
The ratio of the simple shear fracture strain to the intercept of the stress triaxiality dependence curve with zero stress tri-
axiality line defines the value of the fracture strain ratio c. In both loading cases, the mean stresses are theoretically constant
zero. The transferability of the Tresca type of Lode angle function on an octahedral plane to the fracture strain ratio c can be
verified by the calibrated aluminum alloy 2024-T351 [22]. The stress–strain curves for aluminum alloy 2024-T351 are cal-
ibrated using Swift type of hardening rule. The model parameters are listed in Table 1.
A series of numerical simulations in parallel with the experiments are conducted in order to calibrate the material con-
stants for the damage plasticity model. This calibration procedure is reported in detail in Ref. [15]. The choice of experiments
for calibration includes a tensile round bar test, a notched round bar test, a transverse plane strain test and an upsetting test
at low friction condition. This set of tests covers at least three different Lode angles and three different pressure conditions. It
is also reported in [15] for additional tests and numerical simulation to further validate the calibrated material constants.
From the calibration process, the c value is found to be 0.4. The calculated value from the derived equations is 0.43 as
listed in Table 1. This agrees with the calibrated value and suggests the Tresca type of fracture loci on an octahedral stress
plane could be a fairly good approximation for aluminum alloy 2024-T351 with limited number of tests. It is also found the
aluminum alloy 2024-T351 in the form of an extruded bar has some anisotropy due to the manufacturing process [15]. How-
ever, we will focus on the isotropic model in the present paper. Readers are referred to Ref. [26] for an analysis on the aniso-
tropic aspect.

3.2. Hardening effect and Clausing’s test

It was noted from collective experimental results that the fracture strain ration c is affected by the stress–strain curve of
the material. With increasing hardening capability, the fracture strain ratio c increases, which indicates the difference of the

Table 1
The comparison of calibrated and calculated fracture strain ratio for aluminum alloy 2024-T351.

ry0 e0 n ef0 Calibrated c Exact c Eq. (9) Approximately c Eq. (11)


AA2024-T351 compression 302 MPa 0.00387 0.173 0.8 0.40 0.43 0.44
424 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

Table 2
The difference in ductility of the axisymmetric uniaxial tension and the transverse plane strain tension. Experimental data from [23,26].

Steel Yield strength ry0 Tensile strength ruts Fracture strain est Fracture strain eps Fracture strain ratio
(MPa) (MPa) (simple tension) (plane strain) eps =est  c
ABS-C 262.7 428.9 1.04 0.75 0.72
A302-B 371.0 590.2 0.98 0.72 0.73
HY-80 586.1 689.5 1.22 0.73 0.60
HY-130(T) 937.7 986.0 1.06 0.39 0.37
18Ni(180) 1227.3 1289.4 1.00 0.42 0.42
10Ni–Cr–Mo–Co 1254.9 1413.5 1.16 0.36 0.31
18Ni(250) 1710.0 1772.0 0.89 0.15 0.17
X52 (T) 406 555.7 0.96 0.45 0.47

fracture strain between the generalized shear condition and the generalized tension condition is diminishing [22]. For a fixed
rf0 =ry0 value, the fracture strain ratio can be determined for different hardening exponents. Fig. 2 shows increasing c with
respect to increasing hardening exponent n for several values of fracture to yield stress ratio at generalized tension condition.
This trend agrees with the experimental observations.
Furthermore, we verify this trend using Clausing’s experiment [23]. Clausing tested seven structural steels under simple
tension of a round bar and under transverse plane strain tension, which have distinct Lode angles in the stress state. These
steels have different hardening properties. The main finding is the unanimous decrease in the tensile ductilities of the plane
strain tension (grooved plate specimen) from the axisymmetric tension (round bar) for those metals, see Table 2. Compared
with the simple shear or torsion condition, the uniaxial plane strain tension has an additional hydrostatic tensile pressure,
which facilitate the void nucleation and growth and slip band movement. However, the difference of this hydrostatic tension
in the axisymmetric and the plane strain tension is less significant than that in the torsion and the axisymmetric tension.
Unlike the traditional picture (e.g. Hancock and Brown [24] and Bao [25]) that attributes the difference in the ductility of
the plane strain tension to the axisymmetric tension solely to the stress triaxiality and the Lode angle dependence of ductile
fracture is completely overlooked, here, we neglect the pressure differences between the two conditions and explain the dif-
ference using Lode angle dependence function derived above.
Clausing listed the initial yield stresses, the ultimate tensile strength and fracture strain for the seven materials he tested
[23], which allows us to estimate the strain hardening effect by adopting the ratio of the fracture to yield stress and a linear
hardening rule (i.e. n = 1). The analyzed data are listed in Table 2. Without the exact knowledge of c, we use the ratio of eps
(the fracture strain at plane strain tension) and est (the fracture strain at simple tension) as a measurement of the sensitivity
of the Lode angle dependence. The smaller this ratio, the higher sensitivity on Lode angle the ductile fracture of the material.
The last two columns of Table 2 are plotted in Fig. 3. By estimation, the matrix equivalent stress at fracture is taken to be 20%
more of that of the measured ultimate tensile strength to compensate for the geometrical change of the tensile specimen and
the weakening effect to the matrix along the loading path, i.e. rf0 ¼ 1:2ruts . The results are plotted in Fig. 3. Observing the
fracture strains under simple tension for these material does not vary drastically, the fracture to yield stress ratio rf0 =ry0 are
adopted to represent the hardening capability of these materials. The hardening is considered to be linear, i.e. n = 1 for Swift

0.9
γ=(0.866σf0/σy0–1)/(σf0/σy0–1)
0.8
Fracture Strain Ratio – γ

0.7

0.6

0.5

0.4

0.3

0.2
Theoretical Eq. (9)
0.1 Clausing (1970)
Benzerga (2002)
0
1 1.5 2 2.5
σf0 / σy0

Fig. 3. The fracture strain ratio versus the strain hardening capacity. Experimental data from [23,26]. Theoretical curve is plotted for the exact solution with
hardening exponent n = 1.
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 425

type of stress–strain relationship. It is shown that the higher the strain hardening effect, the lower the Lode sensitivity is.
Although the low hardening capability has been known for favoring a shear type of fracture and resulting low ductility,
a quantitative model is still in order. Here, we show in Fig. 3 that the predicted curve of fracture strain ratio achieves
good agreement with the experimental results. We conclude that the difference in ductility between the plane strain
tension and the axisymmetric tension is mainly due to the change in the Lode angle, which is from hL ¼ 0 to hL ¼ p=6
for these steels. Similar to Clausing’s test, Benzerga et al. conducted the tests on X52 steel [26] and their data is also shown
in Table 2.
Hancock and Brown [24] repeated Clausing’s experiment, but found less significant differences between the fracture
strains in simple tension and a flat grooved plate tension. The length to height ratio of the groove used by Hancock and
Brown was 1.0. In Clausing’s experiments, the length to height ratio was 4.0. Benzerga et al. also adopt a 4:1 ratio in their
experiments [26]. Adams conducted a series of tests on doubly grooved plates and suggested the length to height ratio of the
groove need to be great than 10.0 to ensure a plane strain condition for grooved plates [27]. Therefore, the length to height
ratio for specimen grooves in Hancock and Brown’s tests was too low and the plane strain constraint was lost to the diffused
neck in the grooved region.
It could be further assessed the differences in the stress triaxiality between a necked round bar under remote uniaxial
tension and the transverse plane strain condition. The simplest way is by using Bridgman’s correction, the stress triaxiality
T, which is defined as the ratio of the mean stress to the equivalent stress, at the center of the neck is
1  r
T¼ þ ln 1 þ ; ð13Þ
3 2R
where r is the minimum cross-sectional radius at the neck and R is the curvature radius of the neck profile.
pThe
ffiffiffi condition that the stress triaxiality at a necked round bar is the same as the transverse plane strain tension (which is
1= 3) is
1  r 1
þ ln 1 þ ¼ pffiffiffi ; ð14Þ
3 2R 3
and Eq. (14) yields R ¼ 1:81r. For ductile metal, this relationship is in the proper range. Moreover, the pressure dependence
curve is less sensitive to pressure in the high mean stress range as in simple tension and transverse plane strain tension. This
will be shown in the next section.

4. Pressure sensitivity function

The fracture strain is sensitive to the superimposed hydrostatic pressure for ductile materials. Such a behavior has been
long recognized as a material property for metals and granular materials, e.g. [28–33] etc. For simplicity, we adopt Drucker
and Prager’s linear pressure sensitive function [34] as the fracture stress envelope for the matrix
rM ¼ rf0 yp ðpÞ; ð15Þ

where rf0 denotes the equivalent fracture stress at zero pressure and yp is a dimensionless factor and is defined as
yp ðpÞ ¼ 1 þ kp p; ð16Þ

where kp is a material constant having unit the reciprocal of pressure. From numerous experiments, kp is a positive number
for common materials, such as commercial grade metals.
Now, the pressure sensitivity function for fracture envelope can be determined from Eqs. (15) and (4). We write the frac-
ture plastic strain as
" 1=n #
rf0 1=n
ef ¼ e0 ð1 þ kp pÞ 1 : ð17Þ
ry0
Normalizing Eq. (17) with respect to the fracture at zero pressure, the pressure sensitivity function can be determined as
 1=n
rf0
ry0 ð1 þ kp pÞ1=n  1
lp ¼  1=n : ð18Þ
rf0
ry0  1

When ðrf0 =ry0 Þ1=n  1, Eq. (18) degenerates to

lp ¼ ð1 þ kp pÞ1=n : ð19Þ

The value of kp can be estimated from existing results with regard to the fracture strain at various pressure. Here, we
make use of the widely used Johnson–Cook model. The stress triaxiality dependence of the Johnson–Cook model takes
the forms of Hancock and Mackenzie’s modified exponential formula [35] originally due to Rice and Tracey [36]. Knowing
the fracture strain is
426 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

ef ¼ D1 þ D2 expðD3 TÞ; ð20Þ


where T ¼ p=req is the stress triaxiality ratio defined as the negative pressure divided by the equivalent stress. Johnson and
Cook used a linearized integral form for damage evolution due to the complex stress triaxiality path usually experienced by
the material before fracture takes place [21]. We also notice that the damage accelerates toward the fracture point due to the
increasing interaction activities between micro voids and micro cracks. For simplicity, we replace the instantaneous equiv-
alent stress along the plastic loading path with the equivalent stress at fracture by approximation. This allows us to derive an
approximate solution for kp using existing calibrated parameters of Johnson–Cook model for different materials.
We equate the kp ry0 in the present model to the slope of Johnson–Cook model at zero pressure. Knowing ef takes the form
of Eq. (20) and taking derivative of the Ludwig’s relationship with respect to p at p ¼ 0, we have

@ rf D2 D3
kp rf0 ¼ ¼ Bnen1 : ð21Þ
@p p¼0 f0
rf0
At p ¼ 0, ef0 ¼ D1 þ D2 and note ry0 ¼ A for Ludwig model. Therefore, the pressure sensitivity parameter kp can be deter-
mined as
ABnðD1 þ D2 Þn1 D2 D3
kp ry0 ¼ 
2 : ð22Þ
A þ BðD1 þ D2 Þn
For several metals, the Johnson–Cook material parameters are used to calculate the kp . These material parameters are
listed in Table 3. It is found that kp ry0 is about in the range of 0.04–0.20 for common metals.
The exact and approximate solutions of pressure sensitivity functions Eqs. (18) and (19) are plotted for several strain
hardening exponents n with different pressure sensitivity parameters kp for fixed value of rf0 =ry0 ¼ 2:0 in Fig. 4, and for sev-
eral fracture to yield stress ratios rf0 =ry0 with different kp values for fixed hardening exponent n = 0.2 in Fig. 5.
It is noteworthy that, when the damage exponent m = 1, the functional form of the approximate pressure sensitivity func-
tion is identical to the Wilkins model [41], whose damage model can be translated into a pressure dependence of
lp ¼ ð1 þ apÞa with a linear damage rule. For the Wilkins model, the material parameters a and a are fitting parameters ob-
tained from fracture strains measured from experiments. In the present derivation, it is shown that the physical meanings of
the material parameters kp and n in Eq. (19) relate to the fracture stress envelope and the stress–strain curve. By comparing
the two forms, the following relationships between the material parameters can be established for the Wilkins model and
the present model:
a ¼ kp and a ¼ 1=n: ð23Þ
We can further compare the model parameters using the material parameters for aluminum alloy 6061-T561 calibrated
by Wilkins et al. [41]. Since the value a is just the reciprocal of the strain hardening exponent and can be calibrated from
simple tension tests when weakening is ignored, this leave a to be the only parameter to be calibrated for pressure depen-
dence in the Wilkins model. For aluminum alloy 6061-T561, Wilkins et al calibrated the stress–strain curve to be
ry ¼ 285ð1 þ ep =0:008Þ0:1 MPa, a ¼ 0:00133 MPa1 and a ¼ 1:8. The calibrated

numbers n and a do not fit the relationship
@ lp
in Eq. (23) well. Taking calibration error into consideration, we use @p
¼ constant as a constraint, we can derive that
ðp¼0Þ

aa ¼ constant to adjust the value of a to enforce a ¼ 1=n ¼ 10 here. We thus obtain a ¼ kp ¼ 0:00024 MPa1 . This is under-
stood as the calibration was done to fit the range of pressure experienced in their experiments. Within that range, these two
sets of parameters do not differ from each other much, see Fig. 6. Compared with the original calibrated parameters in
Wilkins et al. [41], the new parameters derived from the material hardening exponent is more sensitive to pressure due
to higher exponent of a (from 1.8 to 10). We also have kp ry0 ¼ 0:0684.
Previously, a logarithmic function was proposed for the pressure sensitivity function [14], such that a limiting pressure
exists and above this threshold no damage to the material will occur. The limiting pressure was based on observations that
with extremely high pressure the existing micro or macro cracks can be healed, e.g. in the practice of cold press weld. At a

Table 3
Approximate solution for pressure sensitivity parameter kp ry0 for several metals.

Material A (MPa) B (MPa) n D1 D2 D3 kp ry0 References


OFHC copper 90 292 0.31 0.54 4.89 3.03 0.110 Johnson and Cook [21]
AA 2024-T351 352 440 0.42 0 1.0 1.94 0.201 Teng and Wierzbicki [37]
AA 5083-H116 (90°) 167 596 0.551 0.0261 0.262 0.349 0.040 Clausen et al. [38]
AA 5083-H116 (90°) 143 554 0.526 0.00139 0.415 0.601 0.065 Clausen et al. [38]
AA 5083-H116 (90°) 152 594 0.527 0.0000677 0.335 0.536 0.061 Clausen et al. [38]
AA 6005-T6 270 134 0.514 0.06 0.497 1.551 0.1398 Børvik et al [39]
Armco iron 175 380 0.32 2.20 5.43 0.47 0.046 Johnson and Cook [21]
4340 steel 792 510 0.26 0.05 3.44 2.12 0.1354 Johnson and Cook [21]
Weldox E 460 (coupled) 490 807 0.73 0.0705 1.732 0.54 0.077 Borvik et al. [40]
Weldox E 460 (uncoupled) 490 383 0.45 0.0705 1.732 0.54 0.058 Borvik et al. [40]
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 427

5 5
n=0.10 kpσy0 = 0.20 n=0.20
4.5 σ /σ = 2.0 4.5 σ /σ = 2.0
f0 y0 f0 y0 kpσy0 = 0.20

4 solid line: exact solution 4 solid line: exact solution


dashdot line: approx. solution dashdot line: approx. solution
3.5 3.5

kpσy0 = 0.10 kpσy0 = 0.10


3 3

kpσy0 = 0.05
μp

μp
2.5 kpσy0 = 0.05
2.5

2 2

1.5 1.5

1 1

0.5 0.5

0 0
–3 –2 –1 0 1 2 3 –3 –2 –1 0 1 2 3
p / σy0 p / σy0

5 5
n=0.30 n=0.50
4.5 σ / σ = 2.0 4.5 σ / σ = 2.0
f0 y0 f0 y0

4 solid line: exact solution 4 solid line: exact solution


dashdot line: approx. solution dashdot line: approx. solution
3.5 3.5

kpσy0 = 0.20 kpσy0 = 0.20


3 3
kpσy0 = 0.10 kpσy0 = 0.10
μp

μp

2.5 2.5

kpσy0 = 0.05 kpσy0 = 0.05


2 2

1.5 1.5

1 1

0.5 0.5

0 0
–3 –2 –1 0 1 2 3 –3 –2 –1 0 1 2 3
p / σy0 p / σy0

Fig. 4. The pressure sensitivity functions lp for several hardening exponent values n for fixed value of rf0 =ry0 ¼ 2:0. Solid lines: exact solutions; dash-dot
lines: approximate solutions. (The solid line and dash-dot line overlap when n = 0.1.)

range of moderate pressure, with the linear fracture stress relationship, the pressure sensitivity function can be derived with
a given hardening rule for the matrix.
For generalized tension condition, the fracture stress can be written as
   n
ef0 p
rf0 ¼ ry0 1 þ 1  q log 1  : ð24Þ
e0 plim
Taking derivative of Eq. (5) with respect to p, at p ¼ 0, the slope is

@ rf rf0 nqef0
¼ : ð25Þ
@p p¼0 plim ðef0 þ e0 Þ

When e0  ef0 , Eq. (25) reduces to



@ rf rf0 nq
¼ ; ð26Þ
@p p¼0 plim

or
nq nqry0
kp ¼ or kp ry0 ¼ : ð27Þ
plim plim
For aluminum alloy 2024-T351, kp ry0  0:10. One might perhaps wonder the relationship between the material constant
kp and the yield stress rf0 for common metals. Here, the test results for armor steels under tension superimposed by high
confining pressures by Bridgman [29] are analyzed to determine the likely range of kp .
Bridgman tested a series of armor steels to determine the pressure dependence of fracture strain on superimposed
high pressure [29]. His results have been analyzed and fitted using the logarithmic function for pressure sensitivity [22].
428 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

5 5
n=0.20 n=0.20 kpσy0 = 0.20
kpσy0 = 0.20
4.5 σ / σ = 1.2 4.5 σ / σ = 1.5
f0 y0 f0 y0

4 solid line: exact solution 4 solid line: exact solution


dashdot line: approx. solution dashdot line: approx. solution
3.5 3.5
kpσy0 = 0.10 kpσy0 = 0.10
3 3

kpσy0 = 0.05 kpσy0 = 0.05


μp

μp
2.5 2.5

2 2

1.5 1.5

1 1

0.5 0.5

0 0
–3 –2 –1 0 1 2 3 –3 –2 –1 0 1 2 3
p / σy0 p / σy0

5 5
n=0.20 kpσy0 = 0.20 n=0.20 kpσy0 = 0.20
4.5 σf0 / σy0 = 2.5 4.5 σf0 / σy0 = 5.0

4 solid line: exact solution 4 solid line: exact solution


dashdot line: approx. solution dashdot line: approx. solution
3.5 3.5
kpσy0 = 0.10 kpσy0 = 0.10
3 3

kpσy0 = 0.05 kpσy0 = 0.05


μp

μp

2.5 2.5

2 2

1.5 1.5

1 1

0.5 0.5

0 0
–3 –2 –1 0 1 2 3 –3 –2 –1 0 1 2 3
p / σy0 p / σy0

Fig. 5. The pressure sensitivity functions lp of different values of the fracture to yield stress ratio rf0 =ry0 for fixed hardening exponent n. Solid lines: exact
solutions; dash-dot lines: approximate solutions. (The solid line and dash-dot line overlaps for rf0 =ry0 ¼ 2:5 and 5.0.).

4
Wilkins et al (1980)
present model
3.5

2.5
μp

1.5

0.5

0
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
p/σy0

Fig. 6. A comparison of two set of materials parameters calibrated for aluminum alloy 6061-T561 in the pressure sensitivity function: the parameters
ða ¼ 10 and a ¼ 0:00024 MPa1 Þ for the present model and the parameters ða ¼ 1:8 and a ¼ 0:00133 MPa1 Þ calibrated by Wilkins et al. [41].

Bridgman provided only a linearized plastic hardening modulus (h) for the materials he tested; the hardening exponent n is
not available. We present the results for qry0 =plim in Table 4 and plotted these values versus the fracture strains ef0 in Fig. 7. It
is found that this values lie between 0.5 and 5.0 for most of the armor steels tested by Bridgman.
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 429

Table 4
Material parameters to account for the effect of the hydrostatic pressure from Bridgman’s tests, after Bridgman [29] and Xue [22].

Material ry0 (MPa) h (MPa) ef0 rf0 (MPa) plim (MPa) q qry0 =plim
1-0 690 427 1.187 1197 997 1.500 1.037
2-0 896 352 1.142 1298 1957 3.413 1.563
2-1 724 538 1.181 1359 4624 7.172 1.123
2-2 724 462 0.947 1161 1511 2.124 1.018
2-3 758 365 0.664 1001 749 1.338 1.356
2-5 1655 669 1.157 2429 3959 2.868 1.199
2-6 1862 586 1.120 2518 667 0.757 2.112
2-7 1145 558 1.254 1845 2250 2.240 1.139
3-0 1034 476 0.636 1337 2576 3.589 1.441
4-0 1034 531 0.417 1256 1400 3.382 2.498
4-1 1103 545 0.566 1412 1963 3.936 2.211
4-2 827 572 0.451 1086 2863 6.508 1.881
4-3 862 634 0.210 995 1783 4.967 2.401
4-5 2069 600 0.308 2253 1809 4.062 4.645
5-0-Longitudinal 772 476 1.359 1419 4686 3.978 0.656
7-0-Rolling 965 434 1.407 1576 1764 1.782 0.975
9-2 1241 690 1.198 2067 2380 2.202 1.148
9-3 931 441 1.107 1419 1974 2.166 1.021
9-4 738 455 1.163 1267 2936 3.874 0.973
9-6 965 503 1.224 1581 13316 12.013 0.871
10-1 869 290 0.125 2219 3153 13.419 8.805
16-0 2069 1207 2.036 1458 3935 2.548 0.562
19-1 1655 848 0.854 2379 3327 3.014 1.499
19-2 2069 1138 0.718 2885 6214 4.926 1.640
19-3 2896 1310 0.340 3341 3018 3.500 3.358
19-4 2841 1393 0.240 3175 3971 6.312 4.516

1
10

qσy0/plim=1/εf
lim
qσ /p

0
10
y0

–1
10 –1 0 1
10 10 10
ε
f0

Fig. 7. The pressure sensitivity parameters versus the reference fracture strain for several high strength steels. Test results from Bridgman [29]; curve fitting
results from Xue [22].

For the tested armor steels, the calculated results of the pressure dependence parameters show a monotonic decaying
trend with respect to the reference fracture strain ef0 . This trend can be fit by a power law curve, i.e. the inverse of the ref-
erence fracture strain qry0 =plim ¼ 1=ef0 . Knowing the approximate solution kp ¼ nq=plim Eq. (27) for ductile materials, we
found the follow relationship to be useful in estimating the pressure sensitivity factor kp for high strength steel,
n
kp ¼ : ð28Þ
ry0 ef0
From Eq. (28), we can draw two conclusions on the kp dependence on the material stress–strain relationship. Firstly, the
more hardenable (higher hardening exponent n) the material is; the more pressure sensitive on equivalent fracture stress is
(higher kp ). Secondly, the more energy absorption the material can take before fracture occurs (indicated roughly by the
product of the initial yield stress and the reference fracture strain ry0 ef0 ); the lower the material constant kp is.
430 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

1000 4.5
logarithmic strain curve – Xue (2007) logarithmic strain curve – Xue (2007)
950 linear stress curve (kp=0.000347) 4 linear stress curve (kp=0.000347)
linear stress curve (kp=0.000420) linear stress curve (kp=0.000420)
900
3.5
850
3
σf (MPa)

800
kpσf0σy0 2.5

εf
750
1
2
700
1.5
650
1
600

550 0.5
AA 2024–T351 AA 2024–T351
500 0
–2.5 –2 –1.5 –1 –0.5 0 0.5 1 1.5 2 2.5 –2.5 –2 –1.5 –1 –0.5 0 0.5 1 1.5 2 2.5
p/σ p/σ
y0 y0

Fig. 8. The stress and strain representation of the pressure sensitivity curves for the matrix of aluminum alloy 2024-T351. Solid line: logarithmic curve for
fracture strain [22]; dash-dot line and dash line: linear pressure dependence curves for fracture stress of matrix at two different slopes.

In the moderate range of stress triaxiality (T 2 ½2:5; 2:5), the representations of the fracture envelope based on the in-
verse fracture stress method and the direct fracture strain calibration method are close to each other. The pressure sensitivity
curves are plotted for both stress and strain representation in Fig. 8 for aluminum alloy 2024-T351. Because of the phenom-
enological nature of the description of fracture envelope, both representations are within experimental and calibration errors
and suitable for many practical applications. The derived kp value is 0:000347 MPa1 , however, it was found
kp ¼ 0:00042 MPa1 is more close to the fracture stress envelope. Lines with these two different stress slopes are both plot-
ted for the matrix compression stress–strain curves in Fig. 8 for comparison. The reference fracture stress for matrix rf0 is
760 MPa.

5. Joint effects of the pressure and the Lode angle on fracture envelope

Assuming the fracture envelope is defined by a modified Tresca condition, the fracture loci resemble a right hexagon on
an octahedral plane and a straight line on a meridional plane. The octahedral stress loci are self-similar in the principal stress
space at various pressures, thus, the fracture envelope in the principal stress space can be written as the product of a refer-
ence value and two dependence functions of the fracture stress, i.e.
rM ¼ rf0 yp ðpÞyh ðhL Þ; ð29Þ

where yp and yh are defined in Eqs. (6) and (16), respectively.


Following the same procedure, the fracture envelope can be now presented by the a joint function of the pressure and the
Lode angle, i.e.
8 9
< 1=n " pffiffiffi #1=n =
rf0 3
ef ¼ e0 ð1 þ kp pÞ 1 ; ð30Þ
: ry0 2 cos hL ;

or be expressed in a joint pressure and Lode angle dependence function, i.e.


 1=n h pffiffi i1=n
rf0
ry0 ð1 þ kp pÞ 2 cos3hL 1
lph ðp; hL Þ ¼  1=n : ð31Þ
rf0
ry0 1

In Eq. (30), the fracture strain envelope is expressed in only two parameters, i.e. rf0 and kp , in addition to the three param-
eters used to describe the stress–strain relationship of the matrix, i.e. ry0 ; n; e0 . In the previous model [14], four material
parameters are needed: q and plim for the pressure sensitivity and c for the Lode angle dependence plus a reference strain
ef0 . Here, due to the linearity of the fracture stress on the pressure and the predetermined shape in the maximum shear con-
dition, we are able to reduce the number of model parameters to be calibrated from four to two and still keep reasonable
accuracy by utilizing the stress–strain parameters. Due to the simplicity of the present model, the calibration procedure
can be significantly reduced.
When ðrf0 =ry0 Þð1=nÞ  1, Eq. (31) reduces to the multiplication form:
pffiffiffi !1=n
1=n 3
lph ðp; hL Þ ¼ lp ðpÞlh ðhL Þ ¼ ð1 þ kp pÞ : ð32Þ
2 cos hL
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 431

This multiplication form was suggested previously by hypothesis, such that the pressure sensitivity and Lode angle
dependence curves of the fracture envelope expressed in strain can be found separately [14]. In the present section, the start-
ing point was the modified Tresca condition, which can be expressed in the multiplication form in terms of the equivalent
stress. These two forms do not differ from each other in the moderate range of pressures as long as the condition
ðrf0 =ry0 Þð1=nÞ  1 is satisfied.
Following the same derivation to obtain Eq. (9) and replacing rf0 with rf0 yp ðpÞ, we find the material parameter c is no
longer a constant. It is now a function of the pressure:
 
rf0 yp pffiffi3 1=n
ry0 2 1
cp ðpÞ ¼  1=n : ð33Þ
r f0 yp
ry0 1

For a given ry0 =rf0 , the above function cp increases monotonically with increasing hydrostatic pressure (and hence re-
duced
pffiffiffi Lode angle dependence of the fracture strain), because yp is non-negative. The value cp reaches asymptotically to
ð 3=2Þ1=n when yp approaches infinity, which means when the pressure is so high the difference in fracture strain in general-
ized tension and generalize shear reduces, pffiffiffi but this value is highly depends on the hardening exponent n. On the other hand,
cp becomes zero when ry0 yp =rf0 ¼ 2= 3, which indicates the material has zero ductility when subjected to generalized
shear loads. Further reduction in yp will result in a brittle fracture when the deviatoric stress state is close to generalized
shear and a ductile fracture (quasi-brittle indeed) for deviatoric stress-states close to generalized tension or generalized
compression. When the pressure factor yp reduces to a value below the threshold of yp ¼ ry0 =rf0 , the material breaks in a
brittle manner for all deviatoric stress states.
The relationship of cp and the pressure are plotted for kp ¼ 0:00042 MPa1 for aluminum alloy 2024-T351 and several
other kp values in Fig. 9, where rf0 =ry0 ¼ 2:52. In positive pressure region (right half of Fig. 9), the cp value stabilizes and
reaches the asymptotic value, while at the high stress triaxiality region (left end of Fig. 9, the cp value decrease considerably.
The higher the kp value of the material, the faster the cp value drops.
The three-dimensional fracture strain envelope is shown in Fig. 10 for aluminum alloy 2024-T351. The pressure sensitiv-
ity function is plotted as red dash lines for hL ¼ p=6 and p=6. The Lode angle dependence function is plotted for the pres-
sure to yield stress ratio [2, 1, 0, 1, 2] from bottom to top in Fig. 10 as solid blue lines.
The fracture envelope can also be expressed in other phenomenological forms. For example, Coulomb–Mohr type of frac-
ture stress envelope can be expressed by
6c cos / 3 þ sin /  p 
rM ¼ rf0 yh ðhL Þyp ðpÞ ¼ pffiffiffi 1 þ tan / ; ð34Þ
3 þ sin / 2 3 cos hL  2 sin / sin hL |fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl} c
|fflfflfflfflfflffl{zfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
rf0 yp ðpÞ
yh ðhL Þ

where c and / are the cohesion and the friction angle, respectively. Similarly, the fracture strain envelope can be inversely
obtained from the stress–strain relationship of the matrix. In this case, the fracture strain envelope is not symmetric with
respect to hL ¼ 0 on an octahedral plane. Substitute kp ¼ 0:00042 MPa1 and rf0 ¼ 760 MPa in Eq. (34), we can obtain
tan / ¼ 0:1735 and c ¼ 413 MPa. The deviatoric eccentricity e ¼ ð3  sin /Þ=ð3 þ sin /Þ ¼ 0:8936.
The Lode angle dependence curve can be plotted for aluminum alloy 2024-T351 for the modified Tresca type and the Cou-
lomb–Mohr type of fracture stress envelope in the triaxial plastic strain component space, as shown in Fig. 11. It can be seen

kp=0.0002
0.44 kp=0.0003
0.43

0.42

0.41
γp(p)

0.4 k =0.00042
p

0.39
kp=0.0005
0.38

0.37

0.36

0.35
–3 –2 –1 0 1 2 3
p/σy0

Fig. 9. The dependence of cp with respect to the pressure p. Plotted are for kp ¼ 0:00042 MPa1 for aluminum alloy 2024-T351 and several other kp values.
432 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

σm

μp

μθ

O’
ε3
ε2

O’’

ε1

Fig. 10. The three-dimensional fracture strain envelope for aluminum alloy 2024-T351 using the present model. Red dash lines represent pressure
dependence function and blue solid lines represent the Lode angle function. The symbols e1 , e2 and e3 are the principal plastic strain components. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

ε2 Coulomb–Mohr
modified Tresca
90
2
120 60

1.5

150 1 30

0.5

180 0

210 330
ε ε
3 1

240 300

270

Fig. 11. Comparison of the fracture strain loci for the modified Tresca and the Coulomb–Mohr type of fracture stress envelope on the p-plane (zero mean
stress octahedral plane).
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 433

from Fig. 11 that under the assumption of a Coulom–Mohr type of fracture stress envelope, the fracture strain in generalized
compression is about doubled than the modified Tresca type of fracture stress envelope for aluminum alloy 2024-T351.

6. Numerical results

The numerical integration algorithm is developed and implemented as a user subroutine for the damage plasticity model
in LS-DYNA. Here, we present four numerical simulations results that are representative for different pressures and Lode an-
gles in a laboratory setup. These four loading cases are: (a) an un-notched tensile round bar; (b) a notched tensile round bar;
(3) a plane strain test; and (d) two upsetting tests of solid cylinder. The experimental procedure is well established for these
laboratory scale tests. These simulations are based on the tests using aluminum alloy 2024-T351. These tests are reported in
Xue and Wierzbicki [15]. Parallel numerical studies are conducted in order to calibrate directly the material constants for the
fracture strain envelope in Ref. [15]. In the present study, we adopt the same meshes as previously. The simulation results
obtained herein are similar to those obtained in Ref. [15]. The calibrated material constants for the fracture stress envelope is
used in these simulation. Comparisons of the numerical results with experiments are focused on two aspects, i.e. the fracture
pattern and the load–displacement curve. The results are presented below.

6.1. Un-notched tensile round bar

First, we present the un-notched tensile round bar simulation. Axisymmetric elements are used. The diameter of this
round bar is 9 mm and mesh size is 0.14 mm square in the necked region. The deformation sequence is shown in Fig. 12.
The simulated crack is a typical cup-cone fracture. Compared with the one obtained using the original damage plasticity
model which employed the calibrated fracture strain envelop, the shear lip in the present study is smaller. The shear lip
has only two elements in the radial direction; while previously three element was found in Ref. [15]. Same trend is found
for the 9 mm notched round bar. The experimental and numerical load–displacement curves are plotted in Fig. 13. The

Fig. 12. The deformation sequences of a 9 mm diameter un-notched tensile round bar showing the formation of the neck and the crack propagation from
the center of the round bar to the outer surface. Plotted are damage contours.

Un–notched round bar


35
Experiments
30

25

Simulation
20
Force (kN)

15

10

0
0 1 2 3 4 5 6
Displacement (mm)

Fig. 13. Comparison of the load–displacement curves from the experiments (dash-dot lines) and the numerical simulation (solid line) for the 9 mm
diameter un-notched tensile round bar.
434 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

downward arrows indicate the experimental fracture points and the numerical simulation curve ends at the fracture initi-
ation point.

6.2. Notched tensile round bar

A 9 mm minimum diameter tensile round bar with a 9 mm notch radius is simulated. Axisymmetric elements are em-
ployed. The elements are 0.14 mm long in the radial direction and are close to a square in the region of interest. The defor-
mation sequence is shown in Fig. 14. A shear lip is formed near surface when the crack propagates from the center of the
minimum cross-section to the root of the notch surface. This cup-cone type of fracture is observed in experiments as well
for the notched round bars. Similar to the un-notched round bar, a smaller shear lip is found in the present simulation than
in Ref. [15]. The experimental load–displacement curves are plotted together with the numerical simulation result in Fig. 15.
Due to the increased mean stress at the notched region, damage is more localized and the fracture occurs much earlier com-
pared with that of the un-notched tensile round bar.

6.3. Plane strain test

A transverse plane strain test is performed using doubly grooved flat plate. The grooved section has a length of 100 mm, a
height of 6 mm (including a 0.5 mm radius transitional area on each ends to the outside of the double grooves) and a thick-
ness of 2.5 mm. The specimen is modeled using plane strain elements. The damage evolution in the grooved section is plot-
ted in Fig. 16. The damage in the plane strain section localizes in two shear bands about diagonal to each other. One of the
shear bands dominates and leads to the final crack. The crack initiates at the center of the specimen where lateral constraint
results in higher mean stress at the center. A comparison of the experimental and the simulated load–displacement curves of
the doubly grooved flat plate is shown in Fig. 17.

Fig. 14. The deformation sequences of a tensile round bar with a 9 mm notch radius. Plotted are damage contours.

Notch radius 9mm


40

35

30
Simulation
25
Experiments
Force (kN)

20

15

10

0
0 0.5 1 1.5
Displacement (mm)

Fig. 15. Comparison of the load–displacement curves from the experiments (dash-dot lines) and the numerical simulation (solid line) for the 9 mm notched
tensile round bar.
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 435

Fig. 16. The evolution of the damage contours in the grooved section in a transverse plane strain test.

140

120

Experiments
100

Simulation
Force (kN)

80

60

40

20

0
0 0.1 0.2 0.3 0.4 0.5
Displacement (mm)

Fig. 17. Comparison of the load–displacement curves from the experiments (dash-dot lines) and the numerical simulation (solid line) for the doubly
grooved flat plate.

6.4. Upsetting of cylinders

Upsetting tests of solid cylinder with an original height of 11.25 mm are simulated. The cylinders are compressed be-
tween two frictional plates at friction coefficients of l = 0.02 and l = 1.0 until fully cracked. The low friction coefficient rep-
resents two Teflon sheets at each end of the cylinder and the high friction coefficient represents dry friction condition. Tests
under both friction conditions are performed [15]. The deformation sequences are shown in Figs. 18 and 19. From the defor-
mation sequences, we can see the deformation and damage are uniform in hoop direction at the beginning of the compres-
sion, but the deformation localizes at shear bands under both friction conditions. These shear bands are the initiation sites
for the final cracks. The load–displacement curves for the two friction conditions are compared with the experimental
curves, as shown in Fig. 20. As we can see from Figs. 18 to 20, the case with dry friction condition yields a more significant
barreling effect and the fracture occurs earlier than the case with Teflon friction condition. An inclination angle of 3% to the
vertical axis is introduced to the low friction case to avoid axisymmetry that leads to simultaneous multiple slant cracks. For
dry friction case, no inclination angle is introduced.

Fig. 18. The deformation sequence of an upsetting specimen with an original height of 11.25 mm with a friction coefficient of 0.02. Plotted are the damage
contours.
436 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

Fig. 19. The deformation sequence of an upsetting specimen with an original height of 11.25 mm with a friction coefficient of 1.0. Plotted are the damage
contours.

70
Simulation (μ=0.02)
60
Experiments (Teflon friction)

50
Force (kN)

40
Simulation (μ=1.0)

30
Experiments (Dry friction)

20

10

0
0 1 2 3 4 5 6
Displacement (mm)

Fig. 20. Comparison of the load–displacement curves obtained from the simulations (solid lines) and the experiments (dash-dot lines) for the Teflon and
the dry friction conditions.

The present formulation of ductile fracture allows the damage accumulation to depend on the stress state. Fig. 21 shows
cutaway views of the damage and the plastic strain evolution for a vertical section at diameter for the dry friction condition
case. It can be seen that the plastic strain is the greatest at the center of the specimen, however, the damage is more localized
in the shear bands diagonal in the vertical section. The high pressure at the center of the specimen prevents the damage to
accumulate at the same pace of the plastic strain evolution. The crack starts at the surface edge and propagates through the

Fig. 21. The damage and the plastic strain evolution for a vertical section at a diameter of the cylinder with an original height of 11.25 mm and a friction
coefficient of 1.0.
L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438 437

Fig. 22. The evolution of the vertical stress component in the slip plane where the fracture takes place. The blue zone near surface is the cracked surface.
The central red zone indicates the remaining uncracked portion where the vertical force is redistributed. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

center until a full crack develops. The crack initiation and propagation on the fracture surface can be seen from the redistri-
bution of the vertical stress component in the fracture plane as shown in Fig. 22. At the surface near equator of the barreled
cylinder where the mean stress is the highest, the crack propagates faster than that near the edges. The uncracked zone
resembles an oval shape until full crack develops.
In the present formulation of constitutive equations, no length scale is adopted. It is well-known that the mesh size has a
role to play in the finite element simulation of the present type of damage plasticity model. The key is in the softening phase
where the mathematical model can localize in a band of infinitesimal width. However, in reality, this shear band width is
finite. Because no length scale is provided in the present constitutive treatment, the size of element is the intrinsic charac-
teristic length for the numerical calculation. For ductile fracture problems like the one discussed in the present paper, the
softening is significant in the fracture process zone ahead of the crack tip. The readers are referred to the Section 8 in Ref.
[17] on numerical simulations of compact tension test for more detail.

7. Conclusions

Previously, a fracture strain envelope described by the pressure sensitivity and the Lode angle of stress is proposed under
the damage plasticity framework [14]. In the present paper, an alternative method to determine the fracture strain envelope
from an assumed fracture stress envelope and the hardening rule of the matrix is proposed.
Specifically, a Tresca type of fracture surface is adopted firstly and the Lode angle dependence is discussed. This method is
assessed by the previously calibrated data of aluminum alloy 2024-T351 and Clausing’s experimental data on the transverse
plane strain and the axisymmetric tensile round bars. It was found that the derived expression using a Tresca fracture surface
agrees fairly well with Clausing’s data on materials with different hardening characteristics.
Secondly, the pressure sensitivity of fracture stress envelope is presented by adopting Drucker–Prager type of linear pres-
sure dependence function. Focus is given to the determination of the likely range of the material parameter kp using Bridg-
man’s test data and the functional similarity with Wilkins et al’s model.
Thirdly, the joint effects of the pressure sensitivity function and the Lode angle dependence function are also discussed. It
is found the material parameter c, which is previously assumed to be constant, does not hold constant under the present
assumptions. However, the two approaches do not differ from each other much when the condition ðrf0 =ry0 Þð1=nÞ  1 is
satisfied.
Lastly, the new material parameters for the present model are derived from calibrated material constants for aluminum
alloy 2024-T2351. A series of simulations based on the models previously used to calibrate the material constants for the
fracture strain envelope are performed and good agreement with the experiment results is achieved.

References

[1] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth: Part I Yield criteria and flow rules for porous ductile media. J Engng
Mater Technol – Trans ASME 1977;99:2–15.
[2] Chu CC, Needleman A. Void nucleation effects in biaxially stretched sheets. J Engng Mater Technol – Trans ASME 1980;102:249–56.
[3] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a round tensile bar. Acta Metall 1984;32(1):157–69.
[4] Zhang ZL, Thaulow C, Ødegård J. A complete Gurson model approach for ductile fracture. Engng Fract Mech 2000;67(2):155–68.
[5] Benzerga AA. Micromechanics of coalescence in ductile fracture. J Mech Phys Solid 2002;50:1331–62.
[6] Zhang KS, Bai JB, François D. Numerical analysis of the influence of the Lode parameter on void growth. Int J Solid Struct 2001;38:5847–56.
[7] Kim J, Gao X, Srivatsan TS. Modeling of void growth in ductile solids: effects of stress triaxiality and initial porosity. Engng Fract Mech
2004;71:379–400.
[8] Xue L. Constitutive modeling of void shearing effect in ductile fracture of porous materials. Engng Fract Mech 2008;75(11):3343–66.
[9] Nahshon K, Hutchinson JW. Modification of the Gurson model for shear failure. Euro J Mech A: Solid 2008;27:1–17.
[10] Lemaître J. A continuous damage mechanics model for ductile fracture. J Engng Mater Technol – Trans ASME 1985;107:83–9.
[11] Chaboche JL. Continuum damage mechanics: Part I – general concepts. J Appl Mech – Trans ASME 1988;55:59–64.
[12] Chaboche JL. Continuum damage mechanics: Part II – damage growth, crack initiation and crack growth. J Appl Mech – Trans ASME 1988;55:65–72.
438 L. Xue / Engineering Fracture Mechanics 76 (2009) 419–438

[13] Voyiadjis GZ, Dorgan RJ. Framework using functional forms of hardening internal state variables in modeling elasto-plastic-damage behavior. Int J
Plast 2007;23(10–11):1826–59.
[14] Xue L. Damage accumulation and fracture initiation of uncracked ductile solids subjected to triaxial loading. Int J Solid Struct 2007;44:5163–81.
[15] Xue L, Wierzbicki T. Ductile fracture calibration of aluminium alloy 2024-T351 using damage plasticity theory [submitted for publication].
[16] Xue L, Wierzbicki T. Ductile fracture initiation and propagation modeling using damage plasticity theory. Engng Fract Mech 2008;75(11):3276–93.
[17] Xue L, Wierzbicki T. Numerical simulation of fracture mode transition in ductile plates. Int J Solid Struct, in press, doi:10.1016/j.ijsolstr.2008.11.009.
[18] McClintock FA. Plasticity aspects of fracture. In: Liebowitz H, editor. Fracture an advanced treatise, vol. 3. New York and London: Academic Press; 1971.
p. 47–307 [chapter 2].
[19] Lloyd DJ. The scaling of the tensile ductile fracture strain with yield strength in Al alloys. Scripta Mater 2003;48:341–4.
[20] Lemaître J, Chaboche JL. Aspect phénomènologique de la rupture par endommagement. J Méca Appl 1978;2:317–65.
[21] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to various strains, strain rates, temperatures and pressures. Engng Fract Mech
1985;21(1):31–48.
[22] Xue L. Ductile fracture modeling – theory, experimental investigation and numerical verification. PhD thesis, Massachusetts Institute of Technology;
2007.
[23] Clausing DP. Effect of plastic strain state on ductility and toughness. Int J Fract Mech 1970;6(1):71–85.
[24] Hancock JW, Brown DK. On the role of stress and strain state in ductile failure. J Mech Phys Solid 1983;31(1):1–24.
[25] Bao Y. Dependence of ductile crack formation in tensile tests on stress triaxiality, stress and strain ratios. Engng Frac Mech 2005;72(4):505–22.
[26] Benzerga AA, Besson J, Batisse R, Pineau A. Synergistic effects of plastic anisotropy and void coalescence on fracture mode in plane strain. Modell Simul
Mater Sci Engng 2002;10:73–102.
[27] Adams Jr CM. Effective ductility in castings and weldments. In ductility. Metals Park, Ohio: American Society for Metals; 1967. p. 179–97 [chapter 6].
[28] von Karman T. Strength investigations under hydrostatic pressure. Z Ver Deut Ing 1911;55:1749–57.
[29] Bridgman PW. Studies in large plastic flow and fracture. McGraw-Hill; 1952.
[30] Pugh HLlD. Mechanical behavior of materials under pressure, the application of hydrostatic pressure to the forming of metals. Amsterdam: Elsevier;
1970. p. 522–90 [chapter 10].
[31] Lewandowski JJ, Lowhaphandu P. Effect of hydrostatic pressure on mechanical behavior and deformation processing of materials. Int Mater Rev
1998;43(4):145–87.
[32] Goto DM, Koss DA, Jablokov V. The influence of tensile stress states on the failure of HY-100 steel. Metallurg Mater Trans A 1999;30A:2835–42.
[33] Kao AS, Kuhn HA, Richmond O, Spitzig WA. Tensile fracture and fractographic analysis of 1045 spheroidized steel under hydrostatic pressure. J Mater
Res 1990;5:83–91.
[34] Drucker DC, Prager W. Soild mechanics and plastic analysis for limit design. Quart Appl Math 1952;10(2):157–65.
[35] Hancock JW, Mackenzie AC. On the mechanisms of ductile failure in high-strength steels subjected to multi-axial stress-states. J Mech Phys Solid
1976;24:147–60.
[36] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial stress fields. J Mech Phys Solid 1969;17:201–17.
[37] Teng X, Wierzbicki T. Numerical study on crack propagation in high velocity perforation. Comput Struct 2005;83(12–13):989–1004.
[38] Clausen AH, Børvik T, Hopperstad OS, Benallal A. Flow and fracture characteristics of aluminium alloy AA5083-H116 as function of strain rate,
temperature and triaxiality. Mat Sci Engng A 2004.
[39] Børvik T, Clausen AH, Eriksson M, Berstad T, Hopperstad OS, Langseth M. Experimental and numerical study on the perforation of AA6005-T6 panels.
Int J Impact Engng 2005;32(1–4):35–64.
[40] Børvik T, Hopperstad OS, Dey S, Pizzinato EV, Langseth M, Albertini C. Strength and ductility of Weldox 460 E steel at high strain rates, elevated
temperatures and various stress triaxialities. Engng Fract Mech 2005;72(7):1071–87.
[41] Wilkins ML, Streit RD, Reaugh JE. Cumulative-strain-damage model of ductile fracture: simulation and prediction of engineering fracture tests.
Technical report UCRL-53058, Lawrence Livermore National Laboratory; 1980.

View publication stats

You might also like