You are on page 1of 15

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

net/publication/334198786

Effects of the stress state on plastic deformation and ductile failure:


Experiment and numerical simulation using a newly designed tension‐shear
specimen

Article  in  Fatigue & Fracture of Engineering Materials & Structures · July 2019


DOI: 10.1111/ffe.13084

CITATIONS READS

23 2,758

5 authors, including:

Zhang xue Wei Jian-Feng Wen


East China University of Science and Technology East China University of Science and Technology
3 PUBLICATIONS   92 CITATIONS    34 PUBLICATIONS   542 CITATIONS   

SEE PROFILE SEE PROFILE

Xian-Cheng Zhang
East China University of Science and Technology
274 PUBLICATIONS   5,131 CITATIONS   

SEE PROFILE

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

Crack Growth by Grain Boundary Cavitation View project

Nonlinear accumulative damages of creep, fatigue and even oxidation: Technology Innovation of 3D interaction diagram View project

All content following this page was uploaded by Xian-Cheng Zhang on 15 August 2019.

The user has requested enhancement of the downloaded file.


Received: 31 March 2019 Revised: 3 June 2019 Accepted: 3 June 2019
DOI: 10.1111/ffe.13084

SPECIAL ISSUE CONTRIBUTION

Effects of the stress state on plastic deformation and ductile


failure: Experiment and numerical simulation using a
newly designed tension‐shear specimen

Xue‐Wei Zhang1 | Jian‐Feng Wen1 | Xian‐Cheng Zhang1 | Xiao‐Gang Wang2 |


Shan‐Tung Tu1

1
Key Laboratory of Pressure Systems and
Safety (Ministry of Education), School of
Abstract
Mechanical and Power Engineering, East The stress state is one of the most notable factors that dominates the initiation of
China University of Science and ductile fracture. To examine the effects of the stress state on plasticity and duc-
Technology, 200237 Shanghai, China
2
tile failure, a new tension‐shear specimen that can cover a wide range of stress
State Key Laboratory of Advanced
Design and Manufacturing for Vehicle triaxialities was designed. A fracture locus was constructed in the space of duc-
Body, College of Mechanical and Vehicle tility and stress triaxiality for two typical steels based on a series of tests. It is
Engineering, Hunan University, 410082
observed that the equivalent plastic strain at failure exhibits a nonmonotonic
Changsha, China
variation with increasing the value of stress triaxiality. A simple damage model
Correspondence based on the ductility exhaustion concept was used to simulate the failure
Jian‐Feng Wen and Shan‐Tung Tu, MOE
Key Laboratory of Pressure Systems and
behaviour, and a good agreement is achieved between simulation results and
Safety, School of Mechanical and Power experimental data. It is further shown that consideration of fracture locus cover-
Engineering, East China University of ing a wide range of stress triaxialities is a key to an accurate prediction.
Science and Technology, Shanghai
200237, China.
Highlights
Email: jfwen@ecust.edu.cn; sttu@ecust.
edu.cn • A tension‐shear specimen covering a wide range of stress triaxialities is newly
designed.
Funding information
111 Project of China, Grant/Award Num-
• The obtained fracture locus is a nonmonotonic function of the stress
ber: B13020; Shanghai Pujiang Program, triaxiality.
Grant/Award Number: 18PJ1402300; • A simple strain‐based damage model incorporating the obtained fracture
National Natural Science Foundation of
China, Grant/Award Numbers: 11472105, locus can well predict the ductile fracture behaviour.
51505149 and 51875203; National Key • Consideration of fracture locus covering a wide range of stress triaxialities is
R&D Program of China, Grant/Award a key to accurate prediction.
Number: 2018YFC0808800

KEYWORDS
ductile fracture, ductility exhaustion, fracture locus, notched plate, positive stress triaxiality, stress
state

Nomenclature: α, angel between axial direction of the specimen and normal direction of the minimum notch section; E, modulus of elasticity; ν,
Poisson ratio; σ0.2, 0.2% offset yield strength; σtrue, true stress; εtrue, plastic part of true strain; A, B, ε0, material parameters in Swift hardening law;
tn, minimum thickness of the notch; wn, minimum width of the notch; σe, von Mises equivalent stress; σm, mean stress; σ1, σ2, σ3,, principal
stresses; ω, damage parameter; △ω, incremental damage; △εp, equivalent plastic strain increment; εf, equivalent plastic strain at failure; εp,
equivalent plastic strain; η, stress triaxiality; ηini, initial stress triaxiality; ηavg, average stress triaxiality.

Fatigue Fract Eng Mater Struct. 2019;1–14. wileyonlinelibrary.com/journal/ffe © 2019 Wiley Publishing Ltd. 1
2 ZHANG ET AL.

1 | INTRODUCTION wide range of stress triaxialities for Al‐alloy 2024. A shear


specimen, a tension‐shear specimen and a plate with a
The ductile fracture is strongly dependent on the stress circular hole were designed and performed under tensile
state and is caused by at least two different failure mecha- loading to cover low stress triaxialities. Their experimen-
nisms. For example, at high stress triaxialities, the pre- tal results firstly reveal that the fracture locus constructed
dominant mechanism is void nucleation, growth, and in the space of the equivalent strain at failure and stress
internal necking down of the intervoid ligaments, with triaxiality exhibits a nonmonotonic variation. Subsequent
obvious void growth before coalescence.1-4 On the other tests using various specimens for Q460 steel2 confirm that
hand, at low stress triaxialities, the more operative mech- the variation of the fracture ductility is nonmonotonic
anism can be internal shearing of the intervoid ligaments, with stress triaxiality. Yet another series of tests carried
with very limited void growth until the failure occurrs.4-8 out by Fu et al3 using Al‐6061 samples with various
In addition, these two mechanisms may compete or coop- geometries and shapes give different findings with those
erate at intermediate levels of stress triaxiality.9 of Bao26 and Li,2 ie, the values of fracture ductility under
Most previous experiments involving the effect of shear‐dominated loading are larger than those in the ten-
stress state on ductile fracture were mainly focused on sile tests of the round bars. The main disadvantage of the
high levels of stress triaxiality. Tensile tests were carried above tests is that too many different types of specimens
out on the smooth round bars and notched round with various geometries are needed to construct a frac-
bars,2,3,10-12 flat grooved plates,2,12,13 etc. In these tests, ture locus.
the effective plastic strain at failure decreases monoto- Instead of conducting tests on a variety types of speci-
nously with an increase of stress triaxiality. In recent men, by using a special and complicated test set‐up, Mohr
years, studies on the effect of low stress triaxiality on and coworkers27-31 developed a butterfly specimen to
the ductile failure are attracting increasing attention cover different levels of stress triaxiality by rotating the
and effort. Numerical methods were used to simulate specimen axis relative to the loading direction. Using
void behaviour at low stress triaxialities.14 For example, modified butterfly specimens, recently Huang et al32
unit cells containing voids were employed to investigate experimentally and numerically investigated the fracture
the effect of stress state on the void shape by Torki15 process in detail and proposed a fracture locus signifi-
and Luo.16 Their results suggest that the voids tend to cantly distinct from the observation by Bao and
elongate and rotate under shear‐dominated conditions. Wierzbicki.26 In summary, although a variety of tests
To examine the critical strain at different stress states, have been conducted on different specimens, the relation
Kiran17 and Liu18 performed micromechanic analyses between the stress triaxiality and fracture ductility is still
and proposed that the critical strain is a nonmonotonic inconclusive. In addition, most tests are not easy to con-
function with the stress triaxiality. On the other hand, duct due to their limitations, including requirement of
specimens with new geometries were designed and high machining precision of the specimen, coaxiality of
tested under different combined loadings to cover low the fixtures in a testing machine, requirement of special
stress triaxialities. For instance, Barsoum and equipment, and complicated test set‐up.
Faleskog9,19,20 employed double‐notched tube specimen Many different damage models have been proposed to
to conduct torsion‐tension tests for Weldox 420 and simulate ductile failure behaviour. For instance, Xue33
Weldox 960. Their experimental results clearly show and Nahshon and Hutchinson34 incorporated shear
that the variation of the equivalent plastic strain at fail- effects into the Gurson model35 to simulate localization
ure with the stress triaxiality is nonmonotonic, and dif- and fracture under shear‐dominated stress states. Yet,
ferent failure mechanisms are observed under different the applicability of these models is limited by too many
levels of stress triaxiality. Similar tension‐torsion tests material parameters that are required to be calibrated.
were also carried out using a slightly different geometry In contrast, uncoupled ductile fracture models based on
of specimen for different materials in the literature.21-25 the ductility exhaustion concept were developed with
However, one major drawback of these tension‐torsion inherent simplicity by Xue and Wierzbicki,36 Bai and
tests is that the experimental results are influenced not Wierzbicki,37 Mohr and Coulomb,38 and Hooputra
only by the machining precision of the intricate speci- et al.39 However, the fracture loci described used in the
men but also by the coaxiality of the fixtures in a testing above models have already shown significant difference
machine. It is not uncommon that the manufacturing and therefore may predict contradictory results. To better
techniques and testing machine cannot meet the testing understand the ductile fracture behaviour under multiax-
requirement. ial stress states and to get more insight in the fracture
Alternatively, Bao and Wierzbicki26 conducted a series locus, more experimental evidence with complete numer-
of tests on several different types of specimen to cover a ical analysis are still needed.
ZHANG ET AL. 3

In this paper, a relatively simple tension‐shear speci-


men inspired by the butterfly specimen and can cover a
wide range of stress triaxialities is newly designed in
order to construct the fracture locus. By using this speci-
men, the ductile fracture tests do not require special fix-
tures or test set‐up. A simple damage evolution model
based on the ductility exhaustion concept incorporating
the obtained fracture locus is adopted to simulate the fail-
ure behaviour of two typical steels, ie, low‐carbon steel
SM400A and austenitic stainless steel 316L. Simulated
results are compared with experimental data for the
investigated materials. The effect of mesh sizes and frac-
ture locus on FE results is also investigated. In addition,
the fracture surfaces of specimens in all tests are carefully
examined to confirm the effect of stress state on the duc- (A)
tile failure mechanisms.

2 | EXPERIMENT

2.1 | Materials

Two materials used in this study were JIS SM400A and


316L. The former material is a low‐carbon steel similar
to ASTM A570 commonly used in various industries.
The latter is an austenite stainless steel, which is a widely
used structural material due to its good corrosion resis-
tance, toughness, and weldability. The chemical composi-
tions of two materials are given in Table 1.

2.2 | Specimen
(B)
A new tension‐shear specimen inspired by the butterfly
specimen is designed and shown in Figure 1A. Compared
with the use of butterfly specimen, the major advantage
of the use of this specimen is that neither special equip-
ment nor special test set‐up is required to conduct the
tests. α is the angel between axial direction of the speci-
men and normal direction of the minimum notch section.
Different stress triaxialities can be obtained through
changing the values of α. The minimum thickness, tn, of
the notch is 5 mm, and two U‐type notches are
manufactured with a 0.75‐mm notch root radii. The min-
imum width, wn, of the notch is 2.5 mm. A total of 12
values of α were chosen in this study ranging from 0° to
90° (see Figure 1B). Simple force analysis indicates that
the loading is tension‐dominated for 0° ≤ α ≤ 45° and

TABLE 1 Chemical composition of SM400A and 316L (wt%)


(C)
Material C Si Mn P S Gr Ni Mo

SM400A 0.17 0.32 1.4 0.032 0.035 ‐ ‐ ‐


FIGURE 1 A, Geometry and dimension of the tension‐shear
specimen (unit: mm). B, Test specimens. C, Experimental set‐up
316L 0.054 0.54 1.16 0.032 0.001 16.65 10.03 2.05 [Colour figure can be viewed at wileyonlinelibrary.com]
4 ZHANG ET AL.

shear‐dominated for 45° < α ≤ 90°. All specimens were strain, and A, ε0, and N are material parameters. Material
machined from plates (each 200 mm long × 150 mm parameters, A, B, and ε0 in Equation (1) for two materials,
wide × 40 mm thick) along the rolling direction for inves- are also provided in Table 2. The true stress‐plastic strain
tigated materials. Two identical specimens were prepared curves are depicted in Figure 2.
with each values of α in order to guarantee the reliability
of experimental data.
3.2 | Damage evolution models

2.3 | Experimental set‐up The damage evolution model used in numerical simula-
tions was based on ductility exhaustion concept. Similar
The tests were carried out with RPL50 material testing damage evolution models were also employed in other lit-
machine with a 50‐kN load capacity (see Figure 1C). erature.41-46 A damage parameter ω which keeps track of
The axial displacement was monitored by attaching two damage accumulation is defined.
grating micrometers on fixtures fixed to the specimen at Incremental damage △ω is calculated by the ratio of
a distance of 100 mm. The final axial displacement was the equivalent plastic strain increment, △εp, and equiva-
then taken as the average of the measured values of the lent plastic strain at failure, εf, at each Gauss point, deter-
two grating micrometers. All tests were conducted at mined by Equation (2)
room temperature under displacement control, and each
specimen was loaded until ruptured completely. The Δεp
loading rate, 0.25 mm/min, was chosen to ensure Δω ¼ (2)
εf
quasistatic loading conditions.
The damage accumulation ω is calculated using a simple
3 | FINITE ELEMENT linear damage accumulation rule, determined from
PROCEDURES Equation (3)

3.1 | Constitutive model Δεp


ω ¼ ∑ Δω ¼ ∑ (3)
εf
Both materials were assumed to be elastic‐plastic with
isotropic hardening, and the classical J2 flow theory was When ω becomes unity, the corresponding elements
employed to simulate the plastic behaviour. Uniaxial ten- are assumed to lose the load‐carrying capacity and then
sile tests were carried out using smooth round bars with a a “crack” is assumed to be generated. Although there
diameter of 10 mm at room temperature. Mechanical can be different ways to simulate loss of load‐carrying
properties for 316L and SM400A are given in Table 2. capacity, it is achieved simply by reducing the elastic
Since ductile failure initiation is preceded by large local modulus to a very small value of 1 × 10−10 MPa here.
plastic deformation, a true stress‐plastic strain relation is The above procedure is implemented in the ABAQUS
needed in the finite element analysis. The true stress‐ using the user subroutine USDFLD.
plastic strain data directly determined from tensile tests
were used in the subsequent FE simulations before neck-
ing. The true stress‐plastic strain curves was then fitted by
the Swift hardening law40 so that it can be extrapolated to
cover a higher strain level, which could be expressed as

σ true ¼ Aðεtrue þ ε0 ÞN (1)

where σtrue is true stress, εtrue is the plastic part of true

TABLE 2 Mechanical properties and material parameters of


SM400A and 316L

E, σ0.2, A,
Material GPa ν MPa MPa N ε0

SM400A 200 0.3 254 645.83 0.158 −0.018


FIGURE 2 True stress‐plastic strain curves for SM400A and 316L
316L 210 0.3 260 1114.521 0.375 0.013
[Colour figure can be viewed at wileyonlinelibrary.com]
ZHANG ET AL. 5

3.3 | Finite element meshes and boundary where σm is the mean stress and σe is the von Mises
conditions equivalent stress.
The mean stress σm is calculated as
Numerical simulations were performed using the finite
element software ABAQUS 6.10. Three‐dimensional 1
σ m ¼ ðσ 1 þ σ 2 þ σ 3 Þ (5)
models of the tension‐shear specimen were established 3
using eight‐node brick elements with reduced where σ1, σ2, and σ3 are the principal stresses.
integration (C3D8R). A mesh size sensitivity analysis, The von Mises equivalent stress σe is calculated as
which will be illustrated later, was conducted in order
to guarantee the accuracy of the numerical results. The qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
adopted mesh size over the notch was approximately σ e ¼ pffiffiffi ðσ 1 −σ 2 Þ2 þ ðσ 1 −σ 3 Þ2 þ ðσ 2 −σ 3 Þ2 (6)
2
0.05 mm × 0.16 mm × 0.5 mm. Details of the finite ele-
ment meshes are shown in Figure 3. Axial displacement The ideal specimen used to calibrate the fracture locus
was imposed on one end of the specimen while the other would have at least two features, including uniform dis-
end was constrained in all directions. tribution of stress state and strain and ability to achieve
a wide range of stress states.21 In this subsection, finite
element analysis was performed using the J2‐flow theory
4 | R E S U L T S AN D D I S C U S S I O N without considering material failure in order to analyse
the stress state of tension‐shear specimen and determine
4.1 | Analysis of the stress state of the equivalent plastic strain at failure.
tension‐shear specimen Figure 4 shows the distribution of initial stress triaxial-
ity at elastic stage versus the normalized distance for dif-
The stress state can be characterized by stress triaxiality,
ferent values of α. Paths across the minimum notch
which is defined as
section are illustrated in Figure 3. The ηini is found to
increase with a decrease in values of α. The ηini is the low-
σm
η¼ (4) est at the surfaces and reaches a maximum near the cen-
σe
tre for all cases (see Figure 4A). Figure 4B shows that the
ηini is relatively constant at t/tn between 0.1 and 0.9, but
varies slightly near the notch tip when α ≤ 45° and
α = 90°. This would indicate that the notch significantly
affects the distribution of the stress triaxiality near the
notch tip because of local stress concentration induced
by the notch. By contrast, the ηini is relatively homoge-
neous along the width for 45° < α < 90°. It means that
the effect of notch on the stress triaxiality variations
becomes weak when α is from 45° to 90°. In addition,
stress triaxiality can be used as an indicator of degree of
constraint.47,48 For a given material, the degree of con-
straint is mainly dependent on specimen geometry.
Hence, the degree of constraint changes with the values
of α.
Figure 5 compares the range of positive initial stress tri-
axiality covered by different specimens. These specimens
were conducted under various loading conditions, includ-
ing the tensile loading, tension‐torsion loading and biaxial
loading. For the smooth round bar and the dogbone spec-
imen,26 the ηini is only 1/3 and 0, respectively. The range
of initial stress triaxiality is 1/3 ≤ ηini ≤ 5/3 for the notched
pffiffiffi
round bar49 and 1/3 ≤ ηini≤ 1= 3 for the flat grooved
plate1 through changing the notch or groove radius. The
butterfly specimen can cover the ηini from 0 to 0.58 which
FIGURE 3 Finite element meshes of the tension‐shear specimen is achieved by rotating the specimen axis relative to the
[Colour figure can be viewed at wileyonlinelibrary.com] loading direction.27 For the tension‐torsion specimen,
6 ZHANG ET AL.

the range of the initial stress triaxiality is 0 ≤ ηini ≤ 0.9 for


the double‐notched tube specimen9 and 0 ≤ ηini ≤ 0.6 for
the modified Lindholm specimen.21 The range of ηini for
the biaxially loaded specimen is from 0 to 0.83.1 By con-
trast, the range of ηini for the newly designed tension‐
shear specimen is relatively wide ranging from 0 to 0.68.
It should be noted that stress triaxialities especially in
the low regime cannot be covered by the smooth round
bar, the notched round bar, and the flat grooved plate.
Although the ranges of ηini for the biaxially loaded speci-
men and the double‐notched tube specimen are slightly
wider than that of the proposed tension‐shear specimen,
a special biaxial testing machine or high machining preci-
(A) sion of the specimen is required as mentioned in Section
1. In addition, the range of the ηini for the butterfly speci-
men and modified Lindholm specimen is narrower than
that of the proposed tension‐shear specimen, indicating
that the tension‐shear specimen is a better choice.
In general, compared with smooth round bar, notched
round bar, and other specimens, the stress state and
strain distribution in the tension‐shear specimen show
good uniformity. In addition, a wide range of stress triax-
ialities especially including the low regime can be covered
by changing the values of α of newly designed tension‐
shear specimen. Hence, the proposed tension‐shear spec-
imen does have the two features of the ideal specimen as
mentioned in Section 4.1.
Figure 6 depicts the evolution of the stress triaxiality
(B) versus the equivalent plastic strain at the centre point
of the minimum notch section for different values of
FIGURE 4 Distribution of initial stress triaxiality at elastic stage
versus normalized distance for different values of α. A, Along the α. The location of the centre point is illustrated in
thickness. B, Along the width [Colour figure can be viewed at Figure 3. All curves show a similar tendency. Stress triax-
wileyonlinelibrary.com] iality increases initially, followed by a decrease, then

FIGURE 5 Range of positive initial


stress triaxiality covered by different
specimens. ((1) Biaxially loaded
specimen,1 (2) butterfly specimen,27 (3)
double‐notched tube specimen,9 (4) flat
grooved plate,1 (5) dogbone,26 (6) proposed
tension‐shear specimen, (7) smooth round
bar,26 (8) modified Lindholm specimen,21
(9) notched round bar49) [Colour figure
can be viewed at wileyonlinelibrary.com]
ZHANG ET AL. 7

(A)

FIGURE 6 Evolution of the stress triaxiality versus the


equivalent plastic strain at the centre point of the minimum notch
section for different values of α [Colour figure can be viewed at
wileyonlinelibrary.com]

again increases with increasing the equivalent plastic


strain. Similar sharp transitions in the beginning of the
stress triaxiality versus the equivalent plastic strain curves
were also reported for different types of speci-
mens.2,26,32,45,50 The major reason for the sharp transition
is believed to be that the local stress redistribution occurs
in the beginning of yielding or hardening of materials
(Figure S1). Therefore, the stress triaxiality is not constant
during the whole loading history. In order to describe the (B)
variation of the stress triaxiality, average stress triaxiality
FIGURE 7 SEM fractographs taken from the centre of the
ηave is introduced, defined by minimum notch section revealing different ductile failure
mechanisms at different levels of stress triaxiality for SM400A. A,
1 εf σ m ηave = 0.3. B, ηave = 1.0 [Colour figure can be viewed at
ηave ¼ ∫0 dεp (7) wileyonlinelibrary.com]
εf σ e
Figure 7A clearly shows that the shear ductile rupture
is the main failure mechanism at low stress triaxialities.
where εf is the equivalent strain to fracture.
Elongated shallow shear dimples which are along the
Compared with the SM400A, similar distribution of
shear direction can be observed on the fracture surface.
stress state and plastic response can be observed for
The plastic deformation expedite voids nucleation at the
316L across the minimum notch section.
second‐phase particles or inclusions by particle cracking
or interface decohesion.4,9 The matrix and inclusion
4.2 | Effect of stress states on the ductile interfaces lead to local stress concentrations due to differ-
failure mechanisms ence in stiffness, which cause void generation by
debonding between the inclusions and matrix.51 Voids
The fracture surfaces of specimens in all tests were care- then experience very limited growth and substantial
fully examined by the scanning electron microscope shearing at low stress triaxiality regime. The final rupture
(SEM). The aim is to investigate the effect of stress triax- caused by the internal shearing of intervoid ligaments usu-
iality on the ductile failure mechanisms. ally involves at least one shear band. Material separation
Figure 7 shows two typical fracture surface across the shear band eventually causes voids to turn from
fractographs taken from the centre of minimum notch nucleation to coalescence.4 More information on the pro-
section (see Figure 3) at different levels of stress triaxiality cess of shear ductile rupture for different materials can
for SM400A: (A) ηave = 0.3 and (B) ηave = 1.0, which are be observed in situ by SEM.8,51,52 In addition, the average
corresponding to the α = 80°and 0°, respectively. diameter of the voids is found to be approximately 5 μm.
8 ZHANG ET AL.

The fracture surfaces of tension‐torsion tests for two mate-


rials (Weldox 420 and Weldox 960)9 and tension‐shear
tests for materials (Q460 and Al‐6061)2,3 exhibit a similar
failure mode at low stress triaxialities.
Figure 7B shows a completely different failure mode in
comparison with Figure 7A. Many large deep dimples can
be seen on the fracture surface. The voids are mainly
nucleated at the second‐phase particles or inclusions.
The reason to account for void formation by debonding
between the two phases is the incompatibility of mechan-
ical properties between the matrix material and the inclu-
sions.4 Voids then experience substantial growth in the
form of ellipsoid and are elongated in the loading direc-
tion at high stress triaxiality regime. Void growth domi- (A)
nates the failure process and continuously increases
until deformation localization.9 Meanwhile, slip bands
and microcracks take shape along with the void growth.
This stage corresponds to the void coalescence. In conclu-
sion, the dominating failure process is void nucleation,
growth caused by large deformation, and internal neck-
ing down of intervoid ligaments driven the large void coa-
lesce until the rupture. The processes of void growth and
coalescence were well captured in details using laser‐
drilled materials and X‐ray computed tomography by
Weck et al.53 Moreover, the average diameter of the voids
is approximately 20 μm, which is significantly larger than
that at low stress triaxiality regime. Similar observation
can also be seen in Barsoum and Faleskog.9
(B)

FIGURE 8 Equivalent plastic strain at failure versus stress


4.3 | Fracture locus triaxiality. A, SM400A. B, 316L [Colour figure can be viewed at
wileyonlinelibrary.com]
The equivalent plastic strain at the potential failure loca-
tion corresponding to the measured displacement to frac- triaxiality were localized, referred to the centre point of
ture from extensometers was taken as the equivalent the minimum notch section (see Figure 3), and were cal-
plastic strain at failure. For all tests, it is difficult and culated from the FE results using Equations (4) to (7).
not possible to capture the exact initial failure location Two materials exhibit similar fracture behaviour. The
for the tension‐shear specimen. Based on previous stud- ductility shows great dependence on the stress triaxiality.
ies,2,54 the potential failure location of tension‐shear spec- It can be seen that there are two distinct branches. At
imen is assumed to be at the centre point of the minimum high stress triaxiality regime, the equivalent plastic strain
notch section (see Figure 3). The displacement to fracture at failure decreases with an increase in stress triaxialities.
is determined by the force‐displacement results. There is At low stress triaxiality regime, the equivalent plastic
a significant drop in the force‐displacement results in all strain at failure increases with increasing in stress triaxi-
tests. This sharply drop is taken as a mark for the ductile alities. The reason for that may be the different fracture
failure initiation in this study. The equivalent plastic mechanisms in two regimes, which are discussed in
strain at failure at the potential failure location is Section 4.2.
extracted from simulated results for all tests. This method To develop a simple analytical representation of the
to determine the fracture ductility was also used by Bao26 fracture locus, best fit of the present experiment data
and Huang.32 was determined using different expressions. At low stress
The equivalent plastic strain at failure is plotted versus triaxiality regime, a linear expression is used. At high
the initial stress triaxiality and average stress triaxiality in stress triaxiality regime, an exponential expression is
Figure 8A,B for SM400A and 316L, respectively. Both the employed, which has been commonly used in previous
values of initial stress triaxiality and average stress studies.42,43,55,56
ZHANG ET AL. 9

For SM400A, expressions of fracture locus using initial produces the values of lode parameter ranging from −1
stress triaxiality and average stress triaxiality are given in to 0, which is very similar to that for butterfly
Equations (8) and (9), respectively. specimen,21,58 modified Lindholm specimen,21 and
double‐notched tube specimen.9,21 However, none of
 these experimental samples (including our sample) can
1:00*ηini þ 0:99 for 0 ≤ ηini ≤ 0:20
εf ¼ (8) easily yield a universal fracture locus alone. It will be very
1:78* expð−2:17*ηini Þ for ηini > 0:20 interesting to combine results using different samples and
to construct a fracture locus in the space of ductility, stress
triaxiality, and lode parameter.

0:43*ηave þ 0:98 for 0 ≤ ηave ≤ 0:50
εf ¼ (9)
3:29* exp f−2:05*ηave for ηave > 0:50 4.4 | Comparisons of simulated results
with experimental data of ductile failure
For 316L, similar expressions of fracture locus are
given in Equations (10) and (11). Numerical simulations using ductility exhaustion
 damage concept (Equations (2)‐(3)) and fracture locus
1:89*ηini þ 0:96 for 0 ≤ ηini ≤ 0:20
εf ¼ (10) (Equations (8)‐(11)) were performed. Comparisons
1:56* exp f−0:82*ηini for ηini > 0:20 between the simulation results and experimental data
for 316L and SM400A are presented in Figure 9. Simu-

0 :94*ηave þ 0 :91 for 0 ≤ ηave ≤ 0 :50 lation results using the average stress triaxiality expres-
εf ¼ (11) sions are corresponding well with the experimental
2 :34* exp f−1 :10*ηave for ηave > 0 :50
data at different α, demonstrating that the ductility
It should be noted that the fracture locus may depend exhaustion damage model incorporating the obtained
not only on the stress triaxiality but also on the lode angle fracture locus can well replicate the ductile fracture
according to many studies including the pioneer work by behaviour of investigated materials under different
Bai and Wierzbicki.57 Our tension‐shear specimen levels of stress triaxiality. The FE results using the

(A) (C)

(B) (D)

FIGURE 9 Comparisons of the simulated results with experimental data for SM400A and 316L. A, α = 0° for SM400A. B, α = 65° for
SM400A. C, α = 0° for 316L. D, α = 65° for 316L [Colour figure can be viewed at wileyonlinelibrary.com]
10 ZHANG ET AL.

initial stress triaxiality expressions and without consid-


eration of damage are also provided in Figure 9 for
the purpose of comparison. Clearly, simulations without
consideration of damage can partly reproduce the plas-
tic deform behaviour, but are not able to estimate the
displacement to fracture. On the other hand, numerical
simulations using the initial stress triaxiality expressions
underestimate the displacement to fracture in all cases.
Therefore, simulations using the average stress triaxial-
ity expressions are superior to that using the initial
stress triaxiality expressions.

(A)
4.5 | Necessity of consideration of the
effect of low levels of stress triaxiality

It has been found that there is an opposite tendency at


low stress triaxiality regime when compared with high
stress triaxiality regime. To identify the effect of the
shape of fracture locus on simulation results, numerical
simulations with damage were performed using the
fracture locus which is calibrated only by using the
experimental data of high stress triaxiality regime. Com-
parative simulated results are shown in Figure 10. It is
clear that fracture locus has little effect on the FE
results at high levels of stress triaxiality (see Figure 10
A). However, numerical simulations using the fracture
(B)
locus which is calibrated only by using the experimental
data of high stress triaxiality regime overestimate the FIGURE 10 Effect of fracture locus on simulated results for
displacement to fracture at low levels of stress triaxiality 316L. A, α = 15°. B, α = 75° [Colour figure can be viewed at
(see Figure 10B). Therefore, fracture locus has a signifi- wileyonlinelibrary.com]
cant effect on predicting the ductile failure under shear‐
dominated loading. 4.7 | Prediction of experimental data for
different specimens

Additional tensile tests using tension‐shear specimens


4.6 | Effect of mesh sizes on the simulated with 8‐mm thickness were carried out. Compared with
results the specimens (tn = 5 mm) used to calibrate the fracture
locus, specimens with 8‐mm thickness have a different
For the tension‐shear specimen, the stress/strain gradient stress state. Numerical simulations with damage were
induced by the notch does exist, particularly when performed using the same damage evolution model
necking occurs. For the purpose of a rigorous FE (Equations (2) and (3)) and fracture locus (Equations (9)
analysis, we investigated the effect of the element mesh and (11)). Comparisons of the simulated results with
size on the simulated results. Another two different experimental data are shown in Figure 12 for SM400A
mesh sizes, 0.03 mm × 0.16 mm × 0.5 mm and and 316L. A good correlation between FE results and
0.1 mm × 0.16 mm × 0.5 mm, were employed over the experimental data is achieved, which confirms the predic-
notch. Comparisons of simulation results using different tive capability of the fracture locus.
mesh sizes for 316L are shown in Figure 11. It can be Three moments named moments A, B, and C were
found that basically the simulation results using all differ- marked using cross symbol in Figure 12A for SM400A.
ent mesh sizes coincide well with the experimental data. The corresponding contour plot of damage at different
Particularly, when the mesh size is small enough, the dif- moments and fractured specimen photo (tn = 8 mm) are
ference of different FE results can be neglected. shown in Figure 13. It is clear that damage is mainly
ZHANG ET AL. 11

(A)

FIGURE 11 Comparison of simulated results for 316L using


different mesh sizes [Colour figure can be viewed at
wileyonlinelibrary.com]

(B)

(A)
(C)

(D)
(B)
FIGURE 13 Contour plot of damage at different moments and
FIGURE 12 Comparisons of the simulated results with photo of fractured specimen (tn = 8 mm) for SM400A. A, Moment
experimental data. A, SM400A. B, 316L [Colour figure can be A. B, Moment B. C, Moment C. D, Fractured specimen [Colour
viewed at wileyonlinelibrary.com] figure can be viewed at wileyonlinelibrary.com]
12 ZHANG ET AL.

accumulated and concentrated over the notch as can be A C KN O WL ED G EME N T S


observed in Figure 13A to 13C. Progressive necking along
This work was partly supported by National Key R&D
the thickness can be observed from moments A to C,
Program of China under the contract number
demonstrating that the FE simulation can reproduce the
2018YFC0808800 and the National Natural Science Foun-
necking behaviour. The band of maximum damage at
dation of China under grant nos. 51505149, 11472105,
onset of the fracture (corresponding to moment C) coin-
and 51875203 and sponsored by the Shanghai Pujiang
cides well with the boundary of the fractured parts of
Program (grant no. 18PJ1402300). The work was also sup-
the specimen (see Figure 13D). In addition, numerically
ported by the 111 Project of China (no. B13020). The
predicted deformation of the deformed specimen corre-
helpful discussion on the numerical simulations with
sponds well to the final profile of the fractured specimen,
Prof. Alan Needleman at Texas A&M University is greatly
confirming that the simple damage model incorporating
appreciated. The authors also would like to thank Profes-
the obtained fracture locus is able to reasonably predict
sor Youshi Hong and two anonymous reviewers for their
the ductile failure.
encouragement and helpful comments on an earlier ver-
sion of the paper.

5 | CONCLUSIONS
ORCID
A series of tests using the tension‐shear specimen were
Jian‐Feng Wen https://orcid.org/0000-0001-9999-572X
conducted to study the effect of the stress state on the
Xian‐Cheng Zhang https://orcid.org/0000-0003-4128-7602
ductile failure. Parallel numerical simulations of all tests
Shan‐Tung Tu https://orcid.org/0000-0002-8328-1666
were performed using the ductility exhaustion concept
and the fracture locus. The following conclusions can be
drawn. RE FER EN CES
1. Brünig M, Brenner D, Gerke S. Stress state dependence of duc-
1. A tension‐shear specimen is newly designed, which tile damage and fracture behavior: experiments and numerical
can cover a wide range of initial stress triaxialities simulations. Eng Fract Mech. 2015;141:152‐169.
from 0 to 0.68. Neither special equipment nor special 2. Li WC, Liao FF, Zhou TH, Askes H. Ductile fracture of Q460
test set‐up is required. It basically has a uniform steel: effects of stress triaxiality and lode angle. J Constr Steel
stress triaxiality over the notch. Res. 2016;123:1‐17.
2. The fracture locus in the space of the equivalent plas- 3. Li H, Fu MW, Lu J, Yang H. Ductile fracture: experiments and
tic strain at failure and stress triaxiality exhibits a computations. Int J Plasticity. 2011;27(2):147‐180.
nonmonotonic variation. This may be due to the dif- 4. Pineau A, Benzerga AA, Pardoen T. Failure of metals I: brittle
ferent fracture mechanisms under low and high and ductile fracture. Acta Mater. 2016;107:424‐483.
levels of stress triaxiality.
5. Brünig M, Brenner D, Gerke S. Modeling of stress‐state‐
3. A good agreement is obtained between experimental dependent damage and failure of ductile metals. Appl Mech
data and simulated results by using a simple damage Mater. 2015;784:35‐42.
model incorporating the obtained fracture locus. In 6. Jia LJ, Ge HB, Shinohara K, Kato H. Experimental and numer-
addition, the use of the average stress triaxiality is ical study on ductile fracture of structural steels under combined
superior to that of the initial stress triaxiality in the shear and tension. J Bridge Eng. 2016;21(5). 04016008
FE simulations. 7. Wang LN, Shi YD, Zhang YL, Bai Y, Lei S. Ductile‐to‐brittle
4. Numerical simulations using the fracture locus cali- fracture of CP titanium with torsion deformation. Mater Lett.
brated only by using the experimental data of high 2018;217:263‐266.
stress triaxiality regime may overestimate the dis- 8. Gao B, Zhang G, Guo TF, Jiang C, Guo X, Tang S. Voiding and
placement to fracture in some circumstances. Consid- fracture in high‐entropy alloy under multi‐axis stress states.
eration of fracture locus covering a wide range of Mater Lett. 2019;237:220‐223.
stress triaxialities is a necessity to an accurate 9. Barsoum I, Faleskog J. Rupture mechanisms in combined ten-
prediction. sion and shear—experiments. Int J Solids Struct. 2007;44(6):
1768‐1786.
With the newly developed specimen, it can be antici- 10. Algarni M, Bai YL, Choi Y. A study of Inconel 718 dependency
pated that a re‐examination of the stress state on creep on stress triaxiality and lode angle in plastic deformation and
rupture failure under relatively low stress triaxiality could ductile fracture. Eng Fract Mech. 2015;147:140‐157.
be easily realized, which will be a continuation of the cur- 11. Tu S, Ren X, Kristensen TA, He J, Zhang Z. Study of low‐
rent work. temperature effect on the fracture locus of a 420‐MPa structural
ZHANG ET AL. 13

steel with the edge tracing method. Fatigue Fract Eng M. 29. Mohr D, Henn S. Calibration of stress‐triaxiality dependent
2018;41(8):1649‐1661. crack formation criteria: a new hybrid experimental–numerical
12. Yao D, Cai LX, Bao C. A new fracture criterion for ductile mate- method. Exp Mech. 2007;47(6):805‐820.
rials based on a finite element aided testing method. Mat Sci Eng 30. Mohr D, Oswald M. A new experimental technique for the
A‐Struct. 2016;673:633‐647. multi‐axial testing of advanced high strength steel sheets. Exp
13. Kang L, Ge HB, Fang X. An improved ductile fracture model for Mech. 2008;48(1):65‐77.
structural steels considering effect of high stress triaxiality. Con- 31. Mohr D, Marcadet SJ. Micromechanically‐motivated phenome-
struct Build Mater. 2016;115:634‐650. nological Hosford–Coulomb model for predicting ductile
14. Tvergaard V. Behaviour of porous ductile solids at low stress tri- fracture initiation at low stress triaxialities. Int J Solids Struct.
axiality in different modes of deformation. Int J Solids Struct. 2015;67‐68:40‐55.
2015;60‐61:28‐34. 32. Huang XG, Zhou Z, Zhu YZ, Zhu DP, Lu L. Tension–shear
15. Torki ME, Benzerga AA. A mechanism of failure in shear bands. experimental analysis and fracture models calibration on Q235
Extreme Mech Lett. 2018;23:67‐71. steel. Int J Steel Struct. 2018;18(5):1784‐1800.

16. Luo T, Gao XS. On the prediction of ductile fracture by void coa- 33. Xue L. Constitutive modeling of void shearing effect in ductile
lescence and strain localization. J Mech Phys Solids. fracture of porous materials. Eng Fract Mech. 2008;75(11):
2018;113:82‐104. 3343‐3366.

17. Kiran R, Khandelwal K. A triaxiality and lode parameter depen- 34. Nahshon K, Hutchinson JW. Modification of the Gurson model
dent ductile fracture criterion. Eng Fract Mech. 2014;128: for shear failure. Eur J Mech A‐Solid. 2008;27(1):1‐17.
121‐138. 35. Gurson AL. Continuum theory of ductile rupture by void nucle-
ation and growth: part I—yield criteria and flow rules for porous
18. Liu ZG, Wong WH, Guo TF. Void behaviors from low to high
ductile media. J Eng Mater Technol. 1977;99(1):297‐300.
triaxialities: transition from void collapse to void coalescence.
Int J Plasticity. 2016;84:183‐202. 36. Wierzbicki T, Bao YB, Lee YW, Bai YL. Calibration and evalua-
tion of seven fracture models. Int J Mech Sci. 2005;47(4–
19. Barsoum I, Faleskog J, Pingle S. The effect of stress state on duc-
5):719‐743.
tility in the moderate stress triaxiality regime of medium and
high strength steels. Int J Mech Sci. 2012;65(1):203‐212. 37. Bai YL, Teng XQ, Wierzbicki T. On the application of stress tri-
axiality formula for plane strain fracture testing. J Eng Mater‐T
20. Faleskog J, Barsoum I. Tension–torsion fracture experiments—
Asme. 2009;131(2). 021002
part I: experiments and a procedure to evaluate the equivalent
plastic strain. Int J Solids Struct. 2013;50(25–26):4241‐4257. 38. Bai YL, Wierzbicki T. Application of extended Mohr–Coulomb
criterion to ductile fracture. Int J Fracture. 2009;161(1):1‐20.
21. Graham SM, Zhang TT, Gao XS, Hayden M. Development of a
combined tension–torsion experiment for calibration of ductile 39. Hooputra H, Gese H, Dell H, Werner H. A comprehensive fail-
fracture models under conditions of low triaxiality. Int J Mech ure model for crashworthiness simulation of aluminium
Sci. 2012;54(1):172‐181. extrusions. Int J Crashworthines. 2004;9(5):449‐463.
22. Papasidero J, Doquet V, Mohr D. Determination of the effect of 40. Swift HW. Plastic instability under plane stress. J Mech Phys
stress state on the onset of ductile fracture through tension‐ Solids. 1952;1(1):1‐18.
torsion experiments. Exp Mech. 2014;54(2):137‐151. 41. Kim N, Oh C, Kim YJ. A numerical method to simulate ductile
23. Papasidero J, Doquet V, Mohr D. Ductile fracture of aluminum failure of tensile plates with interacting through‐wall cracks.
2024‐T351 under proportional and non‐proportional multi‐ Fatigue Fract Eng M. 2011;34(3):215‐226.
axial loading: Bao–Wierzbicki results revisited. Int J Solids 42. Nam HS, Oh YR, Kim YJ, Kim JS, Miura N. Application of engi-
Struct. 2015;69‐70:459‐474. neering ductile tearing simulation method to CRIEPI pipe test.
24. Li SC, Yu Q, Pu JJ, Chen F. Study on mechanical properties and Eng Fract Mech. 2016;153:128‐142.
acoustic emission characteristics of metallic materials under the 43. Johnson GR, Cook WH. Fracture characteristics of three metals
action of combined tension and torsion. Eng Fract Mech. subjected to various strains, strain rates, temperatures and pres-
2018;200:451‐464. sures. Eng Fract Mech. 1985;21(1):31‐48.
25. Scales M, Tardif N, Kyriakides S. Ductile failure of aluminum 44. Mirone G, Corallo D. A local viewpoint for evaluating the influ-
alloy tubes under combined torsion and tension. Int J Solids ence of stress triaxiality and lode angle on ductile failure and
Struct. 2016;97‐98:116‐128. hardening. Int J Plasticity. 2010;26(3):348‐371.
26. Bao YB, Wierzbicki T. On fracture locus in the equivalent strain 45. Li C, Zhou Z, Zhu Y, Lu L. A unified damage factor model for
and stress triaxiality space. Int J Mech Sci. 2004;46(1):81‐98. ductile fracture of steels with different void growth and shrink-
27. Dunand M, Mohr D. Optimized butterfly specimen for the frac- age rates. Fatigue Fract Eng M. 2018;41(5):1132‐1145.
ture testing of sheet materials under combined normal and 46. Bae KD, Ryu HW, Kim YJ, Kim JS. Comparison of ductile tear-
shear loading. Eng Fract Mech. 2011;78(17):2919‐2934. ing simulation with complex cracked pipe test data. J Press Vess‐
28. Dunand M, Mohr D. On the predictive capabilities of the shear t Asme. 2017;139(1). 011203
modified Gurson and the modified Mohr–Coulomb fracture 47. Wu D, Christian EM, Ellison EG. Influence of constraint on
models over a wide range of stress triaxialities and lode angles. creep stress distribution in notched bars. J Strain Anal Eng.
J Mech Phys Solids. 2011;59(7):1374‐1394. 1984;19(4):209‐220.
14 ZHANG ET AL.

48. Henry BS, Luxmoore AR. The stress triaxiality constraint and 56. Ohata M, Fukahori T, Minami F. Damage model for predicting
the Q‐value as a ductile fracture parameter. Eng Fract Mech. the effect of steel properties on ductile crack growth resistance.
1997;57(4):375‐390. Int J Damage Mech. 2009;19(4):441‐459.

49. Bridgman PW. Studies in large plastic flow and fracture: with spe- 57. Bai Y, Wierzbicki T. A new model of metal plasticity and frac-
cial emphasis on the effects of hydrostatic pressure. US: Harvard ture with pressure and lode dependence. Int J Plasticity.
University Press; 1952. 2008;24(6):1071‐1096.
58. Španiel M, Prantl A, Džugan J, Ru°žička J, Moravec M, Kuželka
50. Kim NH, Oh CS, Kim YJ, Yoon KB, Ma YH. Comparison of frac-
J. Calibration of fracture locus in scope of uncoupled elastic–
ture strain based ductile failure simulation with experimental
plastic‐ductile fracture material models. Adv Eng Softw.
results. Int J Pres Ves pip. 2011;88(10):434‐447.
2014;72:95‐108.
51. Achouri M, Germain G, Dal Santo P, Saidane D. Experimental
characterization and numerical modeling of micromechanical
damage under different stress states. Mater Design. 2013;50: SU PP OR TI NG IN FOR MAT ION
207‐222. Additional supporting information may be found online
52. Gao B, Xiang Q, Guo TF, Guod X, Tang S, Huang XX. In situ in the Supporting Information section at the end of the
TEM investigation on void coalescence in metallic materials. article.
Mat Sci Eng A‐Struct. 2018;734:260‐268.

53. Weck A, Wilkinson DS, Maire E, Toda H. Visualization by X‐ray


tomography of void growth and coalescence leading to fracture
in model materials. Acta Mater. 2008;56(12):2919‐2928. How to cite this article: Zhang X‐W, Wen J‐F,
54. Gruben G, Morin D, Langseth M, Hopperstad OS. Strain locali- Zhang X‐C, Wang X‐G, Tu S‐T. Effects of the stress
zation and ductile fracture in advanced high‐strength steel state on plastic deformation and ductile failure:
sheets. Eur J Mech A‐Solid. 2017;61:315‐329. Experiment and numerical simulation using a
55. Borvik T, Hopperstad OS, Dey S, Pizzinato E, Langseth M, newly designed tension‐shear specimen. Fatigue
Albertini C. Strength and ductility of Weldox 460 E steel at high Fract Eng Mater Struct. 2019;1–14. https://doi.org/
strain rates, elevated temperatures and various stress triaxial- 10.1111/ffe.13084
ities. Eng Fract Mech. 2005;72(7):1071‐1087.

View publication stats

You might also like