Materials and Design 60 (2014) 198–207

Contents lists available at ScienceDirect

Materials and Design

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

Influence of stress triaxiality and strain rate on the failure

behavior of a dual-phase DP780 steel
D. Anderson a,⇑, S. Winkler b, A. Bardelcik a, M.J. Worswick a
a
Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Ave. W., Waterloo, Ontario N2L 3G1, Canada
b
Dana Power Technologies, 656 Kerr St., Oakville, Ontario L6K 3E4, Canada

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

Article history: To better understand the in-service mechanical behavior of advanced high-strength steels, the influence
Received 12 November 2013 of stress triaxiality and strain rate on the failure behavior of a dual-phase (DP) 780 steel sheet was inves-
Accepted 29 March 2014 tigated. Three flat, notched mini-tensile geometries with varying notch severities and initial stress triax-
Available online 8 April 2014
ialities of 0.36, 0.45, and 0.74 were considered in the experiments. Miniature specimens were adopted to
facilitate high strain rate testing in addition to quasi-static experiments. Tensile tests were conducted at
Keywords: strain rates of 0.001, 0.01, 0.1, 1, 10, and 100 s1 for all three notched geometries and compared to mini-
Limit failure strain
tensile uniaxial samples. Additional tests at a strain rate of 1500 s1 were performed using a tensile split
Triaxiality
High strain rate
Hopkinson bar apparatus. The results showed that the stress–strain response of the DP780 steel exhibited
Dual-phase steel mainly positive strain rate sensitivity for all geometries, with mild negative strain rate sensitivity up to
0.1 s1 for the uniaxial specimens. The strain at failure was observed to decrease with strain rate at low
strain rates of 0.001–0.1 s1; however, it increased by 26% for an increase in strain rate from 0.1 to
1500 s1 for the uniaxial condition. Initial triaxiality was found to have a significant negative impact
on true failure strain with a decrease of 32% at the highest triaxiality compared to the uniaxial condition
at a strain rate of 0.001 s1. High resolution scanning electron microscopy images of the failure surfaces
revealed a dimpled surface while optical micrographs revealed shearing through the thickness indicating
failure occurred via ductile-shear. Finite element simulations of the tests were used to predict the effec-
tive plastic strain versus triaxiality history within the deforming specimens. These predictions were com-
bined with the measured conditions at the onset of failure in order to construct limit strain versus
triaxiality failure criteria.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Increasingly stringent automotive fuel economy and greenhouse
gas emission standards are forcing automotive manufacturers to
reduce the weight of their fleets while maintaining or improving
crashworthiness and occupant safety. Large contributors to vehicle
mass are the chassis and body structure, both of which are typically
fabricated from low strength steel. Substitution of high strength
steels into these components permits thinner sections, ultimately
leading to reduced mass. However, the performance of these alloys
must be investigated to support simulations of their formability
during manufacturing and in-service performance (crashworthi-
ness), such as the work by Abedrabbo et al. [1,2] who considered
⇑ Corresponding author. Tel.: +1 519 888 4567x32413.
E-mail addresses: david.anderson@uwaterloo.ca (D. Anderson), sooky.winkler@
dana.com (S. Winkler), abardelc@uwaterloo.ca (A. Bardelcik), worswick@lagavulin.
uwaterloo.ca (M.J. Worswick).
the alloy examined in this paper subjected to axial crush loading.
Among other important findings, Abeddrabo et al. [1,2] found that
dual phase (DP) 780 steel offered considerably higher energy
absorption compared to drawing quality and high strength low
alloy (HSLA) steels; however, their work did not consider the poten-
tial onset of fracture during axial crush deformation which can be
important in higher strength materials.
Several publications [3–5] have investigated the uniaxial
stress–strain behavior of DP780 steel sheet subjected to a variety
of strain rates. Huh et al. [3] tested 1.0 mm sheet at strain rates
between 0.001 s1 and 100 s1. The results showed a reduction
in elongation up to a strain rate of 0.01 s1 followed by increased
elongation up to 100 s1. Minimal increases in strength were
observed from 0.001 to 0.1 s1; however, strength was shown to
increase from 0.1 to 100 s1. Kim et al. [4] tested 1.4 mm sheet
at a quasi-static strain rate of 0.001 s1 and elevated strain rates
between 0.1 s1 and 200 s1 and demonstrated a reduction in true
failure strain up to 1 s1, increasing failure strains up to 10 s1, and

http://dx.doi.org/10.1016/j.matdes.2014.03.073
0261-3069/Ó 2014 Elsevier Ltd. All rights reserved.
 D. Anderson et al. / Materials and Design 60 (2014) 198–207 199

decreasing again up to 200 s1. An increase in strength was also

(a)
observed from a strain rate of 0.1–100 s1. Winkler et al. [5] tested
12.5
1.56 mm sheet at strain rates between 0.001 and 1 s1 and addi-
tionally at 1500 s1 and showed a reduction in true failure strain
and strength up to a strain rate of 0.1 s1 and increases in both 1.75
as strain rate increased to 1500 s1. These three works show sim-
ilar trends in true failure strain and strength; however, only Kim
et al. demonstrated a reduction in true failure strain between
10 s1 and 100 s1. Only Winkler et al. performed tests at 1500 s1.
The effect of stress triaxiality on the ductility of metals has been (b) 20
well documented since the early work of Bridgman [6], McClintock 4
[7], Rice and Tracey [8], Mackenzie et al. [9], and Hancock and Mac- 1
kenzie [10]. Much of this pioneering work was focused on smooth
3.5 0.875 1.75
and pre-notched axisymmetric round bars, flat specimens, and flat,
grooved plates. Recently, there have been many experimental
studies on the effects of stress triaxiality on the failure behavior
12.5
of aluminum [11–13], iron [13], structural steel [14,15], pipeline
steel [16], and mild steel [13,17,18]. Results of these studies were
consistent with earlier findings and showed that increased stress units: mm
triaxiality resulted in a reduction in ductility. It was noted by Mirza y Notch Radius Triaxiality Gauge Length
and Barton [13] that increases in stress triaxiality had a larger 20 17.5 0.36 10.98
effect on steel compared to aluminum. Although some of these x 4 3.5 0.45 4.66
works [13–15,17] performed tests at elevated strain rates up to z – into page 1 0.875 0.74 1.75
10,000 s1, only Mirza and Barton [13] found a transition from duc-
tile to brittle fracture that was dependent on both strain rate and Fig. 1. Schematic of mini-tensile specimens. (a) Uniaxial sample, and (b) flat,
stress triaxiality for mild steel. notched sample.
The purpose of the current work is to investigate the failure
response of an advanced high strength steel alloy, DP780, to vary-   
a y2
ing stress states and strain rates. Although previous work, as cited ry ¼ r ln 1 þ 1 2 ; ð2Þ
above, has investigated the role of strain rate on the failure 2R a
response of DP780 steel under uniaxial conditions, a review of    
the current literature has not identified currently published work a y2
rz ¼ r 1 þ ln 1 þ 1 2 ; ð3Þ
addressing the combined effect of triaxiality and strain rate on 2R a
the failure response of DP780 steel sheet. A better understanding where rx,y,z are the three principal stress components, r
 is the effec-
of this failure behavior may be a significant contributor to tive stress, R is the radius of curvature due to deformation or pre-
improved predictions of in-service performance and crash notching, and a is half the ligament width, where the ligament
response. To this end, one uniaxial and three flat, notched DP780 width is the distance between the notches and is equivalent to
steel specimen geometries have been tested at seven strain rates the gauge width of the uniaxial specimens (1.75 mm in this work).
to characterize the stress–strain response and the limit strains as Varying y permits the stress calculations along the width of the
a function of triaxiality and strain rate. Finite element analysis of specimen. Eqs. (1)–(3) can be combined to determine the mean
the tests was performed to determine the evolution of effective stress, rm, and subsequently the stress triaxiality, g:
plastic strain and triaxiality during deformation. Additionally,
scanning electron microscopy (SEM) images and optical micro- 1
rm ¼ ðrx þ ry þ rz Þ; ð4Þ
graphs were used to correlate the observed stress–strain response 3
to the microstructure.
r
g ¼ m : ð5Þ
r
2. Experimental procedures
Therefore, the triaxiality along the width of a flat specimen is found
1.56 mm DP780 cold-rolled hot-dip galvanized sheet manufac- from:
tured by Dofasco Inc. was used in this investigation. Fig. 1 shows   
1 a y2
the geometry of the uniaxial and flat, notched mini-tensile samples g ¼ þ ln 1 þ 1 2 : ð6Þ
3 2R a
as well as the axis orientations adopted for the present work. The
mini-tensile uniaxial samples were developed by Smerd et al. At the centre of the specimen this reduces to the familiar form of:
[19] for high strain rate testing and were shown to provide uniaxial 
1 a
constitutive data with a stress–strain response that is in agreement g ¼ þ ln 1 þ ; ð7Þ
3 2R
with standard 50 mm ASTM: E8/E8M samples prior to the onset of
necking. which is identical to that found for axisymmetric specimens.
In order to obtain a range of initial stress triaxialities for this The analytical triaxialities reported herein when discussing the
study, several flat, notched tensile samples were designed based measured data were found from Eq. (7) and are referred to as initial
on the approximation proposed by Bridgman [6]. In his work, since triaxiality evolves with deformation and varies throughout
Bridgman proposed the following corrections to the state of stress the specimen. During deformation the radius of the notch and
in a flat (plane strain) specimen undergoing necking: the ligament width will change and the sheet material will neck
through the thickness affecting the triaxial state of the specimen.
   Moreover, Bridgman assumed constant effective strain across the
a y2
rx ¼ r ln 1 þ 1 2 ; ð1Þ specimen, which may not be accurate. Therefore, this evolution
2R a
of the triaxiality and effective strain will be presented and
200 D. Anderson et al. / Materials and Design 60 (2014) 198–207

discussed with the results of the finite element model, below. In
this work three different notch radii were chosen (17.5, 3.5, and
0.875 mm) to obtain three initial triaxialities (0.36, 0.45, and
0.74). The flat, notched specimens will be referred to using a sim-
plified naming convention indicating the notch radius as a multiple
of the ligament half-width, namely 20a, 4a, and 1a. All samples
were fabricated with the rolling direction oriented along the ten-
sile loading axis. In the present work, flat, notched tensile speci-
mens were used since the sheet thickness did not permit the
manufacture of more traditional axisymmetric bars or grooved
plates.
Low strain rate tensile tests were performed on the uniaxial,
20a, 4a, and 1a specimens at nominal strain rates of 0.001, 0.01,
0.1, and 1 s1 using an Instron model 1331 servo-hydraulic test
machine. Specimen elongation was measured using an Instron
model 2620-601 extensometer with a gauge length of 12.5 mm
and this length was used to calculate the engineering strain, e, of
the specimens. The crosshead velocity, v, was calculated for each
mini-tensile geometry to achieve the desired nominal strain rate,
_ according to:
e,
v
e_ ¼ ; ð8Þ
l
where the gauge length, l, (listed in the table of Fig. 1) was taken as
the length of the reduced or notched section of the tensile geome-
try; therefore, the crosshead velocity reduced as the notch radius
reduced. It should be noted that the notched specimens do not
incorporate a region of constant cross-section and, therefore, do
not possess a region of uniform strain. Consequently, the measured
strains of the notched specimens are the average along the gauge
length and will be lower than the peak strain at the minimum cross
section. As a result, the calculated strains and strain rates reported
herein will be reduced compared to the peak values.
Intermediate strain rate testing was performed on the uniaxial,
20a, 4a, and 1a specimens at nominal strain rates of 10 and 100 s1
using a fast servo-hydraulic load frame. An optical extensometer
with a gauge length of 12.5 mm was used to determine specimen
elongation and crosshead velocity using the same methods and
assumptions of the low strain rate tests discussed above.
High strain rate testing was performed for the uniaxial and 20a
specimens at a strain rate of 1500 s1 using a tensile split Hopkin-
son bar (TSHB) apparatus [20]. For the TSHB apparatus, the grips
were directly incorporated into the tensile bars to avoid wave dis-
tortion caused by threads [19,21]. Strain gauges mounted on the
incident and transmitter bar were used to capture the tensile
waves of the TSHB apparatus, which were then converted to strain
using the standard split Hopkinson bar equations [22].
All testing was performed at room temperature. After testing, a
JSM 840 SEM was employed for detailed visual analysis of the fail-
ure surface and an Olympus BH60 optical microscope equipped
with ImagePro Plus 5.1 image analysis software was used to ana-
lyze the fracture profile. True failure strains, ef, were calculated
by measuring the area of the failure surface, Af, using image anal-
ysis techniques and comparing to the original cross sectional area,
A0, according to:
A0
ef ¼ ln : ð9Þ
Af

1000 1000
(a) Uniaxial (b) 20
900 900

800 800
Engineering Stress [MPa]
Engineering Stress [MPa]

700 700

600 600

500 500

400 400

300 300

200 200

100 Measured 100 Measured

Average Average
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.02 0.04 0.06 0.08 0.10
Engineering Strain Engineering Strain

1000 1000
(c) 4 (d) 1
900 900

800 800
Engineering Stress [MPa]
Engineering Stress [MPa]

700 700

600 600

500 500

400 400

300 300

200 200

100 Measured 100 Measured

Average Average
0 0
0.00 0.02 0.04 0.06 0.08 0 0.01 0.02 0.03 0.04 0.05
Engineering Strain Engineering Strain

Fig. 2. Repeat tests and average for a strain rate of 0.001 s1. (a) Uniaxial, (b) 20a, (c) 4a, and (d) 1a.
 D. Anderson et al. / Materials and Design 60 (2014) 198–207 201

3. Observed stress–strain response
Fig. 2(a–d) shows the measured engineering stress–strain
response from the repeat tests, as solid lines, and the average
response of three or more tests, as solid circles, for the uniaxial
and notched specimens at a quasi-static strain rate of 0.001 s1
to demonstrate the scatter in the data. As the figure shows, the
data exhibits good repeatability. In general, the scatter in the mea-
sured stresses between repeat tests was less than 3% of the mean
value for a given strain, the exceptions being the high strain rate
results for which the scatter was 10%. The effect of strain rate on
the measured engineering stress–strain response for the four spec-
imen geometries can be seen in Fig. 3(a–d) from which it can be
seen that DP780 exhibits mostly positive strain rate sensitivity
under the conditions studied (note that the abscissa is different
for each data set to more clearly demonstrate the data trends).
Each curve represents an average of three or more repeat tests.
For the TSHB tests, failure did not occur on the first pulse for the
uniaxial specimens. High-speed photography of these experiments
revealed that deformation progressed past the material’s ultimate
tensile strength until the onset of localized necking on the first
pulse, followed by complete failure on the second pulse; therefore,
the data shown for a strain rate of 1500 s1 in Fig. 3(a) is termi-
nated at the completion of the first pulse. All 20a samples tested
at 1500 s1 failed on the first pulse.
The effect of initial notch severity (triaxiality) and nominal
strain rate on the nominal stresses operative in the tensile
geometries is shown in Fig. 4, which plots the engineering stress
at 2% engineering strain (this strain level was chosen since it cor-
responded to stresses at or below the peak level for all cases).
The error bars seem disproportionately large for the uniaxial and
20a tests at 1500 s1; this larger error was caused by slight ringing
in the TSHB apparatus at low strains, which damped out at higher
strains. The nominal stresses increased by 46% for the 1a case, the
most severe notch, compared to the stress for the uniaxial case at a
strain rate of 1 s1. Corresponding to this increase in strength there
is a marked decrease in ductility, as shown in Fig. 5. It is of partic-
ular interest to note that the uniaxial specimens exhibited very low
(or even negative) strain rate sensitivity for strain rates up to
0.1 s1, as evidenced by the plateau (or decrease) in the stresses
up to this strain rate (Fig. 4). Beyond this strain rate, the stress level
increased with strain rate.
The measured true failure strains (based on the average reduc-
tion in area of three specimens) are shown in Fig. 5 as a function of
true strain rate. As the figure shows, there is an overall trend
towards increased true failure strain as strain rate is increased with
a slight plateau from 1 to 100 s1 for all tested notch geometries.
However, for the uniaxial samples there does appear to be a reduc-
tion in true failure strain with strain rate up to 0.1 s1 and a pla-
teau or mild oscillation up to 10 s1, whereas above this rate,
failure strain increases with strain rate.
The effect of initial triaxiality on the failure of DP780 can be fur-
ther examined by considering Fig. 6 which shows true failure strain
as a function of initial triaxiality. True failure strain is shown to

1000 1000
(a) Uniaxial (b) 20
900 900

800 800
Engineering Stress [MPa]

Engineering Stress [MPa]

700 700

600 600

500 500

400 Uniaxial - 0.001/s 400

20a - 0.001/s
Uniaxial - 0.01/s 20a - 0.01/s
300 300
Uniaxial - 0.1/s 20a - 0.1/s
200 Uniaxial - 1/s 200 20a - 1/s
Uniaxial - 10/s 20a - 10/s
100 Uniaxial - 100/s 100 20a - 100/s
Uniaxial - 1,500/s 20a - 1,500/s
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15
Engineering Strain Engineering Strain

1000 1000
(c) 4 (d) 1
900 900

800 800
Engineering Stress [MPa]
Engineering Stress [MPa]

700 700

600 600

500 500

400 400

300 4a - 0.001/s 300 1a - 0.001/s

4a - 0.01/s 1a - 0.01/s
200 4a - 0.1/s 200 1a - 0.1/s
4a - 1/s 1a - 1/s
100 4a - 10/s 100 1a - 10/s
4a - 100/s 1a - 100/s
0 0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.00 0.01 0.02 0.03 0.04 0.05
Engineering Strain Engineering Strain

Fig. 3. Average engineering stress–strain response of uniaxial and notched DP780 specimens. (a) Uniaxial, (b) 20a, (c) 4a, and (d) 1a.
202 D. Anderson et al. / Materials and Design 60 (2014) 198–207

1000 1.2
Engineering Stress @ 2% Engineering

900 1.0

True Failure Strain

0.8
800
Strain [MPa]

0.6
700

0.4 0.001/s
0.01/s
600 0.1/s Uniaxial
1/s
0.2 10/s 20
Uniaxial 4
500 100/s
20a 1,500/s 1
4a 0.0
1a 0.3 0.4 0.5 0.6 0.7 0.8
400 Initial Triaxiality
0.001 0.01 0.1 1 10 100 1000 10000
True Strain Rate [/s] Fig. 6. True failure strain as a function of initial triaxiality.

Fig. 4. Engineering stress at 2% engineering strain as a function of log true strain

rate.
the edges of the specimens in the optical micrographs. The uniaxial
specimen displays many small voids likely resulting in rapid coa-
lescence and failure. As initial triaxiality increased, the size of the
1.2
voids increased in agreement with the findings of McClintock [7]
and Rice and Tracey [8] and may be due to rapid growth of voids
1.0 nucleated at the early stages of deformation due to the elevated tri-
axiality, followed by nucleation and coalescence of smaller voids
during final fracture. This duplex character of the void size distri-
True Failure Strain

0.8
0.6
0.4
0.2
0.0
0.001 0.01 0.1 1 10
True Strain Rate [/s]
Fig. 5. True failure strain as a function of log true strain rate.
increase with increasing strain rate and decrease with increasing
initial triaxiality. The data does appear to cluster together at strain
rates from 1 to 100 s1 due to the plateau in true failure strains
shown in Fig. 5. Significant gains in true failure strain are observed
at strain rates of 1500 s1.
4. Fracture surfaces
Fig. 7 shows optical images of the failed specimens and SEM
images of the failure surfaces for all four geometries at a strain rate
of 0.001 s1, with the uniaxial case on the left and increasing initial
triaxiality moving towards the right. The optical images (top row)
are shown with the sheet thickness oriented vertically and the
loading direction oriented horizontally, while the SEM images
(middle, bottom rows) are shown with the specimen thickness in
the vertical direction and the specimen width in the horizontal
direction. The middle images are shown at 80 magnification;
whereas, the images on the bottom are subsets of the central
regions of the middle images shown at 2000 magnification. The
image set shows that the specimens failed due to a ductile-shear
mode as evidenced by the dimpled morphology at the center of
the specimens shown in the SEM images and the shear lips at
bution is most prominent for the highest initial triaxiality (1a). It
is also observed that bands of larger voids and cracks formed along
the transverse direction and became more prominent with increas-
ing triaxiality. These bands may be associated with clusters of mar-
tensitic islands within the dual phase microstructure, which were
observed by Winkler et al. [5], that tend to nucleate damage as
deformation progresses and triaxiality increases. Similar morphol-
Uniaxial
20a ogy was observed by Kim et al. [4] when testing TRIP780 steel.
4a Fig. 8 shows optical images of the failed specimens and SEM
1a images of the failure surfaces for the uniaxial and 20a geometries
100 1000 10000 at strain rates of 0.001 s1 and 1500 s1. The optical images (top
row) are shown with the sheet thickness oriented vertically and
the loading direction oriented horizontally, while the SEM images
(middle, bottom rows) are shown with the specimen thickness in
the vertical direction and the specimen width in the horizontal
direction. The middle images are shown at 80 magnification;
whereas, the images on the bottom are subsets of the central
regions of the middle images shown at 1000 magnification. As
can be seen in the SEM images, there does not seem to be a change
in the morphology of the failure surface at the centre of the spec-
imens as strain rate was increased. The failure patterns appear
remarkably similar at low and high strain rate for the 20a geome-
try indicating little to no change in failure mechanism as strain rate
was increased.
5. Finite element model
A finite element model was developed in LS-DYNA to provide
insight into the stress state of the specimens over the range of
strain rates and triaxialities presented in this work. 8-noded, selec-
tively reduced integration, solid brick elements were used
throughout. Symmetry of the loading and geometry along the XZ-
and XY-planes was exploited and as such only one-quarter of the
geometry was meshed to reduce the number of elements and solu-
tion time. Appropriate boundary conditions were applied to the
symmetry faces. Element size was chosen to provide six elements
through the thickness of the modeled geometry with an aspect
 D. Anderson et al. / Materials and Design 60 (2014) 198–207 203

(Uniaxial) (20 ) (4 ) (1 )

z
x

z
y

250µm (top row) 200µm (middle row, 80x magnification) 10µm (bottom row, 2000x magnification)

Fig. 7. Optical and SEM micrographs of failure surfaces for a strain rate of 0.001 s1.

ratio close to unity in the regions of high deformation resulting in a
nominal element size of 0.13 mm. Element size was increased out-
side of high deformation zones to reduce solution time. A velocity
profile was applied to the end of the specimen to match that used
during the experiments. Implicit time step integration was used for
all strain rates. Care was taken to place nodes corresponding to the
location of the extensometer so as to accurately match the dis-
placement history of the experimental tests. Material damage
and failure were not accounted for in the model and isotropic con-
ditions were assumed. Fig. 9 shows the finite element mesh for
each geometry in which the inset images are local magnifications
to illustrate that element quality was high throughout. Table 1 lists
the number of elements for each geometry as well as the average
solution time. All models were solved using double precision on
a 40-core Intel Xeon cluster computer using 2 cores per model.
The average effective stress-effective plastic strain curve found
from at least three repeat uniaxial tests at each strain rate was
used as the material constitutive relationship for the finite element
models. A smoothed piecewise fit was applied to the measured
data up to necking and an extrapolation was applied up to failure
using the procedure described by Ling [23]. The simulations were
terminated when the gauge displacement of the model matched
that of the measured data. Fig. 10 shows the comparison of the pre-
dicted and measured engineering stress–strain response for all of

0.001s -1 Uniaxial 1,500s -1 Uniaxial 0.001s -1 20 1,500s -1 20

z
x

z
y

250µm (top row) 200µm (middle row, 80x magnification) 10µm (bottom row, 1000x magnification)

Fig. 8. Optical and SEM micrographs of failure surfaces at strain rates of 0.001 s1 and 1.500 s1.
204 D. Anderson et al. / Materials and Design 60 (2014) 198–207

20

Uniaxial
z

1
y

Fig. 9. Finite element meshes.

Table 1 1000
Number of elements and average solution time for each geometry.
900
Geometry #Elements Solution time (s)
800
Engineering Stress [MPa]

Uniaxial 5850 450

20a 2206 102 700
4a 1478 99
1a 3048 395 600

500
Experimental - Uniaxial
400
the geometries at a strain rate of 1 s1. The figure shows good Experimental - 20a
Experimental - 4a
agreement between the measured and predicted data nearly until 300
Experimental - 1a
failure, where some divergence of the predictions is observed. A 200 FEA - Uniaxial
damage model was not implemented for this work and may serve FEA - 20a
to improve the agreement as the specimens begin to fail. The pre- 100 FEA - 4a
dictions for the remainder of the strain rates showed a similar level FEA - 1a
0
of agreement with experiment, but are not shown for brevity. 0.00 0.05 0.10 0.15 0.20 0.25
Fig. 11 shows the predicted histories of effective plastic strain Engineering Strain
versus triaxiality for elements located at the specimen mid-thick-
Fig. 10. Comparison of predicted and measured engineering stress–strain response
ness within the notch region of the four geometries at a strain rate
at 1 s1.
of 0.001 s1. The solid lines represent the effective plastic strain his-
tory at three locations along the ligament width, as indicated in the
schematic at the top of the figure. In the model, the simulations
were run until the experimentally observed elongation at failure levels of engineering strain. The engineering strain corresponding
was reached; hence, the termination of these curves corresponds to the observed peak load and onset of failure are also indicated.
to the onset of failure in the experiments. The dashed red1 lines The uniaxial condition displays no variation in effective plastic strain
show the effective plastic strain as a function of triaxiality along as a function of triaxiality until peak load is reached at which point
the ligament from the centre to the outer notch surface at various the onset of necking results in increased triaxiality at the specimen
centre and variation in effective plastic strain across the minimum
cross section. The uniaxial and 20a predictions show that effective
1
For interpretation of color in Fig. 11, the reader is referred to the web version of plastic strain is lower at the notch surface and increases towards
this article. the centre of the specimen. The 4a prediction shows that effective
 D. Anderson et al. / Materials and Design 60 (2014) 198–207 205

20
y 4
1
Notch/Neck Surface
x Middle
CL
Centre

0.8 0.8
Centre
Neck Middle = 22.8% (failure)
0.7 0.7
Surface Centre
= 22.0% Middle = 9.2% (failure)
0.6 0.6 Notch = 9.0%
Effective Plastic Strain

Effective Plastic Strain

0.5 = 21.0% 0.5 = 8.0%
= 20.0%
0.4 0.4 = 7.0%
= 19.0%
0.3 0.3 = 6.0%
= 18.0%
= 17.0% = 5.0%
0.2 0.2 = 4.9% (peak load)
= 16.0%
= 14.8% (peak load) = 4.0%
0.1 0.1 = 3.0%
(a) Uniaxial = 2.0% (b) 20
0.0 0.0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Triaxiality Triaxiality

0.8 0.8
(c) 4 (d) 1
0.7 0.7

0.6 0.6 Notch

Effective Plastic Strain

Effective Plastic Strain

Middle Centre = 5.3% = 3.9%

0.5 (failure) 0.5 Centre
Notch (failure) Middle
= 5.0%
0.4 0.4 = 3.0%

0.3 = 4.0% 0.3

= 2.3%
(peak load)
= 3.0%
0.2 0.2 = 2.0%
= 2.9%
(peak load)
0.1 = 2.0% 0.1 = 1.0%
= 1.0%
0.0 0.0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Triaxiality Triaxiality

Fig. 11. Evolution of effective plastic strain versus triaxiality as predicted by numerical models from the centre to the outer notch surface of the specimen. (a) Uniaxial, (b)
20a, (c) 4a, (d) 1a.

plastic strain is nearly even across the specimen, while for the 1a ing in through the thickness more than through the width. At the
case, the predicted effective plastic strain is highest at the notch sur- most severe initial notch geometry, 1a, the history is well below
face and lowest at the centre of the specimen. This change in location the dashed-dotted line; therefore, the thickness of the material is
of the maximum effective plastic strain indicates that failure more easily reduced compared to the width. This severe notch
becomes more likely to occur at the notch root as opposed to the cen- geometry constrains the material and prevents a reduction in
tre of the specimen as initial notch severity is increased. The variation cross-section through the width, which leads to higher strains
in triaxiality shown in Fig. 11 highlights the limitations of Bridgman’s and an increased likelihood of failure at the notch root. These
analysis since the predicted triaxialities from the finite element observations may help to explain the SEM images of the fracture
model differ considerably from those calculated using Eq. (7). surfaces shown in the middle and bottom rows of Fig. 7. These
Fig. 12 shows the relationship between the predicted thickness images showed increased void link-up within bands along the
and width strains at the centre of the four specimens. The solid width direction as initial notch severity increased and this banding
lines represent the strain history up to peak load (indicated by spanned the entire fracture surface for the highest initial notch
symbols) and the dashed lines represent the remainder of the his- severity. Therefore, it is proposed that this banding of void damage,
tory up to failure. The path of equal thickness to width strain is and subsequent duplex character of the void size, is exacerbated by
included for reference and is represented by the dashed-dotted the constraint effect of the notch.
line. The uniaxial and 20a geometries display nearly the same his- The effective plastic strain history data shown in Fig. 11 was for
tories and remain at or above the path of equal thickness to width quasi-static conditions at a nominal strain rate of 0.001 s1. For
strain, which indicates that material is drawing in through the brevity, the full history data for the remainder of the strain rates
width more than through the thickness. The 4a geometry remains tested in the present work will not be presented; however, the
below the dashed-dotted line and indicates that material is draw- trends in the predicted triaxiality at the centre of the specimens
206 D. Anderson et al. / Materials and Design 60 (2014) 198–207

0.00 1.2
Thickness Strain = Width Strain
Peak Load

Effective Plastic Strain at Failure

-0.05 1.0

0.8
Thickness Strain

-0.10

0.6
-0.15

0.4 0.001/s
-0.20 Uniaxial 0.01/s
0.1/s Uniaxial
4 1 0.2 1/s
10/s 20
-0.25
100/s 4
20 1,500/s 1
0.0
-0.30 0.5 0.6 0.7 0.8
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 Triaxiality at Failure
Width Strain
Fig. 14. Effective plastic strain at failure as a function of triaxiality at failure and
Fig. 12. Thickness versus width strain obtained from numerical models. strain rate obtained from numerical models.

0.9
reduction) with strain rate observed in Fig. 6. As reduction in area
0.8 increases, there is an associated increase in local plastic strain
0.7
within the neck and an increase in triaxiality as the notch severity
associated with the neck increases. The overlap of the data from 0.1
Triaxiality at Failure

0.6 to 100 s1 can be attributed to the plateau of true failure strain
(Fig. 5) and triaxiality (Fig. 13) at these strain rates.
0.5

0.4
0.3
0.2 Uniaxial
20a
0.1 4a
1a
0.0
0.001 0.01 0.1 1 10
True Strain Rate [/s]
Fig. 13. Triaxiality at failure as a function of log true strain rate.
corresponding to the onset of failure in the experiments are dis-
cussed. Fig. 13 shows the predicted triaxiality at specimen failure
as a function of log true strain rate for each geometry. Interestingly,
the predicted triaxiality at failure first increases with strain rate for
all geometries up to 0.1 s1, but then decreases (or plateaus) at ele-
vated strain rates for the three notched geometries. The uniaxial
specimen shows an increase in triaxiality up to 1 s1 followed by
a slight decrease (or plateau) up to 1500 s1. The rationale for this
decrease in triaxiality at elevated strain rates is unclear; however,
this may be due to inertial effects or material rate sensitivity, for
example. Future work will consider this issue and the effect of
strain rate (and triaxiality) on damage rates within the notched
specimens.
Fig. 14 shows the relationship between the predicted effective
plastic strain at failure and triaxiality at failure as a function of
strain rate at the centre of the specimen. The trends in the data
show the familiar decrease in effective plastic strain at failure as
triaxiality increases for all strain rates. The triaxiality of the uniax-
ial specimen is considerably higher than 0.33 at failure due to the
formation of a neck, which promotes higher triaxiality in a similar
manner to pre-notched specimens. As strain rate increases both
effective plastic strain at failure and triaxiality increase, causing
the curves to shift upwards and towards the right in the figure.
Both trends can be attributed to the increase in failure strain (area
6. Discussion
The uniaxial stress–strain response as a function of strain rate in
the present work was found to agree with that of Huh et al. [3] and
Winkler et al. [5]. The present work is consistent with these two
studies in demonstrating negligible increases in strength up to
approximately 0.1 s1 followed by increased strength at elevated
100 1000 10000
strain rates. Results by Kim et al. [4] did show an increase in failure
strain at 10 s1 however, they also report a reduction in failure
strain at higher rates of 200 s1. The current work and studies by
Huh et al. [3] and Winkler et al. [5] did not observe this drop in fail-
ure strain at elevated rates.
The triaxialities and effective plastic strains at failure shown in
Figs. 13 and 14 were taken from the centre element of the pre-
sented finite element model. In this work it was assumed that fail-
ure would occur at the centre of the specimen; however, this is not
always the case, particularly for sharp notches (high initial triaxial-
ity) as discussed by Alves and Jones [17]. Using the centre element
is likely a conservative assumption and permits a consistent
method to extract effective plastic strains at failure and triaxialities
for input into damage models. High-speed photography of crack
initiation and specimen failure would provide a more quantitative
means to identify the failure location and is under investigation for
future work. It is evident from this work and that of previous
researchers e.g. [15,17] that applying Bridgman’s methods may
lead to significant errors in triaxiality due to the evolution of this
parameter with plastic deformation. These errors could negatively
impact the validity of failure models; therefore, the deformation
history should be accounted for when determining failure locus
parameters.
A relationship appears to exist between triaxiality and failure
surface morphology where high initial triaxiality constrains defor-
mation through the specimen width that encourages martensitic
bands to open up, resulting in a duplex character to the void size
distribution. It is felt that this process is partially responsible for
the dramatic decrease in ductility as initial notch severity
 D. Anderson et al. / Materials and Design 60 (2014) 198–207
increases. Recent work by Sun et al. [24], Paul and Kumar [25], and
Paul [26] confirms this observation. Those studies modelled the
plastic localization of dual phase steels using microstructure-based
representative volume element simulations and found that (using
the microstructural heterogeneity as the only source of imperfec-
tion) the failure mode was closely related to the stress state. Shear
dominant failure was predicted under uniaxial tensile conditions
whereas plane strain conditions resulted in fracture along a plane
perpendicular to the principal axis of loading. Similar dependency
of failure mode on stress state was observed in the current exper-
iments; thus, failure mode of this dual phase steel is strongly
linked to stress triaxiality.
The data presented in the present work will be used as the basis
for a phenomenological based damage and failure model. This
work is currently ongoing. Additionally, digital image correlation
techniques will be implemented in future tests to confirm the
strain distributions predicted by the finite element models with
those observed experimentally.
7. Conclusions
The stress–strain response of DP780 steel was shown to be sen-
sitive to both strain rate and initial triaxiality for the range of con-
ditions tested. The sensitivity of the stress–strain response to
strain rate was near-zero or mildly negative for the uniaxial spec-
imens up to strain rates of 0.1 s1, beyond which the stress showed
a positive dependence on strain rate. The sensitivity of the stress–
strain response to strain rate was nearly zero for the 20a specimens
up to a strain rate of 0.01 s1 followed by positive dependence on
strain rate. The 4a and 1a geometries showed positive rate depen-
dence of the stress for all strain rates.
The loci of effective plastic strain at failure versus triaxiality at
failure exhibits a clear trend to shift upwards (higher failure
strains) and towards the right (higher triaxiality) as strain rate
increases.
Microstructural analysis using SEM images revealed little
change in failure surface morphology as strain rate was increased.
Large transverse cracks formed which are thought to initiate at
clusters of martensitic islands. The extent of the transverse crack-
ing increased as initial triaxiality increased. Optical micrographs
revealed the presence of shear lips for all geometries tested. Based
on the SEM and optical micrographs it can be concluded that this
material fails via a ductile-shear process.
Acknowledgements
Financial support from Automotive Partnership Canada (APC),
the Natural Sciences and Engineering Research Council of Canada
(NSERC), Auto21, Ontario Research Fund, and the Canada Research
Chairs Secretariat are gratefully acknowledged.
207
References
[1] Abedrabbo N, Mayer R, Worswick MJ, van Riemsdijk I. Crashworthiness of
advanced high-strength steel tubes: experimental and numerical results. Int J
Impact Eng 2009;36:1044–57.
[2] Abedrabbo N, Mayer R, Worswick MJ, van Riemsdijk I. Optimization methods
for the tube hydroforming process applied to advanced high-strength steels
with experimental verification. J Mater Process Technol 2009;209(1):110–23.
[3] Huh J, Huh H, Lee CS. Effect of strain rate on plastic anisotropy of advanced
high strength steel sheets. Int J Plast 2013;44:23–46.
[4] Kim JH, Kin D, Han HN, Barlat F, Lee MG. Strain rate dependent tensile behavior
of advanced high strength steels: experiment and constitutive modeling.
Mater Sci Eng, A 2013;559:222–31.
[5] Winkler S, Thompson A, Salisbury C, Worswick M, van Riemsdijk I, Mayer R.
Strain rate and temperature effects on the formability and damage of advanced
high-strength steels. Metall Mater Trans A 2008;39A:1350–8.
[6] Bridgman PW. Studies in large plastic flow and fracture cambridge. Harvard
University Press; 1964.
[7] McClintock FA. A criterion for ductile fracture by the growth of holes. ASME J
Appl Mech 1968;35:363–71.
[8] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial stress fields.
J Mech Phys Solids 1969;17:201–17.
[9] MacKenzie AC, Hancodk JW, Brown DK. On the influence of state of stress on
ductile failue initiation in high strength steels. Eng Fract Mech 1977;9:167–88.
[10] Hancock JW, Mackenzie AC. On the mechanisms of ductile failure in high-
strength steels subjected to multi-axial stress states. J Mech Phys Solids
1976;24:147–69.
[11] Bao Y. Dependence of ductile crack formation in tensile tests on stress
triaxiality, stress and strain ratios. Eng Fract Mech 2005;72:505–22.
[12] Bai Y, Wierzbicki T. A new model of metal plasticity and fracture with pressure
and lode dependence. Int J Plast 2008;24:1071–96.
[13] Mirza MS, Barton DC. The effect of stress triaxiality and strain-rate on the
fracture characteristics of ductile metals. J Mater Sci 1996;31:453–61.
[14] Hopperstad OS, Borvik T, Langseth M, Labibes K, AC. On the influence of stress
triaxiality and strain rate on the behaviour of a structural steel. Part I Exp Eur J
Mech A/Solids 2003;22:1–13.
[15] Borvik T, Hopperstad OS, Berstad T. On the influence of stress triaxiality and
strain rate on the behaviour of a structural steel. Part II Numer Study Eur J
Mech A/Solids 2003;22:15–32.
[16] Ghajar R, Mirone G, Keshavarz A. Ductile failure of X100 pipeline steel –
experiments and fractography. Mater Des 2013;43:513–25.
[17] Alves M, Jones N. Influence of hydrostratic stress on failure of axisymmetric
notched specimens. J Mech Phys Solids 1999;47:643–67.
[18] Bai Y, Teng X, Wierzbicki T. On the application of stress triaxiality formula for
plane strain fracture testing. ASME J Eng Mater Technol 2009;131. 021002-1–
021002-10.
[19] Smerd R, Winkler S, Salisbury C, Worswick M, Lloyd D, Finn M. High strain rate
tensile testing of automotive aluminum alloy sheet. Int J Impact Eng
2005;21:541–60.
[20] Thompson AC, Salisbury CP, Worswick MJ, Mayer R. Constitutive modelling of
dual phase steel sheet and tube. J Phys IV 2006;134:281–6.
[21] Smerd RO. Constitutive behavior of aluminum alloy sheet at high strain rates,
Master’s Thesis: University of Waterloo; 2005.
[22] Gray GT. Classic split-Hopkinson pressure bar testing, mechanical testing and
evaluation, vol. 8. ASM Handbook: ASM International; 2000.
[23] Ling Y. Uniaxial true stress–strain after necking. AMP J Technol 1996;5:37–48.
[24] Sun X, Choi KS, Liu WN, Khaleel MA. Predicting failure modes and ductility of
dual phase steels using plastic strain localization. Int J Plast
2009;25:1888–909.
[25] Paul SK, Kumar A. Micromechanics based modeling to predict flow behavior
and plastic strain localization of dual phase steels. Comput Mater Sci
2012;63:66–74.
[26] Paul SK. Real microstructure based micromechanical model to simulate
microstructural level deformation behavior and failure initiation in DP 590
Steel. Mater Des 2013;44:397–406.