ARTICLE IN PRESS

POLYMER
TESTING
Polymer Testing 27 (2008) 284–295
www.elsevier.com/locate/polytest

Test Method

On the measurement of true fracture strain

of thermoplastics materials
Xinran Xiao
General Motors Corporation, MC 480-106-710, 30500 Mound Road, Warren, MI 48090-9055, USA
Received 5 October 2007; accepted 14 November 2007

Abstract

In tensile testing, engineering thermoplastic materials can sustain large deformation after necking. This property can be
advantageous to the manufacturing process and mechanical performance. However, it causes difficulties in strain
measurement and in the determination of true stress and true strain, particularly at higher strain rates. This paper
discussed the measurement of true fracture strain. The true fracture strain was estimated using three different methods: the
area reduction, elongation over the necked zone, and elongation over the gage length. With strain localization, strain
measured by an extensometer becomes gage length dependent. To address the issue, an algorithm was formulated that
estimates the average fracture strain over an arbitrary gage length from the fracture strain measured over the necked zone.
The proposed algorithm provides a way to estimate the mesh dependence of the fracture strain for finite element (FE)
analysis. This paper also presents an analysis-based two-step correction scheme for the crosshead extension to be used to
supplement the strain measurement from the extensometer.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Strain measurement; Strain rates; Tensile testing; Necking; True fracture strain; Thermoplastics

1. Introduction as well as on accurate measurements of large

deformation and fracture behavior of the material.
The growing application of finite element (FE) A major hurdle in the measurement of stress–
analysis in structural performance evaluations, so- strain responses of thermoplastic materials is related
called virtual testing, has led to greater demands for to necking. Engineering thermoplastics can sustain
material constitutive models. To simulate large very large deformation after necking. This behavior
deformation and fracture behavior of a structural has been observed in polyethylene (PE) [1–7],
component, the material model must be valid over polypropylene (PP) [5–7], rubber-toughened poly-
the entire range of the deformation process for the styrene [8], polycarbonate (PC) [7,9,10] and its
service environment and the loading conditions alloys [11].
under consideration. The development of such Fig. 1 presents a tensile engineering stress–strain
constitutive models is based on good understanding curve of a high-density polyethylene (HDPE). The
tensile experiment was conducted at a quasi-static
Tel.: +1 586 986 5616; fax: +1 586 986 3091. rate in accordance with the ASTM tensile test
E-mail address: [email protected] standard for plastics [12]. As shown, the stress–

0142-9418/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.polymertesting.2007.11.007
 ARTICLE IN PRESS
Engineering Stress (MPa) X. Xiao / Polymer Testing 27 (2008) 284–295 285

25.0 peak stress engineering plastics are often represented by mate-

rial models based on metal plasticity. Therefore, the
20.0 necking
fundamental assumptions in the plasticity theory
15.0 have to be observed in the characterization of large
10.0 deformation behavior of thermoplastic materials.
Similar to metals, the formation of necking
5.0 results in strain localization. Necking causes two
0.0 direct problems in the measurement of stress–strain
0.00 0.50 1.00 1.50 2.00 2.50 relations. Firstly, the strain measured by an
Engineering Strain
extensometer becomes gage length dependent. The
Fig. 1. A typical engineering stress–strain curve of an HDPE extensometer measures the average strain over its
material. gage length. Once necking starts, the deformation of
the specimen over the gage length is no longer
homogenous and, thus, the averaged strain value
strain curve can be divided into four regions: an becomes gage length dependent. Secondly, plasticity
initial monotonic rising region up to the peak stress, models require a stress–strain relation given in
a strain softening region after the peak stress, a terms of true stress and true strain. Based on the
region of large deformation under almost constant constant volume assumption of the plasticity
stress, and a rapid softening region leading to theory, the instantaneous cross-sectional area of a
rupture. In the tensile experiment, it was observed specimen with a uniform cross section can be
that the specimen continued to deform uniformly computed from the elongation. Necking nullifies
after peak load. The tensile instability, i.e. necking, this simple relation and, hence, the determination of
started somewhere after the peak load. The speci- the true stress and the true strain becomes an issue.
men did not fracture immediately after necking but To measure the stress–strain relation after neck-
stabilized when the cross section at the necked area ing requires a method that can capture the full
reduced to a certain size. The necking then displacement field. There are a few techniques
propagated and the specimen elongated at a available. High-speed camera imaging has been
nearly constant load, as in a drawing process, used since earlier days. In this case, the surface of
before final fracture. the specimen is divided into small areas by printed
The capability of large deformation of thermo- grid lines or small circles. The local strain is
plastic materials can be advantageous to the calculated after the test from the relative displace-
manufacturing process and mechanical perfor- ment recorded at each image. Another promising
mance. However, it presents a challenge when method is the digital image correlation (DIC)
attempting to model such behavior. The develop- technique. Parsons et al. [4,10] and Fang et al. [11]
ment of constitutive models for large deformation demonstrated the use of DIC for a variety of
behavior of polymeric materials lags far behind that polymers.
for metallic materials. The deformation mechanisms G’Sell and Jonas [6] took an alternative approach
of polymeric materials are known to be very to measure the stress–strain response beyond the
different from those of metallic materials. Instead necking point. Instead of using a standard tensile
of dislocation motion and interaction, twinning, specimen with a constant cross-sectional area over a
grain boundary slippage mechanics in metals [13], gage length, an hourglass shaped cylindrical speci-
the inelastic deformation of a non-crosslinking men was employed. The minimum instantaneous
polymer is attributed to molecular chain reorienta- diameter was recorded and used to compute the
tion, untangling of entangled chains, etc. [14,15]. radial strain. The axial strain was calculated from
Constitutive models based on the framework of the the radial strain following a relation derived from
viscoelasticity and viscoplasticity theories would be the constant volume assumption. The study also
better representations for the non-crosslinking looked into the issue of the actual strain rate in the
polymers. However, the existing viscoelasticity strain localization region. The signal from the
and/or viscoplasticity models do not cover the diameter transducer was used to control the testing
whole deformation process that spans from viscoe- machine to ensure that the test was conducted at a
lastic stretching to final rupture. In large-scale FE constant true strain rate. G’Sell et al. [7] improved
simulations, particularly in dynamic simulations, the strain measurement using video imaging and
 ARTICLE IN PRESS
286 X. Xiao / Polymer Testing 27 (2008) 284–295

established a video-controlled tensile testing meth-

od. With properly located multiple marks, the
video-controlled method has been used to study
polymers displaying diffuse necking with rectangu-
lar specimens [8].
The above techniques require high-speed imaging
and data-processing systems. They are relatively
complex to be used in routine material testing.
Moreover, the speeds of these methods are still
slower than needed in high strain rate testing.
In the absence of a suitable true strain measure-
ment technique for high strain rate testing, an
alternative method has to be sought. One approach
is to utilize the segment of the stress–strain curve
obtained by valid measurement, the final point Fig. 3. Electric-fiber optical extensometer for high strain rate
determined from the ruptured specimen, and measurement at Axel Product Testing Lab (courtesy of Axel
estimate the missing segment through correlating Product).
FE simulations with experimental results of simple
geometries. In this approach, the determination of 0.01 and 0.1 s1, a clip-on extensometer was used.
the true fracture strain from the ruptured specimen The gage length of the extensometer was 12 mm and
is an important step. the range of the extensometer was about 2.5 mm.
This paper discusses the determination of true The crosshead displacement data were used at
fracture strain of thermoplastic materials that dis- extensions beyond the range of the extensometer.
play uniform necking based on strains measured The tensile test at the three higher rates was
with an extensometer, crosshead extension and on conducted with a high rate load frame, and the
ruptured specimens. strain was measured using an electro-fiber optical
extensometer, both developed and built at the Axel
2. Tensile experiments lab. Fig. 3 shows the strain measurement setup. The
optical extensometer traces the movements of two
An HDPE material was used in this investigation. markers adhered on the gage section of the speci-
The tensile experiment was conducted at Axel men. The strain was calculated based on the relative
Product, a testing laboratory specializing in testing displacement between the two markers. The nom-
polymeric materials. The tensile specimen was the inal center-to-center distance between the two
ISO 8256 type-3 specimen (Fig. 2). It had a nominal markers (gage length) was about 11 mm. The linear
gage width of 10 mm, a gage length of 12 mm and a range of the optical extensometer was about 5 mm.
total length of 80 mm. The specimens were cut from
plaques to dimensions by hand routing. 3. Tensile test results
The tensile tests were conducted at five nominal
strain rates: 0.01, 0.1, 1, 10 and 100 s1. For tests at 3.1. Tensile strain measurement

In a tensile test, extension data were collected

l = 80 from the extensometer as well as from the crosshead
le = 30 displacement. The engineering strain e is the average
linear strain, obtained by dividing the elongation of
l0 = 12 the gage length of the specimen, Dl, by its original
length l0 [13]:
15
10

Dl l  l 0
¼ ¼ . (3.1)
l0 l0
r = 20
Since the data from the extensometer were only
Fig. 2. ISO 8256 type-3 specimen (unit in mm), the tensile valid up to a certain range, the crosshead data were
specimen used at Axel Product Lab. used to supplement the strain data. Assuming that
 ARTICLE IN PRESS
X. Xiao / Polymer Testing 27 (2008) 284–295 287

the entire deformation occurred within the gage method. Nevertheless, such corrections resulted in
length of the specimen, the engineering strain can be discontinuities in the strain rate-history plot, as
estimated from the crosshead extension. shown in Fig. 6. In high strain rate testing, using
Fig. 4 compares the engineering stress–strain specimens with different gage lengths can modify
curves using strain obtained from the extensometer the dynamic response of the test [17,18]. In a
and the strain calculated directly from the crosshead previous work [19], a one-step correction method
extension for a test conducted at a nominal strain was used which simply shifts the strain measured by
rate of 0.1 s1. The strain value calculated from crosshead extension vertically to supplement the
crosshead extension is larger because it includes not extensometer data, as illustrated in Fig. 7. The one-
only the gage length elongation but also the step method results in a discontinuity in the strain-
displacement due to the compliance of the load history curve and the estimated fracture strain was
train and the extension in the area of the specimen also lower [19].
outside the gripping area. To eliminate the unwanted contribution from the
The effect of the compliance of the load train may additional compliances and deformation outside the
be corrected by testing specimens with different gage length, a two-step correction scheme was
gage lengths [16]. Fig. 5 plots the engineering stress– proposed.
strain curves constructed using a similar correction
3.2. Two-step correction for crosshead extension

30 The first step is to correct the contribution

strain from EX from the additional compliances included in the
25
Engineering Stress (MPa)

strain from CH
20 0.20

15
Strain rate (1/s)

0.15

10
0.10
5
0.05
0
0 0.5 1 1.5 2 2.5 0.00
Engineering Strain 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Strain
Fig. 4. Engineering stress–strain curves obtained by strain
measurement from the extensometer (EX) and crosshead exten- Fig. 6. The strain rate versus strain plot corresponding to curve 1
sion (CH). in Fig. 5.

30 1.0 1.0
1
25 2 0.8 0.8
3
Stress (MPa)

20 0.6 0.6
Load (kN)
Strain

15
0.4 0.4
10 StrainZimmer
Strain Zimmer
0.2 Strain(cross
(crosshead) 0.2
Strain head)
5 Load
Load
0.0 0.0
0 0.0035 0.004 0.0045 0.005 0.0055 0.006
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 -0.2 -0.2
Strain Time (s)

Fig. 5. Typical engineering stress–strain curves obtained by Fig. 7. Supplement the strain measured by the optical extens-
combining strain measurement from the extensometer and ometer (Strain Zimmer) with the strain calculated from crosshead
crosshead extension using the conventional correction method. displacement.
 ARTICLE IN PRESS
288 X. Xiao / Polymer Testing 27 (2008) 284–295

crosshead extension. This can be estimated by 1.0

strain from EX
comparing the compliance obtained from the cross- strain from CH
0.8
head extension with that from the extensometer at correction 1
small deformation. As illustrated in Fig. 8, in the 0.6 correction 2

Strain
linear range, the difference in the compliance of the
0.4 ε2CH’
two stress–strain curves is
1CH 1EX 0.2 ε2EX
a ¼ cCH  cEX ¼  , (3.2)
s1 s1 0.0
0 10 20 30 40 50
where e1EX and e1CH are the strains measured by TIME (s)
extensometer and by crosshead extension at an
arbitrary stress s1. cEX and cCH are the compliances Fig. 9. Two-step correction scheme: step 2—determining the
of the stress–strain curve measured by the extens- strain values corresponding to the last valid point of extens-
ometer measurement for the calculation of scale factor b. It
ometer and by the crosshead extension, respectively.
illustrates the crosshead strain after correction 1 and correction 2.
a is the difference in compliance between the two
measurements and, hence, can be used as a
compliance correction factor. value is obtained by a second correction via a scale
The strain from the crosshead is corrected by factor b. b is the ratio of the strain values:
subtracting the contribution of the additional 2EX
compliance b¼ , (3.4)
02CH
0CH ¼ CH  as, (3.3)
where e2EX is the strain measured by extensometer
where eCH is the strain calculated directly from the and 02CH is the strain from the crosshead after
crosshead extension and 0CH is the corrected strain. correction 1. To fully utilize the measurement from
Fig. 9 compares the strain from extensometer, the extensometer, the values corresponding to the
crosshead and correction 1 (compliance correction). last valid point of extensometer measurement
Fig. 9 shows that the curve labeled as ‘‘correction 1’’ should be used, as indicated in Fig. 9.
is much closer to the curve labeled as ‘‘strain from The second correction is obtained by
EX’’ as compared to the curve without correction,
00CH ¼ b0CH , (3.5)
but the two curves are not overlapping. The reason
is that Eq. (3.2) provides an averaged value over a where 0CH and 00CH are the respective strains after
range rather than as a function representing the correction 1 and correction 2. Correction 2 ensures
instantaneous values in terms of difference in that the combined curve is smooth, i.e. maintains its
compliance. The adjustment for the instantaneous continuity in terms of the first derivative at the last
valid point of extensometer measurement.
25 The strain after two-step correction is labeled as
σ1
‘‘correction 2’’ in Fig. 9. As seen, the correction 2
Engineering Stress (MPa)

20 overlaps the last portion of the curve from the

extensometer and extends along the same slope.
15 Combining the strain measured from the extens-
ometer with the strain from the crosshead after two-
10 step correction gives a complete stress–strain curve,
labeled as ‘‘combined’’ in Fig. 10.
5 Strain EX Fig. 11 plots the strain rate–strain plot corre-
Strain CH sponding to the combined curve in Fig. 10. As
0 shown, the two-step correction scheme eliminates
0 0.1 0.2 0.3
ε1EX ε1CH the discontinuity seen in Fig. 6.
Engineering Strain The accuracy of the proposed correction method
was examined by comparing the engineering stress–
Fig. 8. Two-step correction scheme: step 1—determining the
strains measured by extensometer (EX) and from crosshead strain curves of thermoplastic materials measured at
extension (CH) for the calculation of compliance correction two different labs generated in a round robin test
factor a. [18–20]. Both labs used optical extensometers with a
 ARTICLE IN PRESS
X. Xiao / Polymer Testing 27 (2008) 284–295 289

30 gage length of 6 mm. The measurement at the

strain from EX University Dayton Research Institute (UDRI) was
Engineering Stress (MPa)

25 strain from CH made with a 20 mm lens which gave a measuring

combined range up to 120% strain. The optical extensometer
20
at General Motors (GM) lab had a 10 mm lens and a
15 measuring range of 60% strain with a gage length of
6 mm. Previously, a simple shifting method was used
10 and the obtained fracture strain values were lower as
compared to the direct measurements from UDRI
5
[19]. The data generated in the previous study were
0
treated using the current two-step correction. As an
0 0.5 1 1.5 2 2.5 example, Fig. 12 presents the tensile stress–strain
Engineering Strain curves of PC/ABS tested at a nominal strain rate of
40 s1. As shown, the stress–strain curve obtained
Fig. 10. Engineering stress–strain curves obtained by strain
measurement from the extensometer (EX), crosshead extension using the two-step correction is almost identical to
(CH) and by combining the strain from EX and the corrected CH the direct measurement from UDRI.
strain.
3.3. Summary of tensile test results

0.20 The nominal strain rate of the test was estimated by

v
Strain Rate (1/s)

0.15 _ ¼ , (3.6)
l0
0.10 where v is the loading rate of the testing frame and l0
is the nominal gage length of the testing specimen. In
0.05 dynamic testing, the actual strain rate in the specimen
is not constant because the specimen must experience
0.00 acceleration when being loaded from rest to the
0 0.5 1 1.5
Strain
targeted rate. To standardize the definition of strain
rate in high rate tensile tests of polymeric material,
Fig. 11. The strain rate versus strain plot corresponding to the the SAE draft Standard J2749 [21] recommends the
combined curve in Fig. 10. use of the average strain rate within the range of 10–
20% plastic strain. This definition was used in the
current study.
Fig. 13 plots the typical engineering stress–strain
80 curves at five rates. Table 1 provides a summary of
70
Engineering Stress (MPa)

60 40
0.017
Engineering Stress (MPa)

50 0.17
40 30 3
30
30 120
strain from EX 20
20
strain from CH
10 2-step correction
UDRI data 10
0
0 0.2 0.4 0.6 0.8 1
Engineering Strain 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Fig. 12. Tensile stress–strain curves of PC/ABS, tested at 40 s1. Engineering Strain
The curves were obtained by strain measurement from the
extensometer (EX), crosshead extension (CH), by combining the Fig. 13. Engineering stress–strain curves of HDPE. The nominal
strain from EX and the corrected CH strain (two-step correc- strain rates are 0.01, 0.1, 1, 10, 100 s1. The plastic strain rates are
tion), and by direct measurement (UDRI). 0.017, 0.17, 3, 30 and 120 1s, determined according to Ref. [20].
 ARTICLE IN PRESS
290 X. Xiao / Polymer Testing 27 (2008) 284–295

Table 1
Tensile properties of the HDPE material

Nominal strain rate Strain rate (10–20% ep) Peak stress (MPa) Modulus (MPa) Engineering failure strain

0.01 0.017 21.970.5 59370.3 2.8371.29

0.1 0.17 24.970.5 711.8715.3 1.6970.31
1 3 29.171.4 876.2715.9 1.1870.17
10 30 32.870.8 981.5714.6 0.8770.08
100 120 35.470.2 1002.8749.9 0.7970.17

the mechanical properties including the peak tion, the Poisson’s ratio is derived to be 0.5. For a
engineering stress, the Young’s modulus and the metallic tensile specimen, as long as the deformation
engineering failure strain measured at different is homogeneous along the gage length, the constant
rates. As mentioned in Section 1, the strain mea- volume condition yields [13]:
sured by extensometer is an average strain over the
gage length and, therefore, the strain measured after Ai l i ¼ A0 l 0 (4.4)
necking becomes gage length dependent. The failure and the following relationships hold between
strain values listed in Table 1 correspond to a gage the engineering and the true stress and strain
length of 11–12 mm. values [13]:

4. True fracture strain sT ¼ sð1 þ Þ, (4.5)

In large-scale dynamic FE simulations, plastic T ¼ lnð1 þ Þ. (4.6)

materials are usually represented by material models
based on plasticity theory. The constitutive equations The above relations are invalid when necking
in the theory of plasticity are based on true stress– occurs. In addition, large deformations of polymeric
strain curves and, therefore, the input stress–strain materials involve volumetric changes. Unlike me-
curves required by plasticity-based material models are tals, the Poisson’s ratio is not 0.5 for polymeric
true stress versus true (plastic) strain. As mentioned in materials in large deformation. It usually decreases
Section 1, necking nullifies the simple relationship from an initial value of 0.4 to about 0.2–0.3 with
between the elongation and the cross-sectional area of increase in deformation [22].
tensile test specimens and the determination of true Because of the differences in the deformation
stress and true strain becomes an issue. mechanisms, the constitutive equations based on
The engineering strain is defined by Eq. (3.1) and metal plasticity are not quite suitable for polymeric
the engineering stress s [13] is defined as materials. This problem has been recognized [22]
and studied [22–24]. Before suitable material models
P
s¼ , (4.1) become available in commercial codes, one needs to
A0 use the existing material models to represent
where P is the load, A0 is the original cross-sectional polymeric materials. The piecewise linear elastic–
area of the specimen. plastic model is frequently employed [25].
The true stress and true strain are defined as [13] For polymeric materials, since the material
P models based on classic plasticity relations do not
sT ¼ , (4.2) describe their large deformation behavior, but are
Ai
used as a temporary solution, the exact shape of the
Z li true stress–true strain curve is not critical, and a
dl li
T ¼ ¼ ln , (4.3) pseudo-curve might be used [26]. The pseudo-curve
l0 l l0
and the fracture strain may be obtained by
where sT is the true stress, eT is the true strain, l0 is correlating simulations with coupon and/or compo-
the original length, Ai and li are the instantaneous nent tests.
cross-sectional area and length, respectively. The true tensile fracture strain may be esti-
The volume change in metal plastic deformation mated by direct measurement on ruptured speci-
is negligible. Under the constant volume assump- mens. This work investigated the following
 ARTICLE IN PRESS
X. Xiao / Polymer Testing 27 (2008) 284–295 291

Table 2
True fracture strain estimated using various methods

Strain rate (s1) ln(A0/A) 11–12 mm gage Necked zone

0.017 1.6470.12 1.3070.32 1.5170.25

0.17 1.7070.01 0.9970.12 1.4470.04 after test
3 1.6470.09 0.7870.08 1.4570.09
30 1.5070.05 0.6370.04 1.2670.14 after 24 hours
120 1.4470.24 0.5870.09 1.1470.02

three methods to estimate the true tensile fracture Fig. 14. Viscoelastic recovery observed from the shape change of
strain: a specimen after three-point bending test.

(1) direct measurement of area reduction on rup-

where lf is the value of elongation at fracture, and its
tured tensile specimens;
value was estimated from the engineering strain at
(2) average fracture strain over the gage length; and
fracture ef (from crosshead extension with correc-
(3) average fracture strain over the necked length.
tion) using a rearranged form of Eq. (3.1):
l f ¼ l 0 ð1 þ f Þ, (4.9)
4.1. Direct measurement of area reduction
lf can also be estimated by measuring the gage
The true fracture strain eTf can be computed length of ruptured specimens. It should be noted
directly according to its definition [13]: that the measurement on the ruptured specimen
A0 does not include the viscoelastic strain. Fig. 14
Tf ¼ ln , (4.7)
Af compares the shape of a specimen shortly after a
where A0 is the original cross-sectional area and Af three-point bending experiment and after 24 h. As
is the cross-sectional area of the ruptured specimen. shown, the HDPE material exhibited significant
The width and the thickness of ruptured specimens viscoelastic recovery. Consequently, the measure-
were measured using a digital caliper. Because the ment on the ruptured specimens would give a
area of final fracture had been torn into irregular smaller elongation, i.e. a relatively conservative
shapes, it was difficult to measure the true fracture value, as compared to the value calculated using
surface area. Therefore, the measurement was taken Eq. (4.9).
not at the point of final fracture but at an adjacent
area where the cross section of the specimen
4.3. Fracture strain over the necked zone
remained nearly rectangular. The measurements
corresponded to the width and the thickness of the
At quasi-static strain rates, the HDPE tensile
necked zone before fracture. The strain calculated
specimens exhibited uniform necking and sustained
based on this measurement should be treated as a
a large elongation as in a drawing process. Never-
conservative estimation of fracture strain. Table 2
theless, the total elongation decreased with increas-
provides the value of the true fracture strain obtained
ing strain rate, as shown in ruptured specimens in
using this method. It shows that the rate dependence
Fig. 15.
of true fracture strain measured by cross-sectional
The measurements needed to estimate the true
area reduction is relatively small. This is partially due
strain over the necked zone is illustrated in Fig. 15.
to the limitation in the measurement of the fracture
In Fig. 15, ln is the final length of the necked zone
surface area as discussed above.
and l is final gage length. Neglecting the small
deformation outside the necked zone, the original
4.2. Average fracture strain over gage length
length of the necked zone ln0 can be approximated
by subtracting the length of the un-necked zone
The true fracture strain eTf can also be calculated
from the original gage length:
following the definition of Eq. (4.3):
lf l n0 ¼ l 0  ðl  l n Þ, (4.10)
Tf ¼ ln , (4.8)
l0 where l0 is the original gage length.
 ARTICLE IN PRESS
292 X. Xiao / Polymer Testing 27 (2008) 284–295

lastic recovery. Method (2) was based on corrected

crosshead extension and was not affected by the
viscoelastic recovery. In addition, methods (1) and
(3) each had their limitations. As mentioned above,
due to the difficulty in dealing with the torn shape of
ruptured specimens, the final areas measured in
method (1) are actually the values for the cross-
sectional area of the necked zone before the final
fracture. Method (3) measured strain over the
necked zone. As shown in Table 3, the size of
necked zone varied with strain rate at the two lower
strain rates. These factors also contributed to the
deviation of the rate dependence of the fracture
strain of these two methods at the two lower rates.
Theoretically, the fracture strain measured by
Fig. 15. Determining the true fracture strain over the necked
length on specimens after the tensile test. The nominal strain rate area reduction corresponds to an infinitely small
from left to right: 0.01, 0.1, 1, 10, 100 s1. The figure also gage length. Fig. 16 shows that the values of the
illustrates the definition of the necked zone length. fracture strain by area reduction were only slightly
higher than that measured over the necked zone,
The true fracture strain over necked zone was suggesting that the fracture strain may be approxi-
estimated using Eq. (4.8) in the following form: mately considered as independent of gage length,
when the gage length is smaller than the necked
ln
Tf ¼ ln . (4.11) zone, for materials exhibiting uniform necking. This
l n0
assumption helped to formulate a relation to obtain
Table 3 provides the measurements for the length the value of true fracture strain at an arbitrary
of the necked zone and the true fracture strain over gage length.
the necked zone for each specimen. At the two lower
rates, the size of the necked zone was quite 5. True fracture strain for arbitrary gage length
scattered. The values varied from 13 to 4.5 mm.
The size of the necked zone decreased with The tensile test of the HDPE material exhibited
increasing strain rate and was almost constant at necking. Once necking occurs, the deformation is no
the three higher rates with an average value of longer homogeneous and, hence, the strain calcu-
4 mm. lated from extension over a gage length becomes
Fig. 16 plots the true fracture strain measured by gage length dependent. This translates directly into
all three methods as a function of the strain rate. It mesh size dependence in FE analysis.
can be seen that that, when plotted against In this study, it was observed that necking started
logarithmic strain rate, the true fracture strain at roughly 30–40% strain. At lower rates, necking
measured by extensometer over a gage length of appeared to start at a larger strain level. On average,
11–12 mm, method (2), displayed a nearly linear necking occurred at above 30% strain. The work of
relation. The true fracture strain measured by the Parsons et al. [4] also confirmed that necking in
other two methods, methods (1) and (3), followed a HDPE started after the peak load but before the
linear trend similar to that measured by method (1) plateau region, which is comparable to the necking
at the three higher strain rates. At the two lower starting region indicated in Fig. 1. Another im-
strain rates, however, the fracture strain appeared to portant fact is that the propagation of the necked
be insensitive to strain rate. zone is confined within 4 mm length at strain rates
The deviation of the fracture strain from the higher than 0.2 s1. These two conditions lead to the
trend line at the two lower rates measured by following assumption: the elongation Dl over an
methods (1) and (3) may be attributed to viscoelas- arbitrary gage length l consists of two contribu-
tic recovery. These two methods rely on measure- tions:Dl1—uniform stretching up to 30% strain over
ments of ruptured specimens. At lower rates, the the length outside of necked zone; Dl2—elongation
polymer had a greater tendency to viscoelastic within the necked zone (size is about 4 mm). The
deformation and, thus, a larger amount of viscoe- relationship between the gage length and necked
 ARTICLE IN PRESS
X. Xiao / Polymer Testing 27 (2008) 284–295 293

Table 3
Tensile test specimen gage length measurements

Strain rate Sample ID Engineering Gage length Final gage True fracture Necked zone True fracture
(s1) fracture (mm) length (mm) strain over (mm) strain over
strain gage length necked zone

0.02 1 2.97 12.00 47.60 1.38 8.40 1.66

2 2.08 12.00 37.00 1.13 10.10 1.25
3 1.67 12.00 32.00 0.98 6.00 1.47
4 4.58 12.00 67.00 1.72 13.00 1.65
Average 2.83 1.30 9.38 1.51
S.D. 1.29 0.32 2.94 0.19
0.2 1 1.73 12.00 32.80 1.01 4.50 1.73
2 1.98 12.00 35.80 1.09 7.90 1.39
3 1.37 12.00 28.40 0.86 7.10 1.20
Average 1.69 0.99 6.50 1.44
S.D. 0.31 0.12 1.78 0.27
3 1 1.41 11.00 26.50 0.88 4.30 1.53
2 1.35 11.50 27.00 0.85 3.50 1.69
3 1.16 10.90 23.50 0.77 3.60 1.50
R1 1.16 10.26 22.16 0.77 4.06 1.37
R2 0.99 10.91 21.71 0.69 4.51 1.22
R3 1.02 11.01 22.24 0.70 3.81 1.37

Average 1.18 0.78 3.96 1.45

S.D. 0.17 0.08 0.40 0.16

30 1 0.82 11.70 21.30 0.60 4.30 1.17

2 0.96 10.70 21.00 0.67 3.20 1.44
3 0.97 11.40 22.50 0.68 3.40 1.45
R1 0.78 10.26 18.26 0.58 3.06 1.29
R2 0.81 11.21 20.29 0.59 4.61 1.09
R3 0.90 11.38 21.62 0.64 5.08 1.10
Average 0.87 0.63 3.94 1.26
S.D. 0.08 0.04 0.84 0.16
120 1 1.09 11.50 24.00 0.74 3.30 1.57
2 0.75 11.40 19.90 0.56 5.40 0.95
3 0.59 11.90 18.90 0.46 4.00 1.01
R1 0.87 10.26 19.19 0.63 4.36 1.11
R2 0.69 10.78 18.22 0.52 3.78 1.09
R3 0.73 10.69 18.49 0.55 3.69 1.14
Average 0.79 0.58 4.09 1.14
S.D. 0.17 0.09 0.73 0.22

zone is shown in Fig. 15. The mathematical Based on Eqs. (5.1)–(5.3), the failure strain for an
expressions are arbitrary gage length l (l4ln) is derived as
Dl ¼ Dl 1 þ Dl 2 , (5.1) Dl ðl  l n Þ  0:3 þ nf  l n
f ¼ ¼ . (5.4)
l l
Dl 1 ¼ ðl  l n Þ  0:3, (5.2) Eq. (5.4) allows one to estimate the true failure
strain at an arbitrary gage length, i.e. for an
Dl 2 ¼ nf  l n , (5.3) arbitrary mesh size. As an example, Fig. 17 plots
the mesh size dependence of the true fracture strain
where ln is the size of the necked zone, nf is the estimated using Eq. (5.4) for a strain rate of 3 s1.
failure strain measured over the necked zone. Fig. 18 plots the true fracture strain as a function of
 ARTICLE IN PRESS
294 X. Xiao / Polymer Testing 27 (2008) 284–295

2.0 strain rates for mesh sizes of 7 and 10 mm,

calculated using the algorithm discussed above.
For comparison, the true fracture strain measured
1.5
over a known gage length of 11–12 mm is overlaid
True Failure Strain

on the same plot. Judging against these data points,

1.0 Eq. (5.4) provides a conservative estimation.
It should be noted that Eq. (5.4) is valid for l4ln.
However, nf can be used as a conservative estima-
0.5 necked zone tion when lpln.
ln(A0/A)
11 mm gage
0.0
0.01 0.1 1 10 100 1000 6. Summary and conclusions
Strain Rate (1/s)
The tensile properties of an HDPE material
Fig. 16. The true fracture strain measured (1) over the necked
were measured at five nominal strain rates between
zone, (2) by area reduction, and (3) over an 11–12 mm gage
length. 0.01 and 100 s1. The elongation reduced with
increasing strain rate. However, even at the highest
rates tested, the values of the fracture strain were
still greater than the measuring range of the
1
extensometer. To obtain the engineering strain
beyond the range of extensometer, crosshead exten-
0.8 sion had to be used. An analysis based on a two-step
Failure Strain

correction scheme was proposed. This method

0.6
removed the discontinuity in strain rate-history
0.4 plots by the conventional correction method based
on compliance calibration. The stress–strain
0.2
curve obtained using the two-step correction was
0 almost identical to that obtained from the direct
0 10 20 30 40 measurement.
Mesh Size (mm) Necking was observed in all tensile tests at above
30% strain. Necking makes it difficult to obtain the
Fig. 17. Strain limit as a function of the mesh size for strain rate
of 3 s1 determined by Eq. (5.4). true stress–strain values. In this study, the true
fracture strains were estimated using the area
reduction, elongation over the necked zone and
elongation over the gage length. Each method has
2.0 its merits and limitations. With strain localization,
strain measured by an extensometer becomes gage
7mm calculated
length dependent. Fracture strains obtained by the
10mm calculated
1.5 two other methods, on the other hand, were based
True Failure Strain

11mm measured
on measurements with ruptured specimens that did
not include the viscoelastic strain. This affected the
1.0 strain rate dependence of fracture strain obtained by
these two methods, particularly at lower rates.
0.5
Nevertheless, these measurements provided addi-
tional information needed in the formulation of an
algorithm that relates the fracture strain measured
0.0 over the necked zone to the average value over an
0.01 0.1 1 10 100 1000 arbitrary gage length larger than the necked zone
Strain Rate (1/s) for materials exhibiting uniform necking. The
Fig. 18. Rate dependence of true fracture strain as a function of
proposed algorithm provides a method to estimate
strain rate and mesh size, determined by the algorithm discussed the mesh dependence of the fracture strain for FE
in Section 5. analysis.
 ARTICLE IN PRESS
X. Xiao / Polymer Testing 27 (2008) 284–295 295

Acknowledgments [11] Q.Z. Fang, T.J. Wang, H.M. Li, Large tensile deformation
behavior of PC/ABS/alloy, Polymer 47 (2006) 5174–5181.
The author would like to thank Zohir Molhem, [12] Standard Test Methods for Tensile Properties of Plastics,
ASTM Designation: D7638-03, ASTM International, Phila-
Hamid Kia of General Motors Corporation for delphia, USA, 2002.
helpful discussions and valuable suggestions; Andi [13] G.E. Dieter, Mechanical Metallurgy, third ed., McGraw-
Poli and Kurt Miller of Axel Products for tensile Hill, New York, 1986.
testing and photos; Susan Hill of UDRI for sharing [14] I.M. Ward, Mechanical Properties of Solid Polymers, Wiley/
high rate testing data. InterScience, London, UK, 1971.
[15] J.D. Ferry, Viscoelastic Properties of Polymers, Wiley, New
York, 1980.
[16] C. Reder, D. Loidl, S. Puchegger, G. Gitschthaler, H.
References Peterlik, K. Kromp, G. Khatibi, A. Betzwar-Kotas, P.
Zimprich, B. Weiss, Non-contact strain measurement of
[1] R.M. Caddell, T. Bates Jr., G. Yeh, On the tensile behavior ceramic and carbon single fibers by using the laser-speckle
of high density polyethylene subjected to cold rolling, Mater. method, composites part A, Appl. Sci. Manuf. 34 (2003)
Sci. Eng. 9 (1972) 223–229. 1029–1033.
[2] J. Sweeney, P. Caton-Rose, P.D. Coates, The modeling of [17] S. Hill, P. Sjöblom, Practical considerations in determining
large deformation of pre-oriented polyethylene, Polmyer 43 high strain rate material properties, SAE 981136.
(2002) 899–907. [18] X. Xiao, Dynamic tensile testing of plastic materials, Polym.
[3] J. Sweeney, P. Caton-Rose, P.D. Coates, A.P. Unwin, R.A. Test., 2007, in press, doi:10.1016/j.polymertesting.2007.
Duckett, I.M. Ward, Application of a large deformation 09.010.
model to unstable tensile stretching of polyethylene, Int. [19] X. Xiao, M. Leach, Intermediate and high strain rate tensile
J. Plast. 18 (2002) 399–414. testing of plastic materials, in: Proceedings of the 2005
[4] E.M. Parsons, M.C. Boyce, D.M. Parks, M. Weinberg, Society of Experimental Mechanics (SEM) Annual Con-
Three-dimensional large-strain tensile deformation of neat ference and Exposition on Experimental and Applied
and calcium carbonate-filled high-density polyethylene, Mechanics, Portland, USA, June 2005.
Polymer 46 (2005) 2257–2265. [20] S. Hill, M.F. Pinnell, A.J. Minch, Standardization of high
[5] A. Dasari, R.D.K. Misra, On the strain rate sensitivity of strain rate tensile testing of polymers, in: Proceedings of
high density polyethylene and polypropylenes, Mater. Sci. ANTEC 2005.
Eng., A 358 (2003) 356–371. [21] High Strain Rate Testing of Polymers, SAE J2749, a draft,
[6] C. G’Sell, J.J. Jonas, Determination of the plastic behavior 2006-01
of solid polymers at constant true strain rate, J. Mater. Sci. [22] G. Dean, B. Read, Modeling the behavior of plastics for
14 (1979) 583–591. design under impact, Polym. Test. 20 (2001) 677–683.
[7] C. G’Sell, J.M. Hiver, A. Dahoun, A. Souahi, Video- [23] H. Mae, Simulation of fracture behavior of elastomer
controlled tensile testing of polymers and metals beyond the modified polypropylene based on elastoviscoplastic consti-
necking point, J. Mater. Sci. 27 (1992) 5031–5039. tutive equation with craze and tensile softening law,
[8] C. G’Sell, J.M. Hiver, A. Dahoun, Experimental character- IMECE2005-79161, in: Proceedings of ASME World Con-
ization of deformation damage in solid polymer under gress, Orlando, USA, November 2005.
tension, and its interrelation with necking, Int. J. Solids [24] S. Kolling, A. Haufe, M. Feucht, P.A. Du Bois, A
Struct. 39 (2002) 3857–3872. constitutive formulation for polymers subjected to high
[9] G. Gérard Buisson, K. Ravi-Chandar, On the constitutive strain rates, in: Proceedings of 9th International LS-DYNA
behavior of polycarbonate under large deformation, Poly- Users Conference, vol. 15, Dearborn, USA, June 2006,
mer 31 (1990) 2071–2076. pp. 55–73.
[10] E.M. Parsons, M.C. Boyce, D.M. Parks, An experimental [25] X. Xiao, Plastic material modeling for FMVSS201 simula-
investigation of the large-strain tensile behavior of neat tion, SAE Technical Paper, SAE-2002-01-0385, 2002.
and rubber-toughened polycarbonate, Polymer 45 (2004) [26] X. Xiao, Plastic material modeling for vehicle crash
2665–2684. simulation using LS-DYNA, ANTEC 2001.