Composites Part B 97 (2016) 68e83

Contents lists available at ScienceDirect

Composites Part B
journal homepage: www.elsevier.com/locate/compositesb

Creep behavior and modeling of neat, talc-filled, and short glass fiber
reinforced thermoplastics
Mohammadreza Eftekhari, Ali Fatemi*
Mechanical, Industrial and Manufacturing Engineering Department, The University of Toledo, 2801 West Bancroft Street, Toledo, OH 43606, USA

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

Article history: Creep behavior of neat, talc-filled, and short glass fiber reinforced injection molded thermoplastic
Received 29 January 2016 composites were investigated and modeled at room and elevated temperatures. Creep strength
Received in revised form decreased and both creep strain and creep rate increased with increasing temperature. The temperature
6 April 2016
effect was more significant for samples with glass fibers in the transverse to the load direction, as
Accepted 17 April 2016
compared with the longitudinal direction. The LarsoneMiller parameter was able to correlate the creep
Available online 1 May 2016
rupture data of all materials. The MonkmaneGrant relation and its modification were successfully used
to correlate minimum creep rate, time to rupture, and strain at rupture data. The Findley power law and
Keywords:
A. Discontinuous reinforcement
timeestress superposition principle (TSS) were used to represent nonlinear viscoelastic creep curves.
Short glass fiber Long-term creep behavior was also satisfactory predicted based on short-term test data using the TSS
A. Thermoplastic resin principle.
B. Creep © 2016 Elsevier Ltd. All rights reserved.
C. Analytical modelling

1. Introduction When a constant load is applied on the specimen, an initial

The usage of polymeric materials is increasingly rapidly due to
their lightweight, low cost, and capability to be manufactured in
complex geometries at high production rates. Addition of fillers
such as talc or reinforcement such as short glass fibers can improve
the mechanical performance of unreinforced polymers to a high
degree. These materials are typically subjected to loading and
environmental conditions causing creep deformation. This time-
dependent behavior results in molecular rearrangement, the de-
gree of which depends on factors including the type of material,
magnitude of the stress, temperature, and time.
For a linear viscoelastic material under creep condition, the
applied stress is proportional to the strain response at a given time,
independent of the stress. For viscoelastic creep, the behavior ex-
hibits nonlinearity at large strains. While linear viscoelastic
behavior is represented using physical based constitutive equa-
tions, such as Maxwell model, constitutive equations for nonlinear
creep or other viscoelastic behaviors are much more complex. As a
results, they require a large number of functions with higher order
stress terms with a large number of material constants and are
typically empirical in nature [1].
* Corresponding author. Tel./fax: þ1 419 530 8213.
E-mail addresses: mohammadreza.eftekhari@rockets.utoledo.edu (M. Eftekhari),
afatemi@eng.utoledo.edu (A. Fatemi).
elongation occurs which is known as instantaneous strain, ε0. This
is followed by a rapidly decreasing deformation rate known as the
primary stage, followed by a steady-state linear deformation stage,
known as the secondary creep stage, for which the experimental
data is typically of most interest. The final stage involves rapid
deformation at an accelerated rate known as tertiary deformation,
ultimately leading to the specimen rupture. Some materials do not
have the secondary stage, while tertiary creep only occurs at high
stresses and for ductile materials [2]. Instantaneous creep strain, ε0,
can consist of elastic (εe) and plastic (εp) strain depending on the
level of the applied load or stress and can constitute a considerable
fraction of the total strain.
While creep in metallic materials occurs only at elevated tem-
peratures, creep of polymers can be significant at any temperature.
Therefore, creep or relaxation are major concerns in using polymer
and polymer composites in structural applications. Long-term
creep deformation and strength applicable to service condition
are normally obtained by extrapolation of short-term test data
obtained under accelerated testing conditions, such as higher
temperature, stress, and humidity by using prediction models.
A number of studies in the literature have investigated creep
behavior of neat polymers including damage mechanism [3],
nonlinear creep behavior and modeling [4], superposition methods
for long-term creep predictions considering environmental effects
[5e7], and creep rupture behavior and predictions [8,9]. Seltzer
et al. [10] studied the effect of organoclay on creep behavior of

http://dx.doi.org/10.1016/j.compositesb.2016.04.043
1359-8368/© 2016 Elsevier Ltd. All rights reserved.
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
Nomenclature
A, B, C, D, G, H, a, b material constants
CLMP LarsoneMiller parameter constant
E elastic modulus
LMP LarsoneMiller parameter
m, n Findley power law constants
S stress
Su ultimate tensile strength
t time
tR time to creep rupture
T temperature
Tg glass transition temperature
Tm melting temperature
εc creep strain
ε0 instantaneous strain
ε_ min minimum creep rate
4S shift factor
l, h MonkmaneGrant relation constants
l0 h0 modified MonkmaneGrant relation constants
injection molded polyamide-6 using dynamic mechanical analysis
(DMA) in cantilever-bending. Eftekhari and Fatemi [11] reviewed
the studies related to creep behavior and modeling of short fiber
reinforced polymer composites (SFRPCs) at elevated temperatures.
Time-temperature-stress superposition (TTSS) procedure has been
used for extending short-term creep data for long-term creep
predictions. Empirical power law models have been widely used to
represent nonlinear creep strain curves for a wide range of tem-
peratures and stress levels. Prevailing creep damage is crazing
perpendicular to the stress direction where crazes start mainly at
the fiber/matrix interface independent of the interface quality.
This study investigated creep behavior of several injection
molded neat, talc-filled, and short glass fiber reinforced thermo-
plastic composites. Creep tests were conducted for different con-
ditions including stress level, temperature, and fiber orientation.
Empirical fits were used to represent creep rupture data and relate
minimum creep rate data to the applied stress. LarsoneMiller
parameter was used to correlate creep rupture data and power law
models were used to represent experimental creep strain curves.
Semi-empirical models were used to correlate minimum creep rate,
time to rupture, and strain at rupture together. Creep strain master
curves and corresponding shift factor versus stress curves were
developed using short-term creep data according to the timee-
stress superposition (TSS) method. A small number of long-term
creep tests were also conducted to verify the capability of the
developed models for long-term creep predictions based on short-
term creep data. The experimental program conducted, the results
obtained, and the models applied are presented and discussed in
the following sections.
2. Experimental program
2.1. Materials and specimen
Neat, talc-filled, and short glass fiber reinforced thermoplastics
with or without rubber as a modifier were considered for the
experimental study. The neat thermoplastic used is an impact
polypropylene copolymer, here designated as PP. The talc-filled
composites used are a polypropylene-elastomer blend (a thermo-
plastic polyolefin) with 25% by weight (25 wt%) elastomer and
69
35 wt% talc (referred to as PO) and a 40 wt% talc filled poly-
propylene (referred to as PP-T). The short fiber reinforced com-
posites are polypropylene containing 30 wt% short glass fibers
(referred to as PP-G), polyamide-6.6 reinforced with 30 wt% glass
fibers (referred to as PA66), 20 wt% short glass fiber reinforced
modified polyphenylene ether and polystyrene resin (here referred
to as PPE/PS), polybutylene terephthalate with 30 wt% glass fibers
(here referred to as PBT), and polyamide-6 containing 35 wt% glass
fibers and about 10 wt% rubber impact modifier (here referred to as
PA6). Common applications of these materials include those in the
automotive industry. A summary of the materials used in the
experimental study, including constituents, melting temperature
(Tm), and glass transition temperature (Tg) is presented in Table 1.
Mean glass fiber length and diameter for short glass fiber reinforced
composites were obtained as 0.25 mm and 10 microns, respectively.
More details on fiber length, diameter, and orientation distribution
of studied reinforced composites can be found in Ref. [12]. Tensile
and fatigue behaviors of the mentioned materials were studied
previously in Refs. [12e18].
Rectangular plates with thicknesses of 2.8 mm (for PP-T, PP-G,
PA66 and PPE/PS) and 3.8 mm (for PO, PP, PA6 and PBT) and di-
mensions of 100  200 mm were injection molded. Rectangular
strips were cut in 90 (transverse) and 0 (longitudinal) directions
with respect to the mold flow direction with a width of about
25 mm according to Fig. 1(a) and (b). A portion of plaques from
injection molding side with the length of 60 mm was discarded to
prevent end effects. The specimens were machined from the strips
using a CNC milling machine with the geometry shown in Fig. 1(c).
All the specimens were dried before testing to preserve the dry
as molded condition and exclude moisture effect. PA66 and PPE/PS
specimens were dried for 24 h at 80  C and 4 h at 100  C, respec-
tively, in a vacuum chamber. PO, PP, PP-T and PP-G specimens were
dried for 4 h at 80  C, and PBT and PA6 specimens were dried for 6 h
at 120  C and 80  C, respectively, in a regular chamber. All the
specimens were kept in a desiccator prior to testing.
2.2. Experimental method
Creep tests were conducted using a closed loop servo-hydraulic
testing machine which was controlled with a digital controller. A
video extensometer was used to measure creep strain. A pair of
pneumatic grips with 5 kN capacity and adjustable pressure system
which were suitable for testing polymers at low and elevated
temperatures was used. A strain-gaged specimen was used to align
the load train components prior to testing. An environmental
chamber employing an electronic heating element with accuracy of
better than ±1  C was used for performing tests at elevated tem-
peratures. A lever arm test machine with dead weights was used for
long-term creep tests at room temperature. Some tests which were
conducted using the servo-hydraulic test machine were repeated
with the lever arm test machine to verify the repeatability of the
results.
Tensile creep tests were performed according to ASTM D2990
[2] test standard. Creep tests were conducted in the transverse
direction for all the materials and also in the longitudinal direction
for PBT and PA6 for a wide range of stress levels at several tem-
peratures, as reported in Table 1. Short-term tests lasted on the
order of about one hour to three days. A limited number of longer
time creep tests (up to about 40 days) were also conducted for PP-G.
Two repeat tests were conducted for most of the conditions.
3. Creep rupture data and modeling
Stress versus time to rupture data for all the materials at
different temperatures are plotted in Fig. 2. Tests that did not fail
70 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83

Table 1
Summary of the materials with creep test conditions used for the experimental study.

Material Polymer matrix Reinforcement Tm Tg Directions Temperatures Stress range

designation ( C) ( C) studied studied ( C) studied (% Su)

PO Polypropylene þ 25 wt% Rubber 30 wt% Talc >120 <-10 T 23, 85, 125 30e82
PP Impact Polypropylene None 170 4 T 23, 85, 125 40e85
PP-T Polypropylene 40 wt% Talc 165 11 T 23, 85 53e80
PP-G Polypropylene 30 wt% Short Glass Fiber 165 23 T 23, 85, 120 46e85
PA66 Polyamid-6,6 30 wt% Short Glass Fiber 260 55 T 23, 85, 120 75e90
PPE/PS Polyphenylene ether þ Polystyrene 20 wt% Short Glass Fiber 325 135 T 23, 85 60e85
PBT Polybutylene terephthalate 30 wt% Short Glass Fiber 255 50 T, L 85, 125 72e91
PA6 Polyamide-6 þ 10% Rubber 35 wt% Short Glass Fiber 220 40 T, L 85, 125 81e91

L: Longitudinal (fibers along load direction).

T: Transverse (fibers perpendicular to load direction).

Fig. 1. Direction and location of specimens cut from injection molded plaques in (a) the transverse and, (b) the longitudinal direction according to the mold flow direction (arrows).
(c) Specimen geometry used for creep tests (dimensions are in mm).

(run-outs) and are not included in the fits, however, the data points
are shown in the plots. A significant effect of temperature is
observed for all the materials, as expected. For both PBT and PA6,
creep resistant is significantly higher in the longitudinal direction,
as compared to the transverse direction. Also, the effect of tem-
perature is more accentuated in the transverse direction, where
creep resistance is more matrix dependent, as compared with the
longitudinal direction. Also.
To be able to compare the creep rupture behavior of the mate-
rials together, stress versus time to rupture data in Fig. 3 are
normalized with the corresponding ultimate tensile strength, Su,
obtained from tension tests conducted at a displacement rate of
1 mm/min. The data for PO, PP, PP-T, and PP-G are shown in
Fig. 3(a), while the data for PA66, PPE/PS, PBT, and PA6 are shown in
Fig. 3(b). The second group of materials are more creep resistant
than the first group, and the Tg of the second group is also higher
than the first group. PO, PP, PP-T, and PP-G data at 23  C are
correlated very well by normalizing with tensile strength.
Normalizing with tensile strength also correlates all the creep
rupture data of the second group very well, except for PPE/PS data
at 85  C (see Fig. 3(b)).
Tg and Tm of PO are lower than those of PP, PP-T, and PP-G,
therefore, more temperature sensitivity in creep rupture behavior
of PO is expected. The mechanism of creep damage of poly-
propylene, which is mainly due to the crazes perpendicular to the
stress direction, remains unchanged in the presence of reinforce-
ment. PPE/PS has an amorphous matrix, while PA66, PBT, and PA6
have a semi-crystalline matrix. Also, percent weight of short glass
fiber reinforcement for PPE/PS is lower, as compared to the other
SFRPCs, therefore, creep rupture behavior of PPE/PS is more tem-
perature dependent. A craze type failure was observed for PPE/PS
samples, while no evidence of crazing even at elevated tempera-
tures was observed for PA66, PBT, and PA6.
A power fit was utilized to represent the creep rupture experi-
mental data in the form of:
S ¼ AðtR ÞB (1)
where S is the applied stress level, tR is the time to rupture in hours,
A is the intercept at tR ¼ 1, and B is the slope of the fitted line. A and
B are temperature and orientation dependent constants and their
values for all the materials at different temperatures are reported in
Table 2. Correlations of A and B values with Su are shown in Fig. 4
and expressed by the following equations:
A ¼ 0:85 Su with R2 ¼ 0:99 (2)
B ¼ 0:44ðSu Þ0:66 with R2 ¼ 0:8 (3)
These correlations can be used to estimate creep rupture
properties of thermoplastics and thermoplastic composites at
different temperatures and in different mold flow directions.
Different correlative parameters based on the Arrhenius rate
equation [19] which relate temperature and time to rupture
together have been developed over the years. LarsoneMiller [20],
MansoneHaferd [21], SherbyeDorn [22], MansoneSuccop [23],
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83 71

Fig. 2. Effect of temperature on creep rupture of (a) PO, (b) PP, (c) PP-T, (d) PP-G, (e) PA66, and (f) PPE/PS in the transverse direction and, (g) PBT and, (h) PA6 in both transverse and
longitudinal directions. Lines are fits to experimental data. Arrows show tests that did not fail and are not included in the fits.

and MansoneMuralidharan [24] are some of these parameters.
LarsoneMiller Parameter (LMP) has become the most widely used
because of its simplicity. Comparison of correlations with different
parameters did not show considerable differences based on the
experimental data in this study. This parameter is expressed in the
following form:
ðT þ 273Þðlog tR þ CLMP Þ
LMP ¼
1000
where T is the temperature in  C, tR is time to rupture in hours, and
CLMP is a constant. CLMP value was determined by fitting a line to log
tR versus 1/T data for a given material and stress level. Intersection
of the lines for different stress levels at 1/T ¼ 0 defines the value of
constant CLMP.
LarsoneMiller master curves for all the materials are shown
in Fig. 5. As can be seen, the correlations are very good. The
power law fit is used to represent master curves in the following
(4) form:
72 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83

Fig. 3. Normalized creep rupture data with tensile strength for (a) PO, PP, PP-T, and PP-G and, (b) PA66, PPE/PS, PBT, and PA6 at different temperatures.

Table 2
Material, temperature and fiber orientation dependent tensile properties and modeling constants.

Material Direction Temperature ( C) Su (MPa) E (MPa) A B C D n a b

PO T 23 14.7 2123 11.7 0.076 4E-15 14.3 0.500 2E-6 6.33

T 85 4.1 670 2.7 0.133 4E-5 12.5 0.450 2E-2 6.29
T 125 2.1 350 1.0 0.219 2E1 8.6 0.410 2E1 6.89
PP T 23 22.3 1630 21.8 0.070 1E-15 12.2 0.400 1E-5 4.50
T 85 9.5 566 6.8 0.125 7E-9 10.6 0.420 9E-4 4.69
T 125 5 282 3.1 0.205 3E-5 9.8 0.410 1E-1 3.20
PP-T T 23 29.4 4200 23.3 0.052 4E-23 16.6 0.438 1E-9 6.73
T 85 13.7 1550 9.23 0.051 1E-19 19.9 0.318 1E-5 5.56
PP-G T 23 46.8 3446 37.3 0.062 1E-20 12.9 0.375 5E-8 5.00
T 85 21.7 1516 15.8 0.094 2E-13 11.4 0.600 8E-9 7.55
T 120 11.9 966 9.1 0.116 2E-12 13.3 0.762 1E-8 9.51
PA66 T 23 104.7 5250 91.9 0.026 3E-31 15.4 0.237 1E-11 5.71
T 85 58.8 2250 51.0 0.025 5E-52 30.1 0.182 2E-11 6.62
T 120 46.7 1550 40.0 0.028 8E-34 20.7 0.206 6E-11 6.82
PPE/PS T 23 70.8 4400 58.2 0.027 1E-32 17.9 0.288 2E-15 8.39
T 85 39.5 4000 35.3 0.052 5E-20 12.4 0.343 4E-9 5.54
PBT T 85 35.5 2300 31.2 0.023 5E-63 41.6 e
T 125 26.5 1830 22.0 0.036 2E-35 25.6
L 85 64 4600 55.5 0.017 7E-53 30.0
L 125 48.5 3380 41.8 0.023 1E-59 36.3
PA6 T 85 46.5 1800 38.7 0.049 3E-62 38.4 e
T 125 37 1300 32.3 0.022 3E-53 34.6
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83 73

Fig. 4. Correlations of (a) A and, (b) B values with Su for all the materials at different temperatures and fiber orientations. The conditions associated with the symbols are the same as
those in Fig. 3.

Fig. 5. LMP master curves for (a) PO, PP, PP-T, and PP-G in the transverse direction and, (b) PA66, PPE/PS, PA6 in the transverse direction and for PBT in both transverse and
longitudinal directions.
74 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83

Table 3 and PP-G the third stage of creep is an important portion of the
Material and fiber orientation dependent modeling constants. creep curve (see Fig. 6(a)), while for the other materials it is a small
Material Direction CLMP G H l h l0 h0 portion of the creep curve (see Fig. 6(b)). Creep strain of PA6 is
PO T 40 4.92E9 8.03 0.60 4.0 0.90 0.53
considerably higher than PA66, PPE/PS, and PBT (see Fig. 6(b)), due
PP T 31 2.66E7 6.36 0.92 4.5 1.12 0.46 to the effect of elastomer additive (10%wt).
PP-T T 42 4.90E6 4.86 1.00 2.1 0.68 0.27 An increase in temperature results in higher macromolecular
PP-G T 26 1.75E6 5.24 0.83 5.3 0.91 0.50 mobility, therefore, higher creep deformation for all the materials,
PA66 T 55 3.41E5 2.95 1.35 1.7 e e
as expected. An example of this is shown in Fig. 7(a) for PP-G at a
PPE/PS T 32 4.07E4 2.90 1.68 0.4 e e
PBT T 50 4.24E5 3.30 1.00 0.6 e e stress level of 75% of Su. Creep curves of PP-T and PP-G are similar
L 60 2.34E5 2.72 // // at both 23  C and 85  C, as shown in Fig. 7(b). This indicates that
PA6 T 22 1.6E3 1.80 0.82 1.2 e e the type of reinforcement (talc or short glass fiber) has the same
effect on creep deformation of polypropylene, similar to that
observed for creep rupture behavior. Reinforcement or filler does
S ¼ GðLMPÞH (5)
not change the creep mechanism and only results in improved
where S is stress in MPa, and G and H are material constants which creep strength (i.e. reduction of both creep strain and rate), as
are listed in Table 3. The LarsoneMiller creep parameter can be used compared to the neat polymer [26]. Creep strain is higher for the
for long-term creep-rupture predictions at different temperatures, samples tested in the transverse to the mold flow direction, as
provided that stress should be in the range of test stresses [25]. compared to the longitudinal direction, as shown in Fig. 7(c) for
PBT. In addition, as can be observed from Fig. 7(c), the effect of
temperature on creep behavior is more significant in the trans-
4. Creep strain-time behavior and predictions verse direction. This is because of the fact that creep behavior is
more matrix dependent for the samples with transversely ori-
Creep strain versus time curves at 85  C in the transverse di- ented short glass fibers.
rection at several stress levels are shown in Fig. 6. For PO, PP, PP-T

Fig. 6. Creep strain versus time curves at 85  C for (a) PO, PP, PP-T, and PP-G and for, (b) PA66, PPE/PS, PBT, and PA6.
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
Several empirical power law models have been proposed to
describe the nonlinear creep deformation behavior of polymeric
materials. Some of these models were reviewed in Refs. [27] and
[28]. The power law models have evolved over the years in order to
be applicable to a wider range of temperatures, stress levels, and
polymeric materials [29e31]. The Findley power law model [32]
has become a widely used model for describing creep deforma-
tion of polymers and their composites and is given by:
εc ¼ ε0 þ mt n
where εc is creep strain, n is a temperature dependent material
constant, and m is both stress and temperature dependent con-
stant. m is the intercept and n is the slope of the fitted line in the
logelog plot of (εc- ε0) versus time. n values are reported in Table 2
for different temperatures with t and εc in hour and %, respectively.
n is nearly constant for PP, independent of temperature, but for talc-
filled and short glass fiber reinforced materials, larger variation
with temperature is observed. Temperature independency of n for
neat polymers [33,34] and temperature dependency for filled and
reinforced polymers [35] have also been reported in the literature.
Values of ε0 can be estimated using RembergeOsgood equation,
which was used in Refs. [13,36,37] to model the tensile stress-strain
behavior of the studied materials. In order to find the variation of
constant m with stress, the following equation, which was sug-
gested by Hadid et al. [28], was used:
m ¼ a Sb
where a and b are material constants and their values are reported
in Table 2.
Fig. 7. (a) Effect of temperature on creep strain curves of PP-G under stress level of 75% of Su. (b) Comparison of PP-T and PP-G creep strain curves at 23  C and 85  C under stress
level of 62% of Su. (c) Fiber orientation effect on creep strain of PBT at 85  C and 125  C under stress level of 80% of Su.
75
Predictive creep curves using Equations (6) and (7) for PP-G at
different temperatures are presented in Fig. 8. It can be seen that
the Findley model is reasonably accurate in representing the short-
term creep strain curves of PP-G. Similar predictions were observed
for other materials at different stress levels and temperatures.
5. Minimum creep rate data and correlations
Minimum creep rate, ε_ min , is the slope of the portion of creep vs.
(6)
time diagram corresponding to secondary creep and is an impor-
tant parameter widely used for creep design of many components.
Norton power law [38,39] is used to relate minimum creep rate to
stress as:
ε_ min ¼ CðSÞD (8)
where C and D are material, temperature, and orientation depen-
dent constants, with values reported in Table 2 with S and ε_ min in
MPa and %/hours, respectively. Fig. 9 shows ε_ min versus stress data
and the corresponding Norton power law fits at different temper-
atures for all the materials in logelog plots. As can be seen, ε_ min
increases with increasing temperature and stress level for all the
materials. Minimum creep rate versus normalized stress using Su
data are shown in Fig. 10. The data for PO, PP, PP-T and PP-G are
shown in Fig. 10(a), while the data for PA66, PPE/PS, PBT and PA6
are shown in Fig. 10(b). Similar to what was observed for creep
rupture data (Fig. 3), PO, PP, PP-T, and PP-G data at 23  C are
(7) correlated very well by normalizing with tensile strength. Also,
normalizing with tensile strength correlates all the minimum creep
rate data of the second group very well, except for PPE/PS data at
85  C (see Fig. 10(b)).
76 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
Fig. 8. Comparison of predicted creep strain curves using Findley power law model (dash lines) with experimental creep strain curves of PP-G at different stress levels in the
transverse direction at (a) 23  C, (b) 85  C, and (c) 120  C.
PA66, PPE/PS, PBT, and PA6 exhibit small primary and tertiary
creep stages and the secondary creep stage is the main part of the
creep strain curve, particularly for tests conducted at lower stress
levels. Therefore, for these conditions creep strain can be estimated
with a linear line with its slope and intercept to be ε_ min and ε0,
respectively, where ε0 can be estimated using RambergeOsgood
relation explained earlier.
The MonkmaneGrant relationship [40] has been widely applied
to metallic materials. It relates time to rupture, tR, to the minimum
or secondary creep rate, ε_ min , given by:
ε_ lmin tR ¼ h
where l and h are material constants with their values reported
in Table 3 for all the materials, with tR and ε_ min in hours and
%/hours, respectively. The correlations of PP-G data at 23  C, 85  C
and 120  C using this relationship is shown in Fig. 11(a).
As can be seen, the correlation are reasonable and similarly good
correlations were obtained for the other studied materials. l and h
were found to be fiber orientation independent for PBT. Correla-
tions of all data using a single fit by MonkmaneGrant relationship
is shown in Fig. 12(a), expressed as:
ε_ 0:86 with R2 ¼ 0:78
min tR ¼ 2:65
A modified version of MonkmaneGrant relation was suggested
by Dobes and Milicka [41] which also includes strain at rupture, εR,
in the correlations. This model is suitable for materials exhibiting
small secondary creep and large tertiary creep stage and is repre-
sented by the following form:
0
ε_ lmin ðtR =εR Þ ¼ h0 (11)
where the material constants l0 and h0 are reported in Table 3 with
tR and ε_ min in hours and %/hours, respectively. As mentioned earlier,
a significant portion of creep curves of PO, PP, PP-T, and PP-G is in
the third stage of creep and strain at rupture cannot be obtained
using minimum creep rate predictions (Equation (9)), therefore,
Equation (11) is valuable when strain at rupture is needed for these
materials.
As can be seen in Fig. 11(b), correlations of data using the
modified MonkmaneGrant relation is very good for PP-G and
(9) similar good correlations was also obtained for PO, PP, and PP-T. l0 is
suggested to be unity for most metallic materials [42]. For PO, PP,
and PP-G, l0 values are near to unity and is 0.68 for PP-T. A single fit
using Equation (11) for PO, PP, PP-T, and PP-G is shown in Fig. 12(b),
expressed by:
ε_ 0:89
min ðtR =εR Þ ¼ 0:41 with R2 ¼ 0:92 (12)
6. Long-term creep tests and predictions
6.1. Time-stress superposition principle for long-term creep
(10)
predictions
Creep tests usually take a long time and are tedious and
expensive to conduct, therefore, predictive methods for long-term
creep properties based on short-term data are of great interest.
Accelerated or short-term tests can be conducted at elevated
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83 77

Fig. 9. Minimum creep rate versus stress data with corresponding power law fits for (a) PO, (b) PP, (c) PP-T, (d) PP-G, (e) PA66, (f) PPE/PS, (g) PBT, and (h) PA6.
78 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83

Fig. 10. Minimum creep rate versus normalized stress with tensile strength data for (a) PO, PP, PP-T, and PP-G and, (b) PA66, PPE/PS, PBT, and PA6 at different temperatures.

Fig. 11. Correlation of (a) ε_ min and tR using MonkmaneGrant relation and (b) ε_ min , tR and εR using modified MonkmaneGrant relation, for PP-G at 23, 85 and 120  C.

temperatures or under higher stress levels. Time-temperature su-
perposition (TTS) is a well-known procedure frequently applied
either to determine the temperature dependence of the viscoelastic
behavior of polymeric materials, or to expand the time regime at a
given temperature at which the material behavior is studied from a
reference condition. This procedure is developed for prediction of
linear viscoelastic properties of homogeneous polymers and may
not be accurate when applied to semi-crystalline polymers [28]. In
addition, the change of temperature causes a variation of the vol-
ume of the material, and consequently material density change.
Due to these limitations, applying this method to experimental data
of thermoplastic composites may result in inaccurate predictions.
In an analogous manner to the TTS procedure, a timeestress
superposition (TSS) approach has been developed which is used for
prediction of viscoelastic properties at lower stress levels (long-
term creep) from shifting the resultant data at higher stress levels
(short-term creep) at a constant temperature. This method was
been applied successfully for thermoplastic and their composites
[5,7,28,43,44] and is expressed in the following form for creep
strain:
 
t
εðS1 ; tÞ ¼ ε S2 ; (13)
fS
where S1 and S2 are applied stresses and 4S is the stress shift factor.
This method was successfully applied to creep strain data of PO, PP,
PP-T and PP-G at 23  C and 85  C, while the shifting method was not
successful at 120  C for PP-G and 125  C for PO and PP. Therefore, it
appears that there is a temperature limit for TSS principle, like the
limit for Williams-Landel-Ferry (WLF) equation which is used to
relate temperature shifting factor to temperature in TTS, and is
valid for the temperature range between Tg and Tgþ100  C [45].
The creep master curve and corresponding shift factors were
obtained by shifting the short-time (about an hour) creep strain
curves along the time axis with respect to the reference curve at
each stress level and each temperature. The TSS procedure is shown
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
Fig. 12. Correlation of ε_ min and tR using MonkmaneGrant relation for (a) all the studied materials. (b) Correlation of ε_ min , tR and εR using modified MonkmaneGrant relation for PO,
PP, PP-T, and PP-G.
in Fig. 13(a) for PP-G at 85  C for a reference stress of 11 MPa. The
master curves and corresponding shift factors versus stress in a
logelog plot are shown in Fig. 13(b) and (c) for PP-G at 23  C and
85  C. A line was fitted to shift factors versus stress data in a logelog
plot, the same as that in the study of Hadid et al. [28].
As mentioned earlier, only the first one hour data of creep strain
curves were used to generate the master curves and shift factors. It
can be seen from Fig. 13(d) and (e) that the prediction of creep
strain curves using TSS are very good and better than the Findley
model (see Fig. 8) for PP-G at 23  C and 85  C for different stress
levels. Similar good predictions were obtained for PO, PP, and PP-T.
The creep strain master curves for PO and PP-T at 23  C and 85  C
and for PP at 23  C and the corresponding shift factors versus
normalized stress with Su data are shown in Fig. 14. It can be seen
that the shift factor versus normalized stress data are correlated at
23  C for PO, PP, PP-T, and PP-G. At 85  C, however, PO data are not
correlated with PP-T and PP-G.
Obtained time to rupture of the master curve at the reference
stress levels can be shifted using shift factors at each stress level to
predict time to ruptures at other stress levels. As can be seen in
Fig. 15, nearly all the data predicted using this procedure are in
79
scatter bands of ±2, indicating the capability of TSS method in
prediction of creep rupture data.
6.2. Experimental results versus predictions
To evaluate the applicability of predictive models for long time
creep condition, which is of interest in real applications, a limited
number of long-term (up to 40 days) creep tests were conducted for
PP-G at 23  C. As shown in Fig. 16(a), 10 days and 40 days creep
rupture data of PP-G at 23  C are consistent with the extrapolated
creep rupture curve fitted to short-term creep test data. The Lar-
soneMiller model (Equations (4) and (5)) predicts the time to
ruptures at 23  C for PP-G to be 197 and 720 h for the stress levels of
25.5 and 23 MPa, respectively. As compared with the experimental
time to rupture of 273 and 1304 h, the predictions are within a
factor of less than two. The predicted time to rupture data for the
two stress levels using TSS included in Fig. 15, indicates that pre-
dictions for long-term tests are also in the scatter band of two.
The TSS method developed using the first one hour creep strain
curves of short-term tests not only can predict the creep strain of
the rests of those tests, but can also be used for predictions at other
80 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83

Fig. 13. (a) TSS procedure for generating master curve and shift factors for PP-G at 85  C. (b) Master curves and (c) corresponding shift factor versus stress plots for PP-G at 23 and
85  C. Comparison of predictive creep strain (dash lines) using TSS and experimental creep strain curves for PP-G for different stress levels at (d) 23  C and (e) 85  C.

stress levels. As can be seen in Fig. 16(b), the creep strain pre-
dictions for long-term tests are very good for both stress levels of
25.5 and 23 MPa. For long-term creep predictions, Findley [46]
suggested that the constants of the power law model can be ob-
tained from the early creep strain data. The Findley power law
predictions were not as good as those based on TSS, however.
The calculated minimum creep rate of long-term tests from
experimental data are 0.011 and 0.0021%/hour for tests conducted
under stress levels of 25.5 and 23 MPa, respectively. Predicted
minimum creep rate values from Equation (9) are 0.014 and
0.0037%/hour and from Equation (11) are 0.015 and 0.0013 for tests
conducted under stress levels of 25.5 and 23 MPa, respectively.
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83 81

Fig. 14. Creep strain master curves of (a) PO at 23  C and 85  C, (b) PP at 23  C and, (c) PP-T at 23  C and 85  C. Normalized plot of shift factor versus stress using Su for PO, PP, PP-T,
and PP-G at 23  C and 85  C.

or aging effects can be incorporated in the generated TSS master

curves by additionally shifting the curves according to specimen
aging times [6,7,43].

7. Conclusions

Based on the observed experimental behavior and the analysis

performed, the following conclusions can be made:

1) Temperature significantly affected creep behavior of neat, talc-

Fig. 15. Predicted versus experimental time to rupture for PO, PP, PP-T and PP-G at
23  C and 85  C using the TSS technique.
Comparisons between experimental and predictive minimum
creep rate values show that the minimum creep rate correlations
based on short-term creep data give acceptable predictions for
long-term tests.
As thermal and hygrothermal aging affect the mechanical
behavior of some polymers and polymer composites in a beneficial
or a detrimental way [11], they can cause error for long-term creep
predictions based on short-term creep data. To overcome this issue,
short-term creep tests can be conducted on already aged samples
filled, and short glass fiber reinforced thermoplastics. For short
fiber reinforced thermoplastics, the effect of temperature was
more accentuated in the transverse direction, as compared with
the longitudinal direction.
2) For polypropylene composites, the type of reinforcement (talc or
short glass fiber) had the same effect on creep deformation.
Reinforcement or filler did not change the creep mechanism
noticeably and only resulted in improved creep strength (i.e.
reduction of both creep strain and rate), as compared to the neat
polymer.
3) The LarsoneMiller parameter widely used for creep rupture
correlations of metallic materials was used to correlate creep
rupture data of all the materials. This parameter could also be
used to predict long-term creep rupture data, based on short-
term data.
4) The Findley power law model and time stress superposition
(TSS) method represented creep strain curves well. The TSS
method could be used to predict long-term creep curves based
on short-term creep data. This method could also satisfactory
predict creep rupture data.
5) Minimum creep rate data were very well correlated by the
MonkmaneGrant relation and its modified version. Using the
82 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
Fig. 16. (a) Long-term (up to 40 days) creep rupture data superimposed on short-term (up to 3 days) creep rupture curve and, (b) comparison of experimental long-term creep
strain (data points) and predicted creep strain curve using TSS for PP-G at 23  C.
modified version, creep rupture strain, time to creep rupture,
and creep rate are related together.
Acknowledgments
Funding of this study was provided by General Motors. Technical
support of Dr. A.K. Khosrovaneh and Mr. T. Wang is appreciated.
Seyyedvahid Mortazavian and Stephen R. Mellott conducted creep
tests of PO, PP, PBT, and PA6.
References
[1] Drozdov AD. Creep rupture and viscoelastoplasticity of polypropylene. Eng
Fract Mech 2010;77(12):2277e93.
[2] ASTM D2990-09. Standard test methods for tensile, compressive, and flexural
creep and creep-rupture of plastics. West Conshohocken, PA: ASTM Interna-
tional; 2009.
[3] Hamouda HBH, Simoes-betbeder M, Grillon F, Blouet P, Billon N, Piques R.
Creep damage mechanisms in polyethylene gas pipes. Polymer 2001;42(12):
5425e37.
[4] Dean GD, Broughton W. A model for non-linear creep in polypropylene. Polym
Test 2007;26(8):1068e81.
[5] Luo Wb, Wang Ch, Vu-Khanh T, Jazouli S. Time-stress equivalence: application
to nonlinear creep of polypropylene. J Cent South Univ Technol 2007;14(1):
310e3.
[6] Zhao RG, Chen CZ, Li QF, Luo WB. Effects of stress and physical ageing on
nonlinear creep behavior of poly(methyl methacrylate). J Cent South Univ
Technol 2008;15(1):582e8.
[7] Qaiser A, Price J. Estimation of long-term creep behavior of polycarbonate by
stress-time superposition and effects of physical aging. Mech Time-Depend
Mater 2011;15(1):41e50.
[8] Spathis G, Kontou E. Creep failure time prediction of polymers and polymer
composites. Compos Sci Technol 2012;72(9):959e64.
[9] Teoh SH, Cherry BW, Kausch HH. Creep rupture modelling of polymers. Int J
Damage Mech 1992;1(2):245e56.
[10] Seltzer R, Mai YW, Frontini PM. Creep behaviour of injection moulded poly-
amide 6/organoclay nanocomposites by nanoindentation and cantilever-
bending. Compos Part B Eng 2012;43(1):83e9.
[11] Eftekhari M, Fatemi A. Tensile, creep, and fatigue behaviors of short fiber
reinforced polymer composites at elevated temperatures: a literarture survey.
Fatigue Fract Eng M 2015;38(12):1395e418.
[12] Mortazavian S, Fatemi A. Effects of fiber orientation and anisotropy on tensile
strength and elastic modulus of short fiber reinforced polymer composites.
Compos Part B Eng 2015;72:116e29.
 M. Eftekhari, A. Fatemi / Composites Part B 97 (2016) 68e83
[13] Eftekhari M, Fatemi A. Tensile behavior of thermoplastic composites including
temperature, moisture, and hygrothermal effects. Polym Test 2016;51:
151e64.
[14] Mortazavian S, Fatemi A. Tensile and fatigue behaviors of polymers for
automotive applications. Mater Werkst 2015;46(2):204e13.
[15] Mortazavian S, Fatemi A. Fatigue behavior and modeling of short fiber rein-
forced polymer composites including anisotropy and temperature effects. Int J
Fatigue 2015;77:12e27.
[16] Eftekhari M, Fatemi A. On the strengthening effect of increasing cycling fre-
quency on fatigue behavior of some polymers and their composites: experi-
ments and modeling. Int J Fatigue 2016;87:153e66.
[17] Eftekhari M, Fatemi A, Khosrovaneh A. Fatigue behavior of neat and short
glass fiber reinforced polymers under two-step loadings and periodic over-
loads. SAE Int J Mater Manuf 2016;9(3).
[18] Mellott SR, Fatemi A. Fatigue behavior and modeling of thermoplastics
including temperature and mean stress effects. Polym Eng Sci 2014;54(3):
725e38.
[19] Arrhenius SA. Über die Dissociationsw€ arme und den Einflusß der Temperatur
auf den Dissociationsgrad der Elektrolyte. Z Phys Chem 1889;4:96e116.
[20] Larson FR, Miller J. A time-temperature relationship for rupture and creep
stresses. Trans ASME 1952;74:765e71.
[21] Manson SS, Haferd AM. A linear time-temperature relation for extrapolation
of creep and stress-rupture data. Technical Note 2890. NACA; 1953.
[22] Sherby OD, Dorn JE. Creep correlations in alpha solid solutions of aluminum.
Trans AIME 1952;194.
[23] Manson SS, Succop G. Stress-rupture properties of inconel 700 alloy and their
correlation with some time-temperature curve parameters. In: Heat-resisting
metallic materials [Russian translation], IL, Moscow; 1958.
[24] Manson SS, Muralidharan U. Analysis of creep rupture data for five multi-heat
alloys by the minimum commitment method using double het term centering
technique. Palo Alto, CA: EPRI CS 317, Electric Power Research Institute; 1983.
[25] Wilson DJ, Marrone RE, Freeman JW. Larson-Miller and Manson-Haferd
parameter extrapolation of rupture data for type 304 (18Cr-8Ni), grade 22
(2-1/4Cr-1Mo) and grade 11 (1-1/4Cr-1/2Mo-3/4Si) steels. Ann Arbor, Mich-
igan: The University of Michigan, Office of Research Administration; 1968.
[26] Raghavan J, Meshii M. Creep of polymer composites. Compos Sci Technol
1998;57(12):1673e88.
[27] Plaseied A, Fatemi A. Tensile creep and deformation modeling of vinyl ester
polymer and its nanocomposite. J Reinf Plas Comp 2009;28(14):1775e88.
[28] Hadid M, Rechak S, Tati A. Long-term bending creep behavior prediction of
injection molded composite using stressetime correspondence principle.
Mater Sci Eng A Struct 2004;385(1e2):54e8.
[29] Chevali VS, Dean DR, Janowski GM. Flexural creep behavior of discontinuous
thermoplastic composites: non-linear viscoelastic modeling and
83
timeetemperatureestress superposition. Compos Part A Appl S
2009;40(6e7):870e7.
[30] Hadid M, Rechak S, Zouani A. Empirical nonlinear viscoelastic model for in-
jection molded thermoplastic composite. Polym Compos 2002;23(5):771e8.
[31] Kouadri-Boudjelthia A, Imad A, Bouabdallah A, Elmeguenni M. Analysis of the
effect of temperature on the creep parameters of composite material. Mater
Des 2009;30(5):1569e74.
[32] Findley WN. Mechanism and mechanics of creep of plastics. SPEJ 1960;16:
57e65.
[33] Gupta Vidya Bhushan, Lahiri J. Non linear viscoelastic behavior of poly-
propylene and glass reinforced polypropylene in creep. J Compos Mater
1980;14(4):286e96.
[34] Lai JSY, Findley WN. Elevated temperature creep of polyurethane under
nonlinear torsional stress with step changes in torque. Trans Soc Rheol
1973;17(1):129e50 (1957-1977).
[35] David W, Scott JSL. Creep behavior of fiber-reinforced polymeric composites: a
review of the technical literature. J Reinf Plast Comp 1995;14(6):588e617.
[36] Mortazavian S. Fatigue behavior and modeling of short fiber reinforced
polymer composites [Doctoral Desertation]. Toledo, Ohio: University of
Toledo; 2015.
[37] Mellott SR. Tensile, creep, and fatigue behaviors of thermoplastics including
thickness, mold flow direction, mean stress, temperature, and loading rate
effects [Master thesis]. Toledo, Ohio: University of Toledo; 2012.
[38] Norton FH. The creep of steel at high temperature. New York: McGraw Hill;
1929.
[39] Pang F, Wang CH. Activation theory for creep of woven composites. Compos
Part B Eng 1999;30(6):613e20.
[40] Monkman CF, Grant NJ. An empirical relationship between rupture life and
minimum creep rate in creep-rupture tests. ASTM Proc 1956;56:593e620.
[41] Dobes F, Milicka K. The relation between minimum creep rate and time to
fracture. Metal Sci 1976;10:382e4.
[42] Povolo F. Comments on the Monkman-Grant and the modified Monkman-
Grant relationships. J Mater Sci 1985;20(6):2005e10.
[43] Lai J, Bakker A. Analysis of the non-linear creep of high-density polyethylene.
Polymer 1995;36(1):93e9.
[44] Jazouli S, Luo W, Bremand F, Vu-Khanh T. Application of timeestress equiv-
alence to nonlinear creep of polycarbonate. Polym Test 2005;24(4):463e7.
[45] Malcolm LW, Landel RF, Ferreira JD. The Temperature dependence of relaxa-
tion mechanisms in amorphous polymers and other glass-forming liquids.
J Am Chem Soc 1955;77:3701e7.
[46] Findley WN. 26-Year creep and recovery of poly(vinyl chloride) and poly-
ethylene. Polym Eng Sci 1987;27(8):582e5.