Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DOI 10.1007/s11043-012-9168-y
Received: 29 April 2011 / Accepted: 27 January 2012 / Published online: 25 February 2012
© Springer Science+Business Media, B. V. 2012
Abstract The creep lifetime prediction of unidirectional composite materials under step
loading, based on constant loading durability diagram, is analyzed for the two-step creep
loading condition. For this purpose different nonlinear cumulative-damage laws are revis-
ited and applied to predict creep lifetime. One possible approach to accounting for dam-
age accumulation is provided by the continuum-damage mechanics (CDM). However, the
CDM lifetime expression obtained for constant loading condition presents some drawbacks.
Specifically, the upper stress range is not accommodated by CDM form. A modification of
CDM is proposed, forcing the CDM to capture the short-term creep failure. It is proven that
this modified CDM (MCDM) does not yield the same predictions as the Linear Cumulative-
damage law (Miner’s law). Predictions obtained from the nonlinear cumulative-damage laws
are compared against synthetic lifetime generated by a micromechanical model that simu-
lates unidirectional composites under two-step creep loading condition. Comparable devia-
tions from Miner’s law are obtained by the nonlinear cumulative-damage laws.
1 Introduction
Long-term failure prediction of polymers and polymer-based materials under varying load
condition is still a problem not completely solved. One should expect an increasing com-
plexity if cyclic environmental conditions are taken into account. In the end this can be
treated as a damage-accumulation problem, which depends on the loading and environmen-
tal conditions prevailing at each instant.
For multiple-step creep loading, damage may occur at different rates for different stress
levels. Each stress level leads to different creep lifetimes. The damage accumulation dur-
ing each load step reduces the remaining strength, which may be accounted by appropriate
cumulative-damage laws. Almost 30 years after Zhang and Brinson (1986) affirmed “at
present there is no universally accepted method for estimating the creep life of a mate-
rial subjected to multiple stress levels for various periods of time” the situation is almost the
same although meanwhile significant breakthroughs occurred. The constitutive formalism of
continuum mechanics and irreversible thermodynamics makes it possible to account for the
polymer creep and distributed damage (Abdel-Tawab and Weitsman 1998, 2001; Schapery
1999, 2002). A semi-empirical approach for characterization of viscoelasticity, viscoplas-
ticity and damage in different composites has been successfully applied (Varna 2010). More
recently Asadi and Raghavan (2011) measured the influence of damage, i.e. time-dependent
cracking, on the creep of multidirectional laminate [±45/902 ]S . The effect was important to
affect significantly the creep compliance. Direct measurement of creep cumulative-damage
was also performed by using modified conventional small angle X-ray scattering equipment
by Zachariev et al. (2004, 2009).
Few works have reported experimental data on creep failure under multi-step loading
programs applied to polymers or polymer-matrix composites. The obvious reason for this
situation was pointed by Zhang and Brinson (1986) when they assumed that the “main dif-
ficulty . . . is the need for extensive test programs over a long period of time and the need to
have rigorous, but simple, constitutive models to quantify observed behavior”. This remains
true for polymers and polymer-matrix composites (PMC). Those researchers presented creep
and creep failure data under one and two-step creep loading condition applied to polycar-
bonate and polysulfone (Zhang and Brinson 1986). After observing experimental data, they
proposed a new semi-empirical cumulative-damage rule.
The continuum-damage mechanics (CDM), based on original ideas of Kachanov (1958),
Stigh (2006), provides a practical tool to model the effect of damage propagation. In the
classical approach the damage variable ω accounts for cumulative damage; i.e. from un-
damaged state (ω = 0) to failure (ω = 1). However, the CDM lifetime expression, ob-
tained for failure under constant load, does not capture the short-term creep failure di-
agram. Hence, a modification of the classical failure condition is proposed to remedy
this drawback. This new formulation is combined with a general damage growth law
to obtain a modified continuum-damage mechanics (MCDM). As expected, the expres-
sion obtained from MCDM accommodates typical lifetime over a time range of sev-
eral decades (Guedes 2010). A collateral consequence of this modification is that the
MCDM does not yield the same predictions as the Linear Cumulative-damage law (Miner’s
Law) (Miner 1945).
Other cumulative-damage laws are revisited and tested using a micromechanical model to
simulate creep rupture of unidirectional composites, under one- and two-step creep loading,
by assuming that the matrix is viscoelastic and the fiber is elastic. Under these conditions
the use of a micromechanical model permits one to control the damage mechanisms, by
opposition to experimental tests where “small” changes on the test conditions can accelerate
and promote different damage mechanisms.
2 Cumulative-damage laws
The lifetime prediction under variable stress is still a challenging problem for composite
materials. There are several cumulative-damage laws proposed to tackle this problem. Let
us revisit some of them.
Mech Time-Depend Mater (2012) 16:381–396 383
Most used, for its simplicity, is Miner’s rule or Linear Cumulative-damage law (LCD) (Miner
1945). We have
tf
dt
= 1. (1)
0 tc [σ (t)]
For a multi-step creep loading condition the previous formulation can be written as
N
ti
= 1, (2)
i=1
tc [σi (t)]
where ti is the ith step time span and tc [σi (t)] is the failure time under the creep stress
σi (t).
After review of several empirical approaches to cumulative creep damage, Zhang and Brin-
son (1986), based on experimental observations, assumed that the rate of damage accu-
mulation was related to the transient strain component of viscoelastic strains. After some
considerations it was deduced an empirical cumulative-damage formula as
N α
ti
Ki = 1, (3)
i=1
tc [σi (t)]
where r is the power-law exponent of the power-law function of stress intensity factor, Fa
is the normalized failure function and tc is the creep-rupture lifetime at stress level corre-
sponding to Fa . The exponent r can be obtained directly from the lifetime diagram as the
inverse slope of lifetime linear region in the log–log scale (Christensen 2008) (see Fig. 3).
For a multi-step creep loading condition the previous formulation can be written as
N
{Fai }r ti = Far (t)tc (Fa ), (5)
i=1
where ti is the ith step time span, Fai is the respective step stress and tc (Fa ) is the failure
time under the creep normalized stress Fa .
384 Mech Time-Depend Mater (2012) 16:381–396
where z = t/τ , t is the time variable, τ is a characteristic time associated with the process,
Fa is the normalized failure function that applies to a specific controlling failure mode and
j is a material parameter. This material parameter influences the progression of the damage;
if j < 1 the rate of degradation is greatest at beginning but if j > 1 the rate of degradation
increases as function of time while if j = 1 there is no explicit time dependence in the rate
of degradation. The failure criterion is given by Fr = Fa .
For a multi-step creep loading condition the previous formulation can be written as
N
ti j
Fr = 1 − (1 − Fai ) , (7)
i=1
tc (Fai )
where ti is the ith step time span, Fai is the respective step stress and tc (Fai ) is the failure
time under the creep normalized stress Fai .
where C and ν are material constants. This equation leads to a separable differential equation
for ω(t), assuming ω(0) = 0,
t
ν 1+ν
1 − ω(t) ω̇(t) = Cσ ν (t) ⇒ 1 − 1 − ω(t) = C(1 + ν) σ ν (τ ) dτ. (10)
0
ω(t) = 1 − 1 − C(1 + ν) ν
σ (τ ) dτ . (11)
0
Mech Time-Depend Mater (2012) 16:381–396 385
From the previous relationship, the time to failure for creep is readily obtained assuming
σ (t) = σ0 ,
1
tc = . (13)
C(1 + ν)σ0ν
Since the experimental data, spread over a time range of several decades, cannot be accom-
modated by form (13), a new approach was devised (Guedes 2010). The requirements for
more convenient formulations were already discussed (Christensen and Miyano 2007). Par-
ticularly, the upper stress range is hardly accommodated by CDM form. In this context a
modification of CDM, forcing the CDM to capture the short-term failure, is proposed as
follows.
The conventional rupture criterion assumes ω(tR ) = 1 (Stigh 2006). However, before the
stress becomes too large (ω → 1) rupture takes place almost instantaneously, when the net
stress overcomes the material strength. Therefore, instead of considering ω(tR ) = 1, let us
impose that rupture occurs when the net stress equals the strength, given by
Fa
= 1(= Fr ), (14)
1−ω
where Fa is the applied stress normalized to strength and therefore the normalized strength
equals unity. Since the failure condition assumes that Fa = Fr , the remaining strength is
given by
Fr (t) = 1 − ω(t). (15)
Using Kachanov (1958) and Ding et al. (1994) approaches the damage growth law assumes
the following expression:
Fa (t) ν η−1
ω̇(t) = C t , (16)
1 − ω(t)
where C, η and ν are material constants. This equation leads to a separable differential
equation for ω(t), assuming ω(0) = 0:
ν 1 − (1 − ω(t))1+ν t
1 − ω(t) ω̇(t) = CFaν (t)t η−1 ⇒ =C Faν (τ )τ η−1 dτ. (17)
1+ν 0
The previous result can be rewritten to give the remaining strength, using relationship (14),
as
t 1+ν
1
Fr (t) = 1 − C(1 + ν) ν
Fa (τ )τ η−1
dτ , (18)
0
or
C(1 + ν) t
Faν (τ )τ η−1 dτ = 1, (19)
1 − Fr1+ν (t) 0
386 Mech Time-Depend Mater (2012) 16:381–396
For the purpose of this work let us analyze the application of CDM and MCDM to creep
lifetime prediction under a stepwise constant stress level loading condition.
Using the CDM criterion (5) we obtain
N
C(1 + ν) Faiν ti = 1, (21)
i
which leads to
N
ti
= 1, (22)
i
tc (Fai )
1 C(1 + ν) ν η
N
Fai ti = 1, (23)
η 1 − FaN
1+ν
i
which leads to
η
1
N
ti
1 − Faiν+1 = 1, (24)
1 − FaN
1+ν
i
tc (Fai )
1−F v+1
with tcη (Fai ) = η C(1+v)F
ai
v which represents the total creep lifetime at Fa .
ai
This result differs from the LCD law. In this case the v-exponent dictates how far from
the LCD predictions is the MCDM criterion. In fact when the v-exponent tends to large
values, the MCDM predictions become closer to the LCD law.
If the material is loaded for a certain time, less than the time necessary to cause creep fail-
ure, it does produce some damage to the material. For the LCD law the remaining strength
is given by
N
ti
Fr = 1 − , (25)
i
tc (Fai )
Fig. 1 Two-step creep loading; (a) high–low stress case and (b) low–high stress case
Let us analyze the application of the previous cumulative-damage laws for the particular
case of a two-step loading condition, as represented in Fig. 1. Two different cases can occur,
a higher to lower step transition or a lower to higher load step transition.
For the LCD criterion we obtain
t2 t1
+ = 1. (27)
T2 T1
Starting with the following formulation for the creep lifetime:
η 1 − Fa1+ν
tc (Fa ) = η , (28)
C(1 + ν)Faν
let us assume that η = 1. For a two-step creep load case, as depicted in Fig. 1, the CDF
criterion gives
1 − Fa2
1+ν
v
Fa1 t1 + Fa2
v
t2 = Fa2
v
ν , (29)
C(1 + ν)Fa2
or
1 − Fa1
1+v
t1 t2
+ = 1, (30)
1 − Fa2
1+v T1 T2
1−F 1+ν
where Ti = C(1+ν)F
ai
ν is the total creep lifetime under Fai .
ai
For the two-step creep load case, as depicted in Fig. 1, the MCDM gives
η η 1 − Fa2
1+ν
ν
Fa1 t1 + Fa2
ν
t2 = , (31)
C(1 + ν)
or
η η
1 − F a11+ν t1 t2
+ = 1, (32)
1 − F a21+ν T1 T2
j 1−F a 1+ν
where Ti = j C(1+ν)F
i
aiν
.
For the two-step creep load case, as depicted in Fig. 1, the SEI gives
j j
t1 t2
Fr = 1 − (1 − Fa1 ) − (1 − Fa2 ) . (33)
T1 T2
388 Mech Time-Depend Mater (2012) 16:381–396
where Ti is the lifetime under constant load Fai with no particular requirement about the
creep lifetime expression.
From the previous developments, it may be concluded that MCDM and CDF lead to
similar predictions, provided the creep lifetime expression (28) with η = 1 is used.
4 Micromechanical model
σ = (1 − Vf )σm + Vf σf , (37)
where Vf is the fiber volume fraction. From the above equations the composite strain under
creep load σ = σ0 is derived:
Vf Ef σ − Vf Ef ε n 1
1+ ε̇ − B = σ̇ . (38)
Em (1 − Vf ) (1 − Vf ) Em (1 − Vf )
1
ε(0) = σ0 . (40)
(1 − Vf )Em + Vf Ef
Smax
εmax = , (42)
Ef
For a two-step loading case, see Fig. 1, the modified superposition principle (Findley et al.
1989) can be applied to obtain the creep strain
φ(t, σ1 ) t < t1 ,
ε(t) = (44)
φ(t, σ1 ) − φ(t − t1 , σ1 ) + φ(t − t1 , σ2 ) t ≥ t1 ,
where
1−n
σ 1 − Vf Em
φ(t, σ ) = − σ
Vf Ef Vf Ef (1 − Vf )Em + Vf Ef
1/(1−n)
(n − 1)Vf Em Ef
+B t . (45)
(1 − Vf )Em + Vf Ef
In this case it is not possible to obtain a closed form solution for lifetime, i.e. the solutions
must be determined using a numerical method.
The residual stress can be easily calculated, at any instant, from differential expression
(38) written as
σ ε Vf Ef σ − Vf Ef ε n
= 1+ −B Em (1 − Vf ), (46)
t t Em (1 − Vf ) (1 − Vf )
where σ and ε are the stress and the strain increments, respectively, necessary to produce
failure, starting from the stress and strain state registered at the instant considered. Since for
the instantaneous strength we have t → 0 then
Vf Ef
σ = Em (1 − Vf ) 1 + ε. (47)
Em (1 − Vf )
The present model possess a limitation regarding the stress range, i.e. it is only effective
for σ0 > Vf Smax . As Du and McMeeking (1995) discussed, the CML model is just effective
in predicting analytically the creep-rupture time in short-term range. This results from the
390 Mech Time-Depend Mater (2012) 16:381–396
Fig. 2 Creep compliance obtained for the simulated unidirectional composite using the Curtin–McLean
model (CML) model
fact that CML does not include the stress relaxation in broken fibers. This effect becomes
significant in long-term creep. In the present case σl = Vf Smax may be interpreted as the
stress threshold above which creep damage initiates and propagates.
On the other hand the CML model already proved to be effective, emulating similar
relationship between lifetime diagrams under creep loading and constant stress/strain rate
loading, as observed for real polymer-based composites (Guedes 2009).
Table 1 gives the CML model parameters used for the numerical simulation. The creep
compliance for the CML is plotted in Fig. 2. The nonlinearity effect is very pronounced in
this case.
The general behavior of polymers under the effect of the time, changing from glassy to
rubbery state, deform in an asymptotic way as Brinson (1999) discussed in detail. This re-
sponse is in accordance with the features displayed by the CML model. The creep curve
of typical polymers reaches a plateau after a certain time. Therefore the simple power law
cannot capture the long-term range creep behavior. Furthermore it is expected that the rup-
ture stress becomes constant in the rubbery plateau region of the polymer. Following Yang’s
(1996) research, a variation of the usual generalized power based on the Cole–Cole function
is used to model the creep strain under arbitrarily stress σ (t) as
t
D∞ − D0 ∂σ (τ )
ε(t) = D0 σ (t) + dτ, (48)
0 1 + ( t−τ
τ0 n
) ∂τ
Mech Time-Depend Mater (2012) 16:381–396 391
Fig. 3 Creep lifetime for the simulated unidirectional composite using the Curtin–McLean model (CML)
model
where D0 , D∞ , n are material constants and τ0 a function of stress state. For a constant
applied stress, i.e. a creep load, the equation becomes
D∞ − D0
ε(t) = D0 σ0 + σ0 . (49)
1 + ( τt0 )n
From creep data obtained from the CML model, a nonlinear viscoelastic power law was
fitted. The parameters are presented in Table 2. The symbol σR represents the instantaneous
failure stress.
The creep lifetime can then be obtained from the CML model as plotted in Fig. 3. After-
wards the creep lifetime curve was modeled using (27),
1 − Fa6.3
tc (Fa ) = . (50)
0.0252Fa5.3
Although this expression could not accommodate the asymptotic behavior on long-term, this
particular feature is not relevant for the present case. At this point it is important to retain
the fact that the CML model features mimic the PMC creep behavior concerning creep and
creep-rupture.
Let us now analyze what happens when a two-step load is applied. Two cases are con-
sidered as shown in Table 3. The creep lifetime for Fa = 0.8824 is tf = 43.3 h and for
392 Mech Time-Depend Mater (2012) 16:381–396
Fig. 4 Creep strain calculated for Curtin–McLean model (CML) model compared with Cole–Cole model
for: (a) case 1; (b) case 2 with t1 = 45 h
Fa = 0.7843 is tf = 127.0 h. These values were obtained directly from the CML model. In
Fig. 4 is plotted the creep strain evolution for the particular case when the time for step load
transition is t1 = 45 h (see Fig. 1), for Case 1 and Case 2. The creep compliance obtained
from the CML model compare quite well with Cole–Cole model predictions, given by (49).
The evolution of residual strength was also obtained for the CML model, for the same
cases with the time for step load transition equal to t1 = 45 h (see Fig. 1). The CML residual
strength was compared against two cumulative-damage laws, SEI (j = 1) and MCDM, as
shown in Fig. 5. In this case the SEI predictions are quite close to the micromechanical
model residual strengths. The MCDM predictions deviate from the micromechanical model,
this is more notorious for the Case 2.
The simulation was run to build the cumulative-damage curves, for both Case 1 and
Case 2, as shown in Fig. 6. Both damage laws, i.e. SEI and MCDM, predict that for high–
Mech Time-Depend Mater (2012) 16:381–396 393
Fig. 5 Remaining strength obtained for Curtin–McLean model (CML) model compared against strength
evolution integral (SEI) and modified continuum-damage mechanics (MCDM) predictions for: (a) case 1;
(b) case 2 with t1 = 45 h
low load sequences Miner’s rule (LCD) summation is always greater than 1 and for low–high
load sequences it is always less than 1. This is corroborated by the CML cumulative-damage
curves. In this case the cumulative-damage curves predicted by SEI are quite close to those
obtained for the CML micromechanics model.
This behavior was observed in experiments performed on austenitic stainless steel at
700°C (Jansson 1985), as described by Stigh (2006). Similarly to the present study, two
cases were tested; “step-up” and “step-down” with the second stress level higher and lower
than the first stress level, respectively. Like in the present case, step-up registered a lower
life-fraction at rupture than step-down.
It is important to note that the cumulative-damage laws do not replicate the underlying
mechanisms; they simply evaluate damage accordingly to an imposed evolution law. Never-
394 Mech Time-Depend Mater (2012) 16:381–396
Fig. 6 Cumulative creep to rupture for Curtin–McLean model (CML) model compared against strength
evolution integral (SEI) and modified continuum-damage mechanics (MCDM) predictions for: (a) case 1;
(b) case 2
theless Loverich et al. proved that SEI was effective to predict the combined effects of time
and cyclic processes for a polymer-matrix composite (Loverich 1997).
Mikhailov and Namestnikova (2011) developed and applied the concept of temporal
normalized equivalent stress functional, TNESF, for lifetime prediction under creep load-
ing. The mathematical formalism and properties of TNESF were previously detailed by
Mikhailov (2003). The TNESF is a scalar related with mechanical material characteristic,
to be determined directly from durability tests. No information about damage growth is
necessary. Accordingly, the nonlinear (power-type) rule of normalized equivalent stress ac-
cumulation lead to the following failure criterion for two-step loading condition (Mikhailov
and Namestnikova 2011):
b 1/b 1/b
t1 Fa1 t2 Fa1 t2
+ + 1− = 1, (51)
T1 Fa2 T2 Fa2 T2
Mech Time-Depend Mater (2012) 16:381–396 395
where b is a material constant obtained from the durability diagram under constant loading
(see Fig. 3), described by a power-type relation, tf ∼ (Fa )−b . In Fig. 6 are plotted the curves
for (51) (TNESF) for two different exponential constants (b = 4.5 and b = 7.0). The value
obtained from durability diagram (b = 5.3) (see Fig. 3) belongs to the range defined by the
previous values. These values recover MCDM and SEI curves, respectively.
All these results are encouraging indicators about the potential effective performance of
nonlinear cumulative-damage laws, especially SEI, regarding lifetime prediction of compos-
ites under multi-step creep loading. However, the real proof, preformed with polymer-matrix
composites experimental data, is still lacking.
6 Conclusions
The quest on how to predict lifetime under multi-step creep loading of unidirectional re-
inforced composites is still an open problem. The difficulty to tackle these matters arises
from the fact that there is a lack of experimental data under multi-step creep loading on
polymer-based composites. The crucial factors are the time consumed and the cost of these
experiments.
Within this context several cumulative-damage laws were revisited. A modification of
the classical failure condition is proposed in the domain of continuum-damage mechanics,
which lead to a modified continuum-damage mechanics (MCDM). The purpose was to force
the CMD to capture the short-term creep failure, which is not accommodated by CDM
form. Contrary to the classical approach, MCDM does not yield the same predictions as the
Linear Cumulative-damage law (Miner’s Law), i.e. MCDM is a history-sensitive damage-
accumulation rule.
A micromechanical model was used to generate lifetime data of unidirectional compos-
ites, under one- and two-step creep loading. The model assumes that the matrix is viscoelas-
tic and the fiber is elastic. The micromechanical model controls the damage mechanisms,
contrary to the experimental tests where “small” changes on the test conditions can acceler-
ate and promote different damage mechanisms.
The synthetic data generated from the micromechanical model was used to assess the
nonlinear cumulative-damage laws predictions. The damage laws predict that for high–low
load sequences Miner’s rule (LCD) summation is always greater than 1 and for low–high
load sequences it is always less than 1. This was corroborated by the micromechanical
model, with SEI predictions in close agreement with synthetic data. The MCDM predic-
tions on strength evolution showed some deviations. Yet the final proof remains to be done
with experimental data which are lacking for polymer-based composites.
References
Zhang, M.J., Brinson, H.F.: Cumulative creep damage for polycarbonate and polysulfone. Exp. Mech. 26,
155–162 (1986)
Abdel-Tawab, K.H., Weitsman, Y.J.: A coupled viscoelasticity/damage model with application to swirl-mat
composites. Int. J. Damage Mech. 7, 351–380 (1998)
Abdel-Tawab, K.H., Weitsman, Y.J.: A strain-based formulation for the coupled viscoelasticity/damage be-
haviour. J. Appl. Mech. 68, 304–311 (2001)
Schapery, R.A.: Nonlinear viscoelastic and viscoplastic constitutive equations with growing damage. Int. J.
Fract. 97, 33–66 (1999)
Schapery, R.A.: Homogenized constitutive equations for linear viscoelastic unidirectional composites with
growing transverse cracks. Mech. Time-Depend. Mater. 6, 101–131 (2002)
396 Mech Time-Depend Mater (2012) 16:381–396
Varna, J.: Characterization of viscoelasticity, viscoplasticity and damage in composites. In: Guedes, R.M.
(ed.) Creep and Fatigue in Polymer Matrix Composites, pp. 514–542. Wood Head Publishing Ltd, Cam-
bridge (2010)
Asadi, A., Raghavan, J.: Influence of time-dependent damage on creep of multidirectional polymer composite
laminates. Compos., Part B, Eng. 42, 489–498 (2011)
Zachariev, G., Rudolph, H.-V., Ivers, H.: Damage accumulation in glass fibre reinforced polyoximethylene
under short-term loading. Composites, Part A, Appl. Sci. Manuf. 35, 1119–1123 (2004)
Zachariev, G., Rudolph, H.-V., Ivers, H.: Prediction of creep-damage accumulation in glass fibre reinforced
polyoximethylene. Mater. Sci. Eng. A, Struct. Mater.: Prop. Microstruct. Process. 510–511, 425–428
(2009)
Kachanov, L.M.: On the time to failure under creep conditions. Izv. Akad. Nauk SSSR, Otd. Teh. Nauk 8,
26–31 (1958)
Stigh, U.: Continuum damage mechanics and the life-fraction rule. J. Appl. Mech. 73, 702–704 (2006)
Guedes, R.M.: A cumulative damage law based on continuum damage mechanics approach. In: 14th Euro-
pean Conference on Composite Materials, 7–10 June 2010, Budapest, Hungary (2010). Paper ID: 945
Miner, M.A.: Cumulative damage in fatigue. J. Appl. Mech. 12, A159–A164 (1945)
Christensen, R.M.: A physically based cumulative damage formalism. Int. J. Fatigue 30, 595–602 (2008)
Reifsnider, K.L., Stinchcomb, W.W.: A critical element model of the residual strength and life of fatigue–
loaded composite coupons. In: Hahn, H.T. (ed.) Composite Materials: Fatigue and Fracture (ASTM
STP 907), pp. 298–313. American Society for Testing and Materials, Philadelphia (1986)
Reifsnider, K.L., Case, S.C., Duthoi, J.: The mechanics of composite strength evolution. Compos. Sci. Tech-
nol. 60, 2539–2546 (2000)
Reifsnider, K.L., Case, S.C.: Damage Tolerance and Durability in Material Systems. Wiley-Interscience, New
York (2002)
Christensen, R.M., Miyano, Y.: Deterministic and probabilistic lifetimes from kinetic crack growth—
Generalized forms. Int. J. Fract. 143, 35–39 (2007)
Ding, J.L., Liu, K.C., Brinkman, C.R.: A phenomenological high-temperature deformation and life prediction
model for advanced silicon nitride ceramics. In: Noor, A.K., Reifsnider, K.L. (eds.) Durability and
Damage Tolerance, vol. 43, pp. 177–190. American Society of Mechanical Engineers, New York (1994)
Du, Z.Z., McMeeking, R.M.: Creep models for metal matrix composites with long brittle fibres. J. Mech.
Phys. Solids 43, 701–726 (1995)
McLean, M.: Creep deformation of metal-matrix composites. Compos. Sci. Technol. 23, 37–52 (1985)
Curtin, W.A.: Theory of mechanical properties of ceramic-matrix composites. J. Am. Ceram. Soc. 74, 2837–
2845 (1991)
Koyanagi, J., Kiyota, G., Kamiya, T., Kawada, H.: Prediction of creep rupture in unidirectional composite:
Creep rupture model with interfacial debonding and its propagation. Adv. Compos. Mater. 13, 199–213
(2004)
Findley, W.N., Lai, J.S., Onaran, K.: Creep and Relaxation of Nonlinear Viscoelastic Materials. Dover, New
York (1989)
Guedes, R.M.: Relationship between lifetime under creep and constant stress rate for polymer-matrix com-
posites. Compos. Sci. Technol. 69, 1200–1205 (2009)
Brinson, H.F.: Matrix dominated time dependent failure predictions in polymer matrix composites. Compos.
Struct. 47, 445–456 (1999)
Yang, Q.: Nonlinear viscoelastic-viscoplastic characterization of a polymer matrix composite. Ph.D. disser-
tation, Free University of Brussels (V.U.B.) (1996)
Jansson, S.: Damage, crack growth and rupture in creep. Ph.D. Thesis, Division of Solid Mechanics, Chalmers
University of Technology, Göteborg, Sweden (1985)
Loverich, J.: Life prediction of composite armor in an unbonded flexible pipe. MSc dissertation, Virginia
Polytechnic Institute and State University. Blacksburg, Virginia (1997)
Mikhailov, S.E., Namestnikova, I.V.: History-sensitive accumulation rules for life-time prediction under vari-
able creep loading. Arch Appl Mech. (2011). doi:10.1007/s00419-011-0511-6
Mikhailov, S.E.: Theoretical backgrounds of durability analysis by normalized equivalent stress functional.
Math. Mech. Solids 8, 105–142 (2003)