Mechanics Research Communications 36 (2009) 428–436

Contents lists available at ScienceDirect

Mechanics Research Communications

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

Non-isothermal low-cycle fatigue analysis of elasto-viscoplastic materials

Heraldo da Costa-Mattos a,*, Pedro Manuel Calas Lopes Pacheco b
a
Laboratório de Mecânica Teórica e Aplicada, Programa de Engenharia Mecânica, Universidade Federal Fluminense, Rua Passo da Pátria 156,
24210-240 Niterói, RJ, Brazil
b
Departamento de Engenharia Mecânica, Centro Federal de Educação Tecnológica Celso Suckow da Fonseca, Av. Maracanã 229,
Maracanã – 20271-110 Rio de Janeiro, RJ, Brazil

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

Article history: A non-isothermal elasto-viscoplastic model, developed within the framework of contin-
Received 14 March 2008 uum damage mechanics, is proposed to investigate the influence of thermomechanical
Received in revised form 2 December 2008 coupling on fatigue life of metallic materials. The numerical simulation of constant-ampli-
Available online 16 December 2008
tude low-cycle fatigue testing of a 316L stainless steel at room temperature shows that the
hypothesis of isothermal processes may be inadequate when cyclic inelastic deformations
are involved. The results show that the specific heat can decrease when the fatigue damage
Keywords:
is high, causing a localized increase of the temperature even if the amount of heat gener-
Low-cycle fatigue
Elasto-viscoplasticity
ated is not so high and, hence, promoting a reduction in lifetime.
Heat transfer Ó 2008 Elsevier Ltd. All rights reserved.
Thermomechanical coupling
Continuum damage mechanics

1. Introduction

Inelastic cyclic deformation promotes heating of metallic structural elements. For high loading rates and/or high ampli-
tudes of inelastic deformation, a considerable amount of heat can be generated (Simo and Miehe, 1992; Pacheco and Costa-
Mattos, 1997; Stablere and Baker, 2000; Rosakis et al., 2000; Longère and Dragon, 2008). However, in traditional low-cycle
fatigue models, the variation of the material temperature due to thermomechanical coupling is not considered and unreal
life predictions may be obtained. Indeed, there are situations where such coupling cannot be neglected and a physically real-
istic model must consider it. Since temperature variation can interfere on the fatigue phenomena and most classical low-cy-
cle fatigue models only take into account isothermal processes, the ASTM standard for low-cycle fatigue testing (ASTM,
1980) establishes that the gradient of temperature during the testing program must not exceed a range of ±2 K. For high
inelastic amplitudes the standard recommends the use of cooling devices and low loading frequencies maintain the speci-
men temperature within the established range. However, this is a difficult condition to achieve in real mechanical compo-
nents in operation. In this paper, a continuum damage mechanics model is proposed to study the thermomechanical
coupling effects on the life prediction of metallic structures submitted to cyclic inelastic loadings. A thermodynamic ap-
proach permits a rational identification of the thermomechanical coupling in the mechanical and thermal equations. Numer-
ical simulations of austenitic stainless steel (AISI 316L) bars submitted to cyclic loadings are presented and analyzed.

2. First and second laws of thermodynamics

To set up a general constitutive theory it is necessary to consider aspects of the first and second law of thermodynamics
since heat transfer and dissipative behavior must be taken into account. Considering the hypothesis of small transformations

* Corresponding author. Tel./fax: +55 21 2629 5585.

E-mail address: heraldo@mec.uff.br (H. da Costa-Mattos).

0093-6413/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechrescom.2008.12.003
 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436 429

and under suitable regularity assumptions it is possible to consider the following expressions as local versions of the first law
(FLT) and second law of thermodynamics (SLT)

FLT : qe_ ¼ div q þ r : e_ þ q r; SLT : d ¼ r : e_  qðw_ þ shÞ

_ þ q:g P 0 ð1Þ

where a dot denotes the material time derivative; q is the mass density; r the stress tensor; e the total strain (the symmetric
part of the displacement gradient); e the internal energy per unity mass, hd the absolute temperature; s the total entropy per
unit mass; w ¼ e  hs the Helmholtz free energy per unit mass; q the heat flux vector and g ¼ gradðlog hÞ, log h standing for
the logarithm of the absolute temperature. is the rate of energy dissipation per unit volume. The second law of thermody-
namics makes a distinction between possible processes ðd P 0Þ and impossible processes ðd < 0Þ. The possible processes
may be reversible (the rate of energy dissipation d is always equal to zero) or not. This local version of the SLT does not ex-
clude the possibility of unusual behaviors such as a decreasing temperature if heat is added to the medium. To exclude the
possibility of such kind of unusual behavior, here we only consider fluids that always satisfy a further restrictive constraint:

d1 ¼ r : e_  qðw_ þ shÞ
_ P 0; d2 ¼ q  g P 0 ð2Þ

Obviously, if the above relations are satisfied, then the local version of the SLT presented in (1) will also be satisfied. It is
also simple to verify that (2) leads to the classical heat conduction inequality q  gradðhÞ P 0 since the absolute temperature
h is a positive quantity. This relation implies that heat flows in the direction of decreasing temperature when q is parallel to
the temperature gradient. The quantity d1 , defined in (2), is usually called the intrinsic dissipation and the quantity d2 , the
thermal dissipation. From the definition of the Helmholtz free energy per unit mass w and of the intrinsic dissipation d1 it is
possible to obtain the following alternative local form for the first law of thermodynamics which will be useful to analyze the
thermomechanical couplings in the next sections:

ow_
div q þ qr ¼ d1  qh ð3Þ
oh

3. Elasto-viscoplastic model

The model used in this paper is developed within the framework of the thermodynamics of the irreversible processes.
Such model is a generalization of the classic elasto-viscoplastic model proposed by Lemaitre and Chaboche (1990) for iso-
thermal processes. For an elasto-viscoplastic material, the thermodynamic state is completely determined by a finite set
of state variables, also called thermodynamic or independent variables. The state variables are the so-called observable vari-
ables – the total strain e and the absolute temperature - and a set of internal variables related to dissipative mechanisms –
the plastic strain variable ep , the isotropic hardening variable p, the kinematic hardening variable c and the damage variable
D. The macroscopic quantity D ð0 6 D 6 1Þ represents the material local degradation. When D = 0 the material is in a virgin
state and when D = 1 the material is completely damaged. In this model, the damage variable can be interpreted in terms of
energy or as a measure of the link between material points and is not related to a volume fraction of voids. A further discus-
sion about this interpretation of the damage variable can be found in Costa-Mattos et al. (1992). No matter the interpretation
of the damage variable, the analysis of the different possibilities the free energy can be affected by the damage variable
(which is the goal of this paper) is crucial to define how it can affect the specific heat C and provides a rational methodology
to understand the adequate form of free energy and specific heat in coupled thermo-elasto-viscoplasticity with isotropic
damage.
The elasto-viscoplastic behavior is characterized by two thermodynamic potentials: the Helmholtz free energy w and the
potential of dissipation / . The Helmholtz free energy is expressed as:
qwðe; ep ; p; c; D; hÞ ¼ ð1  DÞ½W e ðe  ep ; hÞ þ W p ðp; hÞ þ W c ðc; hÞ  W h ðhÞ ð4Þ

where the energy densities W e , W p , W c and W h have the following form:

E m aE
W e ðe  ep ; hÞ ¼ ½ðe  ep Þ : ðe  ep Þ þ ½trðe  ep Þ2   ðh  h0 Þtrðe  ep Þ ð5Þ
2ð1 þ mÞ ð1  2mÞ 1  2m
Z h
1 a q
W p ðp; hÞ ¼ bðp þ edp Þ; W c ðc; hÞ ¼ ðc : cÞ; W h ðhÞ ¼ q p1 logðnÞdn þ p2 h2 ð6Þ
d 2 h0 2
E, m, a, b, d and a are temperature-sensitive material parameters, p1 w and p2 are positive constants and h0 a reference
temperature. The so-called thermodynamic forces ðr; Bp ; Bo ; BD ; sÞ, associated to the state variables ðe; p; c; D; hÞ, are defined
from , as follows (a discussion about the role of the thermodynamic forces can be found in Lemaitre and Chaboche, 1990):

ow ow ow ow ow
r¼q ; Bp ¼ q ; Bc ¼ q ; BD ¼ q ; s¼ ð7Þ
oe op oc oD oh

An additional set of constitutive equations, called the evolution laws, which characterizes the evolution of the dissipative
processes, is obtained from the potential of dissipation / :
430 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436

o/ o/ o/ o/

e_ p ¼ ; p_ ¼ p ; c_ ¼ c ; D_ ¼ D ð8Þ
or oB oB oB
The potential of dissipation / is supposed to be a function of ðr; Bp ; Bo ; BD Þ with the following form:
* +nþ1
p D k Fðr; Bp ; Bc ; BD ; e; ep ; p; c; D; hÞ
/1 ð c
r; B ; B ; B ; e; e p
; p; c; D; hÞ ¼ ð9Þ
nþ1 k

where F is the yielding function given by:

Fðr; Bp ; Bc ; BD ; e; ep ; p; c; D; hÞ ¼ f ðr; Bp ; Bc ; DÞ þ gðBD ; e; ep ; p; c; D; hÞ ð10:1Þ

p c c p
f ðr; B ; B ; DÞ ¼ Jðr þ B Þ þ B  ð1  DÞrp ð10:2Þ
u 2 1
D
gðB ; e; e p
; p; c; D; hÞ ¼ c c
½ðB : B Þ  ½að1  DÞ ðc : cÞ þ ½ðBD Þ2  ðW e ðe  ep ; hÞ þ W p ðp; hÞ þ W c ðc; hÞÞ2  ð10:3Þ
2a 2S0
with Jðr þ Bc Þ ¼ ½1:5ðr þ Bc Þdev : ðr þ Bc Þdev 1=2 being the von Mises equivalent stress where ()dev represents the deviatoric
part of a tensor. hFi ¼ maxf0; Fg. Using the state laws (7) it is possible to verify that the perturbation term
gðBD ; e; ep ; p; c; D; hÞ in (10) is always equal to zero, but its derivative is not necessarily zero (this can be more easily verified
using the alternative expressions (A.3) and (A.4) for the state laws presented at the appendix). Hence it introduces additional
couplings in the evolution laws (8), mainly for the variable c and D. Hence, the elasto-viscoplastic materials are characterized
p
by an elastic domain in the stress space where yielding doesn’t occur (e_ ¼ 0, p_ ¼ 0 and c_ ¼ 0, if Fðr; Bp ; Bc ; BD ; cÞ < 0. rp , u, k
and n are material parameters that depend on the temperature. If D ¼ 0 and the study is restricted to isothermal processes,
the expressions proposed for the free energy w and the yielding function F in this paper coincide with the classic elasto-
viscoplasticity model with nonlinear kinematic and isotropic hardening presented in Lemaitre and Chaboche (1990).
In the proposed theory, the elastic region can be represented as a sphere with radius R
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
R¼ 2=3½Bp þ ð1  DÞrp  ¼ ð1  DÞf 2=3½bð1  edp Þ þ rp g ð11Þ
centered at the point X
X ¼ Bc ¼ ð1  DÞac ð12Þ

in the space of the principal directions of the deviatoric stress tensor. Such a geometric interpretation with can be found in
Lemaitre and Chaboche (1990), in the case of elasto-viscoplasticity with nonlinear kinematic and isotropic hardening. The
main difference here is the way the damage variable also affects the elastic domain. The internal forces Bc and Bp (and, hence
X and R) are affected not only by the internal variables associated to the irreversible deformations (p and ep ) but also by the
damage variable D.
The damage and plasticity can be seen as competing mechanismspffiffiffiffiffiffiffiffi - the radius R of the elastic region can get bigger due to
the plastic deformation (hardening due to plasticity, the term f 2=3½bð1  edp Þ þ rp g in (11) is an increasing positive quan-
tity) but also can get smaller as the damage increases (what describes the softening behavior caused by damage). Initially,
the hardening induced by the plastic deformation is more important and, as the damage increases, the softening occurs until
the ‘‘collapse” of the elastic region (radius R ¼ 0 when D ¼ 1). The damage also affects the kinematic hardening ðX ¼ 0 when
D ¼ 1). The range of validity of the model may be enlarged by introducing a second kinematic variable (Lemaitre and
Chaboche, 1990; Chaboche and Rousselier, 1983)
1 2
c ¼ c1 þ c2 ) Bc ¼ Bc þ Bc ð13Þ

and the center of the elastic domain can then be written as

X ¼ X1 þ X2 ð14Þ

The main idea in this theory is to characterize the material behavior through the free energy and the rate of energy dis-
sipation d. Consequently, additional constitutive information about the thermal dissipation must be given. In this theory this
information is obtained through the following constitutive relation between q and g:

q ¼ K gradðhÞ ð15Þ

K is a positive scalar function of the absolute temperature h. Eq. (14) is a generalized version of the Fourier law with K
being the thermal conductivity. It can be reduced to the usual Fourier law (K = constant) if small deviations of temperature
from a fixed ho is assumed.
The state laws (7), the evolution laws (8) and the Fourier law (14) form a complete set of constitutive equations. Intro-
ducing these constitutive equations into the local form (3) of the free energy, it is possible to obtain the so-called heat
equation:

div ðKgradðhÞÞ  qCh_ þ qr ¼ d1  acpT ð16Þ

 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436 431

acpT is called the thermal coupling and C is the specific heat, given by the following expressions:
" #
ow o/ ow o/ ow o/ ow o/ p
d1 ¼ q : þ þ : þ q ¼ r : e_ þ Bp p_ þ Bc : c_ þ BD D_ ð17Þ
oep or op oBp oc oBc oD oBD
! !
or p oBp oBc oBD _ o2 w p o2 w o2 w o2 w o2 w _
acpT ¼ h : ðe_  e_ Þ  p_  : c_  D ¼ qh : ðe_  e_ Þ þ p : e_ p þ p_ þ : c_ þ D ð18Þ
oh oh oh oh oeoh oe oh opoh ocoh oDoh
!
os o2 w h o2 W e o2 W p o2 W c
C ¼ h ¼  2 h ¼ ð1  DÞð Þ þ þ þ p1 þ p2 h ð19Þ
oh oh q oh2 oh2 oh2

It can be shown that the set of constitutive equations formed by (7), (8), and (15) will always verify the SLT inequalities
(9), provided all material parameters are positive functions of the absolute temperature. Since d1 P 0, it has a role in the heat
Eq. (16) similar to a heat source in the classical heat equation for rigid bodies. The term acpT in the right hand side of the heat
equation, can be positive or negative. It is important to remark that the strong coupling between the damage and other
mechanisms shown in (4) is necessary not only to and adequate description of the strain-softening phenomenon, but also
of the specific heat. From (15) it can be verified that the specific heat C decreases as D ! 1. It can also be shown that the
thermal coupling acpT goes to zero as D ! 1. Hence, the temperature oscillations are reduced when the damage is high
(since acpT ! 0), although it increases due to the term d1 P 0. If, in the model, the damage is supposed to affect the parcel
W h of the free energy ðW h ðD; hÞ ¼ ð1  DÞf ðhÞÞ, then the specific heat will decrease even more (C ? 0 as D ! 1 if the above
expression is adopted), what eventually is not physically reasonable.

4. Elasto-viscoplastic bars

In order to better understand the features of the model and the role of the thermomechanical couplings, a very simple
problem is considered: a straight bar immersed in a medium with constant temperature and subjected to prescribed axial
displacements at the extremities is considered in this section. This simple geometry offers a good enlightenment in the anal-
ysis of the thermo-coupling effects in fatigue life of metallic components and permits a direct comparison with experimental
results obtained from specimens of traditional low-cycle fatigue testing. For this one-dimensional case (x denotes the axial
direction), the set of Eqs. (7) and (8) with two kinematic hardening variables can be reduced to the following set of equations
(see appendix).
  
F d
¼ ð1  DÞE  ep  aðh  h0 Þ ð20Þ
Ao Lo
 
Y ¼ ð1  DÞ bð1  edp Þ þ rp ð21Þ
1 1 2 2 1 2 1 2
X ¼ ð1  DÞða1 c Þ; X ¼ ð1  DÞða2 c Þ; X ¼ X þ X and c ¼ c þ c ð22Þ
d d
BD ¼ E½1=2ð  ep Þ  aðh  h0 Þð  ep Þ þ bðp þ ð1=dÞedp Þ þ ð3=4Þ½a1 ðc1 Þ2 þ a2 ðc2 Þ2 
Lo Lo
Z h
p
 q½ p1 logðnÞdn þ 2 ðh  h0 Þ2  ð23Þ
h0 2

n
j r x  Xj  Y rx  X
e_ p ¼ ð24Þ
k jrx  Xj
p_ ¼ je_ p j ð25Þ
c_ 1 ¼ ð3=2Þe_ p  ðu1 =a1 ÞX 1 p;_ c_ 2 ¼ ð3=2Þe_ p  ðu2 =a2 ÞX 2 p_ ð26Þ
D
B
D_ ¼ p_ ð27Þ
S0
When there is no internal source present and the traditional hypothesis adopted in the study of fins are considered, the
heat Eq. (15) can be written in the alternative form (27), where X and Y are auxiliary variables, respectively related with the
kinematic and isotropic hardening. h0 is the initial temperature of the bar which is the same as the environment temperature
h1 . In the following analysis a linear dependency on temperature is considered for the material parameters k, n, rp , u1 , u2 ,
a1 , a2 , S0 , E, a, b, d, K, q. The constants h, P and Ao are, respectively, the convection coefficient, the perimeter and the area of
the cross section of the bar.
!
o oh hP or oY 2 oX oBD _
ðK Þ  ðh  h1 Þ  qCh_ ¼ re_ p þ Y p_ þ X c_  BD D_  h ðe_  e_ p Þ þ p_ þ c_  D ð28Þ
ox ox Ao oh oh 3 oh oh
     "  2  
h oE oa oa h od 1 1
C ¼ ð1  DÞ 2e ðh  h0 Þ þ a þ E  ð1  DÞ b pþ þ
q oh oh oh q oh d dð1 þ pdÞ

ob od ð1=d þ pÞ dp
2 e þ p1 þ p2 h ð29Þ
oh oh d
432 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436

The right side of Eq. (27) contains the thermomechanical coupling terms. In the analysis of problems with prescribed dis-
placement (u) loadings, the equilibrium equation and the strain-displacement relation must also be considered:
or ou r
¼ 0; ¼e¼ þ ep þ aðh  h0 Þ ð30Þ
ox ox ð1  DÞE
with the following boundary conditions: uðx ¼ 0Þ ¼ 0; uðx ¼ Lo Þ ¼ uL ðtÞ, where Lo is the length of the bar.

5. Numerical simulations

To study the thermo coupling effects in fatigue life of metallic components, the following analysis consider AISI 316L stain-
less steel rods with 5 mm diameter and 50 mm length submitted to prescribed cyclic displacement loadings (triangular shape).
The bar has an initial temperature of 293 K (h0 ) and is immersed in a medium with a constant temperature of 293 K (h1 ). The
material parameters for two distinct temperatures (Lemaitre and Chaboche, 1990; Peckner and Bernstein, 1977) are: for 293 K
(20 °C)  E= 196 GPa, rp = 82 MPa, k = 151 MPa, n = 24, b = 60 MPa, d = 8, a1 = 108.3 GPa, u1 = 2800, a2 = 4.5 GPa, u2 = 25,
a = 15.4  106/K, C = 454 J/Kg K, K= 13 W/m K. For 873 K (600 °C)  E =150 GPa, rp = 6 MPa, k = 150 MPa, n = 12, b=
80 MPa, d = 10, a1 = 17.5 GPa, u1 = 350, a2 = 1.0 GPa, u2 = 15, a= 18.0  106/K, C = 584 J/Kg K, K = 21 W/m K. A linear depen-
dence on temperature is adopted in the analysis. The coefficient S0 presents a dependency with plastic deformation amplitude,
p
and can be adequately represented by the following equation (Pacheco and Costa-Mattos, 1997) – S0 ¼ kf enf e with
p p 11
e ¼ maxt je ðtÞj - where kf and nf are material parameters. For AISI 316L, kf = 1.524  10 Pa and nf = 64.4. Fig. 1 shows a
comparison between experimental data from a low-cycle fatigue test and model fatigue life prediction for the 316L stainless
steel at room temperature. In the simulations, a combination of slow loadings rates and high convection coefficients is used
(T = 100 s and h = 10 kW/m2 K, where T is the loading period) to guarantee the temperature variation restriction of ±2 K of
the ASTM standard. This figure shows the good agreement between experimental data and predictions. For ordinary mechan-
ical components in operation this is a difficult condition to achieve and wrong predictions can be obtained if the thermome-
chanical coupling is not considered. As an example, consider a bar submitted to prescribed displacement loading amplitude
of ±0.75 mm at one end (±1.5% mean strain) and with constant temperature boundary conditions at both ends. To study
the influence of loading frequency and heat removal characteristics in the low-cycle fatigue life predictions, six thermome-
chanical loading conditions with different loading periods (T) and heat convection coefficients (h) are considered and are
shown in Table 1.
By symmetry considerations, only one half of the bar (0 6 x 6 L/2) is studied with the following boundary conditions: at
x = 0, u = 0 and h = hi ; at x = L/2, u = u(t) – triangular loading with an amplitude of 0.375 mm – and q = 0.
Fig. 2 shows temperature and damage evolution as a function of the number of loading cycles (N) at the bar midpoint
(x = L/2) where a localization phenomenon is observed. Situation (1) can be regarded as a testing of a specimen in compliance
with the ASTM standard for low-cycle fatigue, where the variation of temperature does not exceed the ±2 K range. In the

Fig. 1. e  N curve for the stainless steel 316L (293 K) obtained with the ASTM recommendations for low-cycle fatigue test.

Table 1
Thermomechanical loading conditions studied.

Condition T (s) h (W/m2 K)

(1) 100 10,000
(2) 100 1000
(3) 100 100
(4) 100 10
(5) 10 100
(6) 10 10
 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436 433

Fig. 2. Temperature (a) and damage (b) evolution at the bar midpoint for the six conditions considered.

other situations, this requirement is not obeyed and the temperature rise is enough to decrease the fatigue resistance of the
material. The predicted lifetimes using conditions (5) and (6) are, respectively, 79% and 74% of the lifetime considering con-
dition (1), which indicates a strong influence of the loading frequency in the amount of heat generated and, as a consequence,
in the life of the bar. The thermomechanical coupling can be seen as a feedback phenomenon. The heat generated by the
mechanical process causes an increase of temperature, which promotes a decrease in the mechanical strength. Therefore,
the plastic strain amplitude tends to increase causing a greater temperature rise and so on. It is also important to observe
that the temperature boundary conditions can lead to a localization process which can accelerate the feedback phenomenon.
In the following analysis, condition (6), which presents the more dramatic life prediction reduction, is explored in more
detail. Fig. 3 shows the stress and plastic strain evolution at the bar midpoint (x = L/2). From the stress curve it can be ob-
served the initial hardening and the subsequent softening behavior induced by the damage. Fig. 4 shows the localization of
the temperature and damage variable at the bar midpoint.

Fig. 3. Stress (a) and plastic strain (b) evolution at bar midpoint for condition (6).

Fig. 4. Temperature (a) and damage (b) evolution at different points for condition (6).
434 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436

Fig. 5. Temperature (a), damage (b) and accumulated plastic strain (c) distribution for condition (6).

Fig. 5 shows the distribution of the temperature, damage and accumulated plastic strain at different cycles. It is worth to
note that the mechanic behavior strongly depends on the temperature distribution. Other mechanical and thermal condi-
tions lead to different predictions. For example, adiabatic temperature boundary conditions at both ends lead to a homoge-
neous temperature distribution and different life predictions are observed for the six defined conditions. In this case a
reduction of 3% and 6% is predicted for conditions (4) and (6), respectively. In real life structures and mechanical compo-
nents, thermal boundary conditions between adiabatic and constant temperature can be expected. In tensile tests, the per-
turbation in the temperature field caused by the end connections geometry is sufficient to strongly localize deformation in
the middle of the bar, even if adiabatic conditions are considered.

6. Conclusion

In this paper, an internal variable theory was proposed to study the thermomechanical coupling effects in elasto-visco-
plastic materials subjected to cyclic inelastic deformations. The preliminary simulations show that it is important to consider
the thermomechanical coupling effects in low-cycle fatigue projects of mechanical components, particularly when the load-
ing frequency is high. In these situations, if the thermomechanical effect is not included in the model, wrong predictions can
be obtained and unexpected failures may occur. Eventually, when damage is not very high, the controlling phenomenon is
convection and temperature does not increase too much even if heat is generated due to the thermomechanical couplings.
Nevertheless, the fatigue damage and plastic deformations are strongly localized at the middle of the specimen. In this case
w, the specific heat can decrease very fast as the damage variable tends to its limit value, causing a localized increase of the
temperature even if the amount of heat generated is not so high and, hence, promoting a reduction in lifetime.
Further, experimental studies must be performed in order to fully understand the physical mechanisms involved. Most of
the experimental studies so far are restricted to coupled thermo-elasticity and do not account for the effect of inelastic defor-
mations. Infrared-thermography methods are promising tools for experimental studies of such kind of thermomechanical
couplings (Pastor et al., 2008 for instance). The way the free energy is affected by the damage variable defines how it will
affect the specific heat C (see Eqs. (4) and (15)), and it is a key to propose physically realistic expressions for w. In classical
 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436 435

heat conduction, generally conceived for rigid bodies, the Biot modulus (Bi) is used to define if a lumped-parameter analysis
is reasonable or not. The Biot modulus is defined as Bi ¼ ðhVÞ=ðSKÞ, where S is the surface area of the body and V its volume.
Bi can be seen as the ratio of the conductive (internal) resistance to heat transfer to the convective (external) resistance to
heat transfer. It is usually considered that, if Bi 6 0:1, a lumped-parameter analysis will introduce an error in the tempera-
ture smaller than 5%. Hence, it is usually supposed that a small value of Bi implies that the temperature gradients within the
system are relatively small. This is true for rigid bodies but surely not in the case of damageable inelastic materials under-
going cyclic deformations. In the case of tensile specimens in low-cycle fatigue tests there is a strong localization of temper-
ature at the center and the gradients can be very high even if the Biot modulus is small.
Finally it is important to remark that an adequate modeling of the inelastic strain (and damage) localization is very
important in order to perform a realistic prediction of the thermomechanical couplings. Eventually, for a better description
of the localization phenomenon, a gradient-enhanced damage model (such as in Costa-Mattos et al., 1992 or Costa-Mattos
and Chimisso, 1997), based on a microstructure theory, should be used. However, the proposed model is reasonably suitable
to a first step towards a rational methodology to understand the adequate form of free energy and specific heat in coupled
thermo-elasto-viscoplasticity with isotropic damage.

Appendix A. One-dimensional constitutive equations

The goal of this appendix is to summarize the procedure to obtain the one-dimensional constitutive equations (19)–(26)
from the general state laws and evolution laws (7), (8). From (4)–(7) it is possible to obtain the following general state laws
 
E m
r ¼ ð1  DÞ ðe  ep Þ þ ½trðe  ep Þ1 ðA:1Þ
2ð1 þ mÞ ð1  2mÞ
Bp ¼ ð1  DÞ½bð1  edp Þ ðA:2Þ
ci i
B ¼ ð1  DÞ½ai c  ði ¼ 1; 2Þ ðA:3Þ
D p
B ¼ ½W e ðe  e ; hÞ þ W p ðp; hÞ þ W c ðc; hÞ ðA:4Þ
o oW h ðhÞ
s ¼ ð1  DÞ ½W e ðe  ep ; hÞ þ W p ðp; hÞ þ W c ðc; hÞ þ ðA:5Þ
oh oh
and from (8–10) we have the following general evolution laws
ðr þ Bc Þdev
e_ p ¼ ð3=2Þhf =kin ðA:6Þ
Jðr þ Bc Þ
p p
p_ ¼ hf =kin ¼ ½ð2=3Þðe_ : e_ Þ1=2 ðA:7Þ
i p i
c_ ¼ e_ þ ðui =ai ÞðBc pÞði
_ ¼ 1; 2Þ ðA:8Þ
D
D_ ¼ ðB =S0 Þp_ ðA:9Þ
For a round elasto-plastic specimen with gauge length Lo and cross section Ao submitted to a one-dimensional loading
such that a push-pull low cycle fatigue test, the tensors r and e are supposed to be given by the following expressions
0 1 0 1
rðtÞ ¼ ðFðtÞ=Ao Þ eðx; tÞ ¼ ðdðtÞÞ=Lo
r¼B
@ 0
C
A e¼B
@ m eðx; tÞ
C
A
0 m eðx; tÞ

where FðtÞ is the external force and dðtÞ the prescribed elongation. m is a constant such that 0 < m < 0:5. Considering the
i i
material isotropy and that trðep ¼ trðBc Þ ¼ trðci Þ ¼ 0, the variables ðrÞdev ; ep ; Bc and ci can be expressed by
0 1 0 1
2
3
rðtÞ ep ðx; tÞ
B rðtÞ C B ep ðx;tÞ C
ð rÞ ¼ B
@  3
C;
A ep ¼@  2 A
ep ðx;tÞ
 r3ðtÞ  2
0 1 02 i
1
2
3
rðtÞ 3
c ðx; tÞ
B C B i C
 r3ðtÞ
i
 Bc ¼ B
@
C;
A ci ¼ B
@  c ðx;tÞ
3
C
A
rðtÞ i
3  c ðx;tÞ
3

i
where X i and ci are auxiliary variables. It is simple to verify that the components of the tensor variables ðrÞdev ; ep ; Bc and ci
are not independent, in this case. From the definition of the von Mises equivalent stress, it is possible to conclude that
Jðr þ Bc Þ ¼ jr  X 1  X 2 j. Hence, using (10.2), we have: f ¼ jr  X 1  X 2 j  Y, with Y being the auxiliary variable defined in
(20). Finally, using the above results and (A.6), it is possible to obtain Eq. (23):
436 H. da Costa-Mattos, P.M.C.L. Pacheco / Mechanics Research Communications 36 (2009) 428–436

* +n
jr  X 1  X 2 j  Y r  X1  X2
e_ p ¼
k jr  X 1  X 2 j

The deduction of the other one-dimensional constitutive equations (20)–(23) and (25)–(27) is reasonably simple using
the expressions presented at this appendix.

References

ASTM E606-80, 1980, Standard recommended practice for constant-amplitude low-cycle fatigue testing. In: ASTM Standards, vol. 03.01, pp. 629–641.
Chaboche, J.L., Rousselier, G., 1983. On the plastic and viscoplastic constitutive equations – Part I and II. J. Pressure Vessel Technol. 105, 153–164.
Costa-Mattos, H.S., Chimisso, F.E.G., 1997. Necking of elasto-plastic rods under tension. Int. J. Non-Linear Mech. 33 (6), 1077–1086.
Costa-Mattos, H.S., Mamyia, E.N., Fremond, M., 1992. A simple model of the mechanical behavior of ceramic-like materials. Int. J. Solid Struct. 20 (24), 3185–
3200.
Lemaitre, J., Chaboche, J.L., 1990. Mechanics of Solids Materials. Cambridge University Press.
Longère, P., Dragon, A., 2008. Plastic work induced heating evaluation under dynamic conditions: critical assessment. Mech. Res. Commun. 35, 135–141.
Pacheco, P.M.C.L., Costa-Mattos, H.S., 1997. Modeling the thermomechanical coupling effects on low-cycle fatigue of metallic materials. In: 5th ICBMFF, 5th
International Conference on Biaxial/Multiaxial Fatigue and Fracture, Cracow, Poland, pp. 291–301.
Pastor, M.L., Balandraud, X., Grédiac, M., Robert, J.L., 2008. Applying infrared thermography to study the heating of 2024-T3 aluminium specimens under
fatigue loading. Infrared Phys. Technol. 51 (6), 505–515.
Peckner, D., Bernstein, I.M., 1977. Handbook of Stainless Steels. McGraw-Hill.
Rosakis, P., Rosakis, A.J., Ravichandran, G., Hodowany, J., 2000. A thermodynamic internal variable model for the partition of plastic work into heat and
stored energy in metals. J. Mech. Phys. Solid. 48, 581–607.
Simo, J.C., Miehe, C., 1992. On the coupled thermomechanical treatment of necking problems via finite element methods. J. Appl. Meth. Eng. 33, 869–883.
Stabler, J., Baker, G., 2000. On the form of free energy and specific heat in coupled thermo-elasticity with isotropic damage. Int. J. Solid Struct. 37 (34), 4691–
4713.