Composite Structures 94 (2012) 3043–3051

Contents lists available at SciVerse ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Bending and creep buckling response of viscoelastic functionally graded

beam-columns
Ehab Hamed ⇑
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, The University of New South Wales, UNSW Sydney, NSW 2052, Australia

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

Article history: The viscoelastic creep response of flexural beams and beam-columns made with functionally graded
Available online 8 May 2012 materials is numerically investigated. The paper highlights the challenges associated with the modeling
and analysis of such structures, and presents a nonlinear theoretical model for their bending and creep
Keywords: buckling analysis. The model accounts for the viscoelasticity of the materials using differential-type
Beam constitutive relations that are based on the linear Boltzmann’s principle of superposition. The model is
Buckling general in terms of its ability to deal with any material volume faction distribution through the depth
Column
of the beam, and with different linear viscoelastic laws, boundary conditions, and loading schemes.
Creep
Functionally graded materials
The governing equations are solved through time stepping numerical integration, which yields an expo-
Viscoelasticity nential algorithm following the expansion of the relaxation function into a Dirichlet series. A numerical
study that examines the capabilities of the model and quantifies the creep response of functionally
graded beam-columns is presented, with special focus on the stresses and strains redistribution over time
and on the creep buckling response. The results show that the creep response of such structures can be
strongly nonlinear due to the variation of the viscoelastic properties through the depth, along with
unique phenomena that are not observed in homogenous structures.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction viscoelastic materials that creep at ambient temperatures are used

The use of functionally graded materials (FGMs) in a broad
range of engineering applications is rapidly increasing with many
aspects of this concept being researched worldwide. In structural
applications, the FGM concept is attractive mainly due to the
advantages of such structures over layered sandwich-type struc-
tures in terms of eliminating interface problems and mitigating
stress concentrations. Applications have recently been reported
for sensors, armors, dental implant, and rotating blades [1–3],
and the concept has been recently examined even for rigid con-
crete pavement systems [4]. Sandwich panels made from function-
ally graded core materials, as well as composite laminates and
cement composites with variable fibers volume fraction have also
been investigated [5–7].
Many research studies with regards to functionally graded
structures (FGSs) were conducted with focus on vibration of
beams, buckling, fracture behavior, and nonlinear behavior of
plates and shells [2,3,8–13]. However, the viscoelastic response
of FGSs, as well as the development of analytical models for their
analysis, has hitherto not been fully investigated. In many cases,
⇑ Tel.: +61 2 93859765; fax: +61 2 93859747.
E-mail address: e.hamed@unsw.edu.au
in a FGS like polymers and cementitious materials; and in other
cases, applications are made for structures that are exposed to high
temperatures, which initiate creep in many metals and other
materials. This paper focuses on the viscoelastic creep behavior
of flexural beams made of functionally graded materials, including
the effects of geometric nonlinearities where creep buckling effects
may dominate the structural response (beam-columns).
The viscoelastic creep behavior of beams and beam-columns
made of FGMs is associated with some unique physical phenomena
that result from the gradual variation of the material volume frac-
tion and the corresponding variation of the viscoelastic character-
istics through the thickness of the member. Although the gradual
structure eliminates interfacial stress concentrations which may
cause debonding failures over time in layered beams [14–16],
creep may cause distortions and rotations of the cross-section, as
well as, gradual variations of its flexural rigidities and stresses
distribution with time [6]. In addition, because of the potential
extreme nonlinearity in typical creep constitutive relations (e.g.,
power law creep with different powers) [2], varying the micro-
structure from one material to another leads to variations in the
viscoelastic properties that may not necessarily follow the same
path of variation of the materials volume fraction or the elastic
properties (e.g., modulus of elasticity or Poisson’s ratio). Thus,
the flexural viscoelastic response of functionally graded beams

0263-8223/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2012.04.029
3044 E. Hamed / Composite Structures 94 (2012) 3043–3051
and its structural modeling need to be clarified to better utilize the
advantages of such structures.
Under axial compression loading (columns or beam-columns),
the structural response can be dominated by buckling of the mem-
ber, which becomes more critical when the load is constant with
time. In such cases, the deformations caused by creep, as well as
the consequent increase or modification of the internal stresses
may lead to loss of stability under a sustained load that is signifi-
cantly smaller than the elastic buckling load (so called ‘‘creep buck-
ling’’). Alternatively, the creep deformations may not necessarily
lead to buckling failure but they may increase the internal stresses
and decrease the residual strength of the member when additional
loads are to be applied. The dependence of the creep strains on the
level of stresses, which vary with time, along with the structural
sensitivity to geometric imperfections that cannot be avoided in
practice, make accurate prediction of the nonlinear behavior of
functionally graded beam-columns a challenging and difficult task.
Few research studies have focused on the viscoelastic behavior
of FGSs. Paulino and Jin [17,18] and Mukherjee and Paulino [19]
showed that the elastic–viscoelastic correspondence principle can
be used for the analysis of FGSs. Nevertheless, this principle was
implemented only for simple structures with single relaxation
time, and where a closed form solution of the corresponding elastic
problem exists. However, the application of the correspondence
principle for materials with multiple relaxation times (required
to fit test data), or for cases where a closed form solution of the
elastic problem is hard to be achieved, can be problematic. Further
limits of the viscoelastic correspondence principle that include
restrictions on the material viscoelastic characteristics and on the
boundary conditions were discussed by Jin [20] and Paulino and
Jin [17], which reveal the need for general modeling and solution
approaches. Hilton [6] and Hilton et al. [21] studied the material
modeling of viscoelastic FGMs including the development of an
optimization procedure for their design; yet, with no specific refer-
ence to the structural modeling and analysis. Pan et al. [22] used
the finite element approach to investigate the fracture and crack-
ing behavior of viscoelastic FGMs, and Khan and Muliana [23] stud-
ied the thermo-viscoelastic material characterization of FGMs
where the continuous variation of the thermal and mechanical
properties is approximated as an assembly of homogeneous layers.
Chen et al. [24] presented a section analysis of a beam made with
viscoelastic FGMs; yet without generalization of the model for dif-
ferent creep laws, and without consideration of the geometric non-
linearity. Hilton [25] investigated the failure probability of
viscoelastic functionally graded columns under dynamic loading,
but without a detailed description of the variation of stresses and
deformations with time under sustained loading. Other research
studies have focused on the creep buckling response of function-
ally graded shells [26].
This paper aims to provide insight into the viscoelastic flexural
and creep buckling behavior of beam-columns made of function-
ally graded materials through the development of a general theo-
retical model and numerical tools for their viscoelastic analysis.
The characterization of the material properties is based on the ba-
sic assumption that the microscopic length scale is much larger
than the molecular scale, but at the same time, much smaller than
the characteristic length of the macroscopic sample [1]. So that the
heterogeneous FGM can be treated as a continuum in the micro-
scopic scale, while macroscopic effective properties can be used
for each material point through the thickness, along with assumed
smoothing variation in the gradation direction. In the microscopic
level, the classical role of mixture is used here for the evaluation of
the effective material properties, while the lamination theory is
adopted for the macroscopic level. In this sense, the model consid-
ers the functionally graded member as a composite member with
gradual variation of the volume fraction of its constituents through
the depth. Creep is accounted for through Boltzmann’s principle of
superposition [27], which yields a generalized Maxwell rheological
model after the expansion of the relaxation function into a Dirich-
let series. The model accounts for the geometric nonlinearity
through large deformation kinematic relations. In addition, the pa-
per focuses on gradation of the materials through the depth of the
section only. Other assumptions made in the creep modeling and
analysis are included in the mathematical formulation.
2. Mathematical formulation
The model uses the lamination theory with first order shear
deformations based on Timoshenko’s beam theory for the model-
ing of the functionally graded beam-column (FGBC). Despite the
potential distortion of the section that may result from the gradual
variation of the mechanical properties, it was shown by Sankar
[28] that beam models appear to agree well with two dimensional
elasticity solutions of slender FGBs under transverse loading. In or-
der to account for the gradual variation of stresses and strains with
time, an incremental time-stepping analysis is conducted. For this,
the time of concern t is subdivided into nt discrete times with
Dtr = tr – tr1 (r = 1, 2, . . . , nt). The incremental nonlinear kinematic
relations for the FGBC at time tr follow the assumption of large dis-
placements, with moderate rotations and small strains as follows:
1
Dexx ðx; z; t r Þ ¼ Du;x ðx; t r Þ þ ðDw;x ðx; tr ÞÞ2
2
þ w;x ðx; t r1 ÞDw;x ðx; t r Þ  zD/;x ðx; t r Þ ð1Þ
Dcxz ðx; t r Þ ¼ Dw;x ðx; t r Þ  D/ðx; t r Þ ð2Þ
where exx and cxz are the in-plane normal and engineering shear
strains respectively; w, u and / are the vertical displacement, in-
plane displacement at the reference line (mid-depth), and rotation
of the cross section, respectively; D is the incremental time operator;
and (),x denotes a derivative with respect to x. The sign conventions
for the coordinates, deformations, and loads of a FGBC are shown in
Fig. 1. The constitutive relations and the characterization of the vis-
coelastic properties in the material point level are discussed next,
followed by the structural modeling that includes the formulation
of the governing equations and their solution procedure.
2.1. Constitutive relations
Using Boltzmann’s principle of superposition [27], the time-
dependent stress–strain relations of each point through the depth
of the member take the following history integral form:
Z t
dexx ðx; z; t 0 Þ 0
rxx ðx; z; tÞ ¼ Rxx ðz; t  t0 Þ dt ;
to dt 0
Z t
dcxz ðx; t 0 Þ 0
sxz ðx; z; tÞ ¼ Rxz ðz; t  t0 Þ dt ð3Þ
to dt 0
where rxx and sxz are the in-plane normal and shear stresses respec-
tively; Rxx(z, t  t0 ) and Rxz(z, t  t0 ) are the normal and shear relax-
ation functions, which represent the stress at time t caused by a unit
constant strain imposed at time t0 . The thermo-mechanical material
behavior of the structural components is not accounted for, in order
to clarify the creep effects.
x,u mx qz nx h/2
z,w h/2
L
Fig. 1. Geometry, loads, and sign conventions.
 E. Hamed / Composite Structures 94 (2012) 3043–3051 3045

The creep laws that appear in Eq. (3) require the storage of the This formulation enables the description of the viscoelastic behav-
complete strain histories and consequently huge computational ior under variable loading history and accounts for the variation
resources that are impractical for large structures or structures with of stresses and geometry with time through a step-by-step analysis.
many degrees of freedom. Therefore, a differential-type form of However, to be able to implement it into the structural model, char-
Boltzmann’s principle, which was presented and further developed acterization of the z-dependent spring moduli is required, which is
in detail in Zienkiewicz and Watson [29], Taylor et al. [30], and discussed in the subsequent.
Bazant and Wu [31], is adopted with further enhancement to the The relaxation times (which correspond to gl(z)/el(z) and fl(z)/
case of FGMs. For this, the relaxation functions need to be expanded Gl(z) with gl and fl as the dashpot constants) are chosen in ad-
into Dirichlet series, which then yield a rheological material model vance considering the time of interest. A suitable choice is given
that corresponds to the generalized viscoelastic Maxwell model by [31]:
(one in the normal direction and one in shear) as follows [29–31]:
T nl ðzÞ ¼ T n1 ðzÞal1 ; T sl ðzÞ ¼ T s1 ðzÞbl1 ð13Þ
X
N
n
 xx ðz; tÞ ¼
Rxx ðz; tÞ ffi R El ðzÞet=T l ðzÞ þ ENþ1 ðzÞ ð4aÞ where T n1 ðzÞ; T s1 ðzÞ,
a and b are parameters that are chosen to cover
l¼1 the entire time of concern. Once these parameters are chosen, the
moduli of the springs can be determined by the least squares or
X
S
s other curve fitting method to fit a test data or known expressions
 xz ðz; tÞ ¼
Rxz ðz; tÞ ffi R Gl ðzÞet=T l ðzÞ þ GSþ1 ðzÞ ð4bÞ
l¼1
of Rxx(z, t) and Rxz(z, t) at each point through the depth (any given
z). The least squares method proposed in [31] is adopted here,
where Rxx ðz; tÞ and Rxz ðz; tÞ are the approximated relaxation func- which is based on minimization (dU/dEu = 0) of the following
tions, El(z) and Gl(z) are the moduli of the lth Hookean spring in expression:
the generalized Maxwell chain for the modeling under normal
and shear tractions respectively, N and S are the number of units X X
N1
U¼  xx ðt i Þ  Rxx ðti Þ2 þ x1 ðElþ1  El Þ2
½R
in each action, and T nl ðzÞ and T sl ðzÞ are the relaxation times. i l¼1
Using Eqs. (4a) and (4b), the incremental model presented in
X
N2
Bazant and Wu [31] can actually be applied for the viscoelastic þ x2 ðElþ2  2Elþ1 þ El Þ2
creep analysis of FGBCs but with z-dependent spring moduli. For l¼1
clarity and completeness, the incremental formulation is repre- X
N3
sented here, which yields the following incremental constitutive þ x3 ðElþ3  3Elþ2 þ 3Elþ1  El Þ2 ð14Þ
relations: l¼1
Drxx ðx; z; t r Þ where x1, x2 and x3 are weight functions that are determined to
Dexx ðx; z; t r Þ ¼ þ De00 ðx; z; t r Þ ð5Þ
E00 ðz; t r Þ achieve best fitting of the data. The same procedure is also used
for evaluating the spring moduli in shear.
Dsxz ðx; z; t r Þ However, as fitting the data at every z is impossible, a number
Dcxz ðx; t r Þ ¼ þ Dc00 ðx; t r Þ ð6Þ
G00 ðz; t r Þ of discrete points need to be selected through the depth of the
where E00 and G00 are the pseudo normal and shear moduli, and De00 beam. One may take advantage of the known or estimated distri-
and Dc00 are the incremental creep strains. Eqs. (5) and (6) are based bution path of the material volume fraction to make the calculation
on a numerical time integration of the generalized Maxwell model of the Maxwell chain moduli at two points only (at z = h/2 and
assuming a constant strain rate at every time interval, which yields z = h/2), and to obtain a continuous variation of the spring moduli
an exponential algorithm as follows: that follows the same path. However, this procedure can be inaccu-
XN   Tn rate in many cases because the algebraic system of equations from
n l
E00 ðz; t r Þ ¼ 1  eDtr =T l El ðzÞ þ ENþ1 ðzÞ ð7Þ which the spring moduli are calculated (minimization of Eq. (14))
l¼1
D t r
exhibits an ill-conditioned system in terms of sensitivity to the
input data [32]. When this sensitivity is combined with the nonlin-
XS   Ts earity associated with the different creep laws of the constituent
s l
G00 ðz; tr Þ ¼ 1  eDtr =T l Gl ðzÞ þ GSþ1 ðzÞ ð8Þ materials, the values of the spring constants that can be obtained
l¼1
Dt r
through direct calculation at the intermediate points can be differ-
N 
ent to those obtained from assuming a variation in accordance
1 X 
De00 ðx; z; tr Þ ¼
n
1  eDtr =T l rl ðx; z; tr1 Þ ð9Þ with the material volume fraction. Therefore, the method proposed
00
E ðz; t r Þ l¼1 here is based on calculation of the spring constants via minimiza-
tion of Eq. (14) at a number of selected points (between 5 and 10
1 XS 
s
 points) through the depth, along with an interpolation function
Dc00 ðx; tr Þ ¼ 00 1  eDtr =T l sl ðx; z; tr1 Þ ð10Þ for each spring modulus that passes through the selected points.
G ðz; t r Þ l¼1
Depending on the variation of the volume fraction, the smoothing
or interpolation function may take different forms. In general, the
The stresses at each unit of the Maxwell chain are determined as
following interpolation function can be used with a curve fitting
follows:
procedure for the determination of its constants:
n n
rl ðx; z; tr Þ ¼ eDtr =T l rl ðx; z; tr1 Þ þ ð1  eDtr =T l Þ El ðzÞ ¼ c1 z þ c2 z2 þ c3 z3 þ c4 z4 þ c5 ð15Þ
n
Tl The same procedure is also used for the determination of Gl(z). In
 El ðzÞDexx ðx; z; t r Þ ð11Þ
Dt r many cases, the procedure outlined above is necessary because
the relaxation function may be known only for specific combina-
s s
sl ðx; z; tr Þ ¼ eDtr =T l sal ðx; z; tr1 Þ þ ð1  eDtr =T l Þ tions of the constituent materials, bearing in mind that each point
through the depth includes different material constituents and dif-
T sl
 Gl ðzÞDcxz ðx; t r Þ ð12Þ ferent viscoelastic properties, which cannot be experimentally
Dt r characterized.
3046 E. Hamed / Composite Structures 94 (2012) 3043–3051

The procedure outlined in Eqs. (14) and (15) can be used for the 2.2. Structural modeling and analysis
characterization of the spring moduli provided that the relaxation
functions Rxx and Rxz are known for any combination of the constit- The incremental equilibrium equations of a beam-column with
uent material. However, as this is not always the case, the charac- geometric nonlinearity can be derived based on the variational
terization of the relaxation functions is discussed here. In general, principle of virtual work as follows:
test data is available for individual phase of the constituent mate- Z Z
rials of a FGS and not for the composite structure. This requires the ðrxx þ Drxx ÞdDexx dV þ ðsxz þ Dsxz ÞdDcxz dV
V V
use of micromechanical model to predict the relaxation function of Z x¼L
the composite structure at each point through the gradation direc-  ½ðqz þ Dqz ÞdDw þ ðnx þ Dnx ÞdDu
tion. Here, the classical role of mixture is adopted for its simplicity, x¼0

where the relaxation functions are given by the following expres- þ ðmx þ Dmx ÞdD/dx ¼ 0 ð22Þ
sion assuming two-phase materials for brevity and clarity:
where qz, nx, and mx are external distributed loads and bending mo-
Rxx ðz; tÞ ¼ V a ðzÞRaxx þ ð1  V a ðzÞÞRbxx ð16Þ ments, respectively (see Fig. 1). Concentrated loads and moments
can also be included in Eq. (22), but for brevity they are not pre-
Rxz ðz; tÞ ¼ V a ðzÞRaxz þ ð1  V a ðzÞÞRbxz ð17Þ sented here. Using Eqs. (1), (2) and (22), the following nonlinear
incremental equilibrium equations are obtained:
where and Raxx Rbxx
are the normal relaxation functions of the two
DNxx;x ¼ Dnx ð23Þ
constituent materials a and b, Raxz and Rbxz are the relaxation func-
tions in shear, Va(z) is the variable volume fraction of material a,
DV xx;x þ ððNxx þ DNxx ÞDw;x Þ;x þ ðDNxx w;x Þ;x ¼ Dqz ð24Þ
which can take different quadratic, exponential, power, or trigono-
metric shapes.
DMxx;x  DV xx ¼ Dmx ð25Þ
The relaxation functions of each phase material are generally
determined based on the compliance function (J) owing to the fact where Nxx, Vxx, and Mxx are the axial force, shear force, and bending
that creep tests are much easier to conduct than relaxation tests. moment respectively. Note that all quantities appear in Eqs. (23)–
This is achieved through the following convolution integrals [27]: (25) are function of x and t, which are omitted for brevity.
Z t
The constitutive relations in the cross-section level are deter-
J ixx ðt  nÞRixx ðnÞdn ¼ t for t > 0; mined using the classical definition of the stress resultants and
0 using Eqs. (5) and (6):
J ixx ð0ÞRixx ð0Þ ¼ 1 for t ¼ 0 ði ¼ a; bÞ ð18Þ Z h=2
DNxx ðx; tr Þ ¼ bDrxx ðx; z; tr Þdz
Z t h=2
J ixz ðt  nÞRixz ðnÞdn ¼t for t > 0; Z h=2
00
0 ¼ bE ðz; tr ÞðDexx ðx; z; tr Þ  De00 ðx; z; tr ÞÞdz ð26Þ
J ixz ð0ÞRixz ð0Þ ¼ 1 for t ¼ 0 ði ¼ a; bÞ ð19Þ h=2

Eqs. (18) and (19) are solved numerically or analytically by the aid Z h=2

of a Laplace transformation for each material phase. DMxx ðx; t r Þ ¼ bDrxx ðx; z; t r Þzdz
h=2
Depending on the materials involved (metals, plastics, poly- Z h=2
mers, etc.), the compliance function can take different and com- ¼
00
bE ðz; t r ÞðDexx ðx; z; t r Þ  De00 ðx; z; t r ÞÞzdz ð27Þ
bined forms of exponential laws, power laws, logarithmic laws, h=2
hyperbolic sin functions, and other forms [27]. For brevity and sim-
Z h=2
plicity, the empirical power function law, which is able to describe
DV xx ðx; t r Þ ¼ j bDsxz ðx; z; tr Þdz
creep in many materials, is demonstrated here. Nevertheless, note h=2
that the modeling concepts proposed in this paper can deal with Z h=2
00
any desired compliance function. The compliance functions ¼j bG ðz; t r ÞðDcxz ðx; t r Þ  Dc00 ðx; t r ÞÞdz ð28Þ
according to the power law take the following form: h=2

1 1 where b and h are the width and depth of the section respectively.
J ixx ðtÞ ¼ i
þ Sixx t mi ; J ixz ðtÞ ¼ i
þ Sixz tgi ði ¼ a; bÞ ð20Þ Substitution of the kinematic relations (Eqs. (1) and (2)) and the
E G
incremental creep strains (Eqs. (9) and (10)) into Eqs. (26)–(28)
where Ei and Gi (i = a, b) are the modulus of elasticity and the shear leads to the following constitutive relations:
modulus of the phase materials, Sixx , Sixz , mi and gi (i = a, b) are nor- 8 9 2 38 9 8 9
2
< DNxx > < DN c >
> > 1
mally determined to best fit a test data. Once the compliance func- = A11 B11 0< Du;x þ 2 ðDw;x Þ >= > =
6 7
tions of each phase material are known, the relaxation functions are DM xx ¼ 4 B11 D11 0 5 D/;x  DM c ð29Þ
>
: >
; >
: >
; :> >
;
calculated via Eqs. (18) and (19). For the power law that is given in DV xx 0 0 jA55 Dw;x  D/ DV c
Eq. (20), Neilsen [33] presented the following expressions obtained
from solving Eqs. (18) and (19): where A11 ; B11 ; D11 and A55 are the extensional, extensional-bending,
flexural, and shear viscoelastic rigidities of the functionally graded
X1
ð1Þk tkmi ðSixx Ei cð1 þ mi ÞÞk beam-column, j is the shear correction factor, and DNc, DMc and
Rixx ðtÞ ¼ Ei ;
k¼0
cð1 þ kmi Þ DVc are incremental effective forces that result from creep. The vis-
coelastic rigidities take the following form:
X1
ð1Þk tkg i ðSixz Gi cð1 þ g i ÞÞk
Rixz ðtÞ ¼ Gi ð21Þ Z h=2 Z h=2
k¼0
cð1 þ kg i Þ A11 ðt r Þ ¼
00
bE ðz; t r Þdz; B11 ðtr Þ ¼
00
bE ðz; t r Þz dz ð30; 31Þ
h=2 h=2
where C is the gamma function, and Eq. (21) is valid for i = a or b.
The relaxation functions obtained from Eq. (21) can be back substi- Z h=2 Z h=2
00 00
tuted into Eqs. (16) and (17) to obtain the relaxation function of the D11 ðt r Þ ¼ bE ðz; tr Þz2 dz; A55 ðt r Þ ¼ bG ðz; tr Þdz ð32; 33Þ
h=2 h=2
FGM at any point through the depth of the member.
 E. Hamed / Composite Structures 94 (2012) 3043–3051 3047

Note that the definition of the rigidities according to Eqs. (30–33) b xx þ Dmx  w; DNxx  Nxx Dw;
DMxx;x ¼ D V ð41Þ
x x
applies to an arbitrary FGM including continuous and stepwise lay-
ered materials. b xx;x ¼ Dq
DV ð42Þ
z
The incremental effective creep forces, which result from sub-
stitution of the incremental creep strains (Eqs. (9) and (10)) into DNxx;x ¼ Dnx ð43Þ
the second term of Eqs. (26)–(28), take the following form:
Z h=2 X
N Du;x ¼
DNc ðx; t r Þ ¼
n
b ð1  eDtr =T l Þrl ðx; z; t r1 Þdz ð34Þ  
h=2 B11 ðDMxx þ DMc Þ  D11 ðDNxx þ DNc Þ þ ðA11 D11  B211 Þ 12 ðDw;x Þ2  w;x Dw;x
l¼1

A11 D11  B211
Z h=2 X
N
n ð44Þ
DM c ðx; t r Þ ¼ b ð1  eDtr =T l Þrl ðx; z; t r1 Þz dz ð35Þ
h=2 l¼1 b xx ¼ DV xx þ w;x DNxx þ N xx Dw;x as the effective shear force.
where D V
The corresponding general boundary conditions are:
Z h=2 X
S
s
DV c ðx; t r Þ ¼ j b ð1  eDtr =T l Þsl ðx; z; t r1 Þdz ð36Þ fDNxx ¼ DN 
or Du ¼ Du ð45Þ
h=2 l¼1
fDM xx ¼ DM 
or D/ ¼ D/ ð46Þ
These effective creep forces depend on the response from the previ-
ous time step. Thus, they are known functions at the beginning of
b xx ¼ DP
fD V 
or Dw ¼ Dw ð47Þ
each time step and are assumed constants within it. The integrals
appear in Eqs. (30–36) are numerically evaluated using the Trape- where the over-bar designates prescribed forces and deformations
zoidal rule in order to account for any distribution of the material at the boundaries; f = 1 for x = L and f = 1 for x = 0.
constants and the resulted stresses and strains through the depth. Due to the variation of the equivalent viscoelastic rigidities with
Due to the variation of the pseudo shear modulus through the time (Eqs. 30–33), and the variation of the effective creep forces
depth of the beam, the shear correction factor appears in Eqs. with time and space (Eqs. (34)–(36)), Eqs. (39)–(47) exhibit a
(29) and (36) no longer equals to the well-known values reported different boundary value problem at each time increment that do
for homogeneous beams. For this, the shear correction factor is not have a closed form solution. Because of this, the general
evaluated based on the method proposed by Bert [34] for the elas- numerical multiple shooting method is used for their solution
tic analysis of non-homogeneous beams. The approach is based on [36], which is applicable for any combination of boundary condi-
comparing the elastic strain energy of the actual beam with that of tions and any distribution of the effective creep forces along the
an equivalent Timoshenko beam, and the shear corrector factor member. In general, the incremental procedure outlined above
reads: for the solution in time allows for increasing the time step interval
hR i2
h=2  Þdz throughout the analysis. However, in order to account for the rapid
h=2
ðA11 b  B11 a
j¼ ð37Þ increase of the deformations rate at certain points due to the geo-
Z   B11 a
A55 h=2
ðA11 b  Þ2 metric nonlinearity, to consider the fact that the rate of creep is
00 dz
b h=2 G normally higher for a short period of time after loading, and to ac-
 are defined as the partial extensional and exten-
 and b count for gradually increasing or decreasing loads, the time step is
where a
kept relatively small and almost constant through the entire anal-
sional-bending viscoelastic rigidities, which are given as follows:
Z Z ysis. Nevertheless, at certain points or time domains, a reduction of
z z
ðtr Þ ¼ 00  rÞ ¼ 00 the time step is required, and in other cases, an increase is allowed.
a bE ðz; t r Þdz; bðt bE ðz; tr Þz dz ð38Þ
h=2 h=2 In this paper, user intervention was used when necessary. The
appropriate time step for a given loading is selected so that the dif-
Note that the shear correction factor becomes time-dependent due
ference between the critical time that causes creep buckling based
to the dependent of the pseudo moduli on the time step interval
on the selected time step and one-half of this step is of minor sig-
(Eqs. (7) and (8). The procedure outlined through Eqs. (37) and
nificance. The analysis is conducted up to a certain time (the crit-
(38) is only an approximation because it applies an expression of
ical time) where the rate of deformation of the system begins to
the shear correction factor obtained from an elastic analysis to a vis-
increase rapidly towards infinity or when the deformations exceed
coelastic one; yet, it is adopted here for simplicity. In fact, the shear
prescribed limits [37,38]. The instantaneous response can be ob-
correction factor in a viscoelastic material no longer depends on the
tained by choosing a number of fairly small time steps between
material and geometric properties only, rather it also depends on
t = 109 and t = 108 h for example with initial conditions of
the stress and strain histories (see [35]). However, this effect is
rl(x, z, t0) = sl(x, z, t0) = 0.
not considered here.
The incremental governing equations are formulated by substi-
3. Numerical examples
tution of the incremental constitutive relations (Eq. (29)), into the
equilibrium equations (Eqs. (23)–(25)) noting that terms of higher
Two numerical examples are presented. The first example
order product of the incremental displacements and forces may be
investigates the viscoelastic bending behavior of a simply sup-
neglected due to the use of sufficiently small increments in time.
ported functionally graded beam without the effect of axial load-
The equations are formulated as first order differential equations
ing, while the second example focuses on the creep buckling
in terms of the unknown incremental deformations and forces
b xx  as follows: phenomenon.
½Du; Dw; D/; DN xx ; DM xx ; D V
b xx þ jA55 D/ þ w; DNxx þ DV c
DV 3.1. Beam under bending only
x
Dw;x ¼ ð39Þ
jA55 þ Nxx
The length of the investigated beam is 1000 mm with a rectan-
gular cross section of 100 mm height and 30 mm width. The beam
A11 ðDMxx þ DM c Þ  B11 ðDNxx þ DNc Þ
D/;x ¼  ð40Þ is subjected to a uniformly distributed load of 10 N/mm that is
A11 D11  B211 applied at t = 0 and remains constant with time. The compliance
3048 E. Hamed / Composite Structures 94 (2012) 3043–3051
functions of the two constituent materials (a and b) in the normal
direction follow Eq. (20) with Ea = 50 GPa, Saxx ¼ Ea =20; 000
n n
h a ; ma ¼ 0:4; and Eb ¼ 100 GPa; Sbxx ¼ Eb =70; 000 h b ; mb ¼ 0:1.
The following power variation function of the materials volume
is considered, with a volume fraction exponent of K = 3:
 K
h  2z
V a ðzÞ ¼ ; V b ðzÞ ¼ 1  V a ðzÞ
2h
The relaxation functions of the two materials in the normal direc-
tion (Raxx and Rbxx ) are obtained via Eq. (21), while those of the
FGM are obtained from Eqs. (16) and (17). Fig. 2 shows the time var-
iation of the relaxation function at five equally spaced selected
points through the depth of the beam, which reveal the different
creep behavior. The spring moduli of the Maxwell chain modeling
of the FGM are determined at these five points by the minimization
of Eq. (14), while the relaxation times are determined through Eq.
(13) with T n1 ¼ 0:1 h, and a = 5. Seven Maxwell units are used to
fit the relaxation function at the different points. The interpolation
function that appears in Eq. (15) is used to obtain a continuous var-
iation of the Maxwell constants through the depth of the beam. The
Maxwell constants that correspond to the shear behavior are as-
sumed to follow the elastic relation with the same relaxation times
as those in the normal direction:
El ðzÞ
Gl ðzÞ ¼
2ð1 þ mðzÞÞ
where m is Poisson’s ratio that is assumed to vary in accordance with
the variation of the material volume fraction as follows:
 K
h  2z
mðzÞ ¼ ðma  mb Þ þ mb
2h
where the Poisson’s ratios of the two materials are taken as ma = 0.2
and mb = 0.3.
The time variation of the peak deflection and peak strains ap-
pear in Fig. 3. The results are normalized with respect to the
instantaneous response, which yields w(x = L/2, t = 0) = 0.85 mm,
cxz(0, 0) = 0.0836‰, exx(L/2, z = h/2, 0) = 0.44‰ and exx(L/2,
z = h/2, 0) = 0.35‰. It can be seen that due to the varying viscoelas-
tic characteristics through the depth of the beam, the normalized
amplifications of the deflection and strains due to creep are differ-
ent. The peak compressive strain at the top of the beam cross-sec-
tion (z = h/2) after 500 h since first loading is about 2.61 the
instantaneous one, while the tensile strain at the bottom (z = h/2)
is only about 1.74 the instantaneous strain. This is because the
FGM examined here exhibits stronger viscoelastic behavior at the
upper part of the cross-section than at the lower one (with respect
to z = 0) as shown in Fig. 2. The peak long-term deflection and
shear strain are 2.2 and 2 times the instantaneous ones, respec-
tively. These differences between the time behavior of the defor-
mations and the normal and shear strains reveal the need for a
detailed structural modeling and analysis of beams made with
Fig. 2. Normal relaxation functions at five points through the depth of the beam.
ð48Þ
Fig. 3. Normalized creep response with respect to the instantaneous response.
FGMs. The shear correction factor that is obtained via Eq. (37) at
t = 0 equals 0.8061, which is slightly smaller than the classical va-
lue of 5/6 for a homogenous rectangular section. With time, the
shear correction factor drops down to 0.7768 at t = 500 h.
The distributions of the maximum normal stresses and strains at
t = 0 and t = 500 h through the depth of the beam appear in Fig. 4.
The results show the nonlinear stress distribution and their signif-
icant redistribution over time. Thus, unlike simply supported
beams made with a homogeneous material where creep increases
the strains only, beams made with viscoelastic FGMs undergo shift-
ð49Þ
ing of the neutral axis (zero strains) that is associated with changes
in the distribution of the normal stresses. These stress and strain
redistributions, which are quantitatively described by the proposed
model, mainly depend on the variation of the viscoelastic material
properties through the depth of the beam cross-section. In the case
ð50Þ
studied here, where the material exhibits stronger viscoelastic
behavior at the upper part of the section, the lower part tends to re-
strain the creep deformations of the upper part, resulting in a signif-
icant increase of the stresses at the bottom point from 33.5 MPa at
t = 0 to 47.6 MPa at t = 500 h. This is an increase of about 42% in the
stress, which may lead to a significant reduction in the load carrying
capacity of the member. As a result of this increase, there is a shift-
ing down of the neutral axis (as also indicated by Hilton [6]) from
z = 5.67 mm at t = 0 (measured positive downwards from the refer-
ence line at the mid-depth of the beam) to 15.32 mm at t = 500 h.
This shifting leads to a release of the compressive stresses at the
upper points as a result of the increase in the area under compres-
sion. In fact, a similar behavior occurs also in reinforced concrete
beams, where the internal steel reinforcement tends to restrain
the concrete creep deformations [15].
In order to shed light on the influence of the variation pattern of
the viscoelastic properties through the depth of the beam, Fig. 5
shows the variation of the peak normalized (with respect to the
instantaneous response) deflection, and longitudinal tensile strain
and stress after 500 h of creep obtained with different volume frac-
tion exponents (K). The results show that for the case in hand, the
peak long-term deflection, stress, and strain significantly increase
in a decreasing rate with the increase of K. The increase is signifi-
cant, and indicates that different combination and gradation func-
tions should be considered for the analysis and design of FGBs,
when there are uncertainties regarding the actual variation of the
material volume fraction (which is commonly the case).
3.2. Beam-column under eccentric axial loading
The second example investigates the geometrically nonlinear
response and the creep buckling of functionally graded beam-
columns under sustained eccentric compressive axial loading.
The same material and geometrical properties as in the first exam-
ple are adopted (with K = 3). Three load scenarios are examined:
(1) a load that passes through the mid-depth of the beam; (2) a
 E. Hamed / Composite Structures 94 (2012) 3043–3051
(a)
Fig. 4. Distribution of strains (a) and stresses (b) at x = L/2 (Legend:
Fig. 5. Peak normalized creep response with different volume fraction exponents.
load that is 5.67 mm below the mid-depth; (3) a load that is
11.34 mm below the mid-depth. Based on Section 3.1, the instan-
taneous neutral axis (centroid) is located 5.67 mm below the
mid-depth, which can also be simply calculated through the
following expression:
Z ,Z
h=2 h=2
00
z0 ¼ E ðz; 0Þz dz E00 ðz; 0Þdz ð51Þ
h=2 h=2
Thus, the second load scenario is actually a centric loading (e = 0),
while the first and third scenarios are loadings with negative
(e = 5.67 mm) and positive (e = 5.67 mm) eccentricities, respec-
tively. The magnitude of the applied axial load is chosen as 0.3Ncr,
where Ncr = 1542 kN, and corresponds to the classical buckling
load of a Timoshenko composite beam, which is calculated as fol-
lows [39]:
N 0cr
Ncr ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð52Þ
1 þ N0cr =A55
where N0cr ¼ p2 Db 11 =L2 , and Leff is the effective buckling length; D
b 11
eff
is the reduced flexural stiffness that is given by ðA11 D11  B211 Þ=A11 .
The time variations of the peak normalized deflection at mid-
span for the three loading scenarios described above are shown
in Fig. 6. The results show that at this level of loading, creep buck-
ling occurs for the first and second loading scenarios. As mentioned
in Section 2, creep buckling is characterized by a rapid increase of
the deformation rate towards infinity. The critical time for which
this phenomenon occurs is characterized here as the time when
the deflection exceeds 1.5 times the depth of the beam (though
other definitions may also apply [37,38]). In this example, it equals
464 h with e = 5.67 mm and 911 h with e = 0. The geometrically
nonlinear response of FGBC is also associated with continuous
shifting of the neutral axis, which increases or decreases the effec-
tive load eccentricity. In the case studied here, the neutral axis
3049
(b)
____
t = 0, - - - t = 500 h).
Fig. 6. Variation in time of the normalized peak deflection with different eccen-
tricities and Nxx/Ncr = 0.3.
shifts downwards over time, resulting in an increase in the load
eccentricity (for the first and second scenarios) in addition to the
increase due to the geometric nonlinear effects (P  D). These ef-
fects continuously increase the deflections and the vulnerability
of the structure to buckling failures.
As shown in Fig. 6, the direction of the eccentricity of the load
exhibits a critical role, which is not the case in homogeneous struc-
tures or symmetric FGS. For eccentricities that are opposite to the
direction of shifting of the neutral axis (scenario 1), the structure is
much more susceptible to creep buckling failures. For positive
eccentricities like the third scenario, Fig. 6 shows that the instanta-
neous deflection of the beam-column is upwards (negative one).
Without shifting of the neutral axis, the deflection is supposed to
increase with time and to remain negative. However, due to the
shifting, the effective eccentricity of the load decreases and the
load becomes a centric one at certain time (11 days in this case)
where the deflection is zero. Beyond this point, the load becomes
an eccentric one (with negative eccentricity) and tends to increase
the positive deflections downwards. For higher load levels, creep
buckling may also occur at this stage, but in the case examined
here with relatively low sustained load (0.3Ncr), no creep buckling
is observed.
To further clarify the results and the structural response, the
distributions of the normal stresses and strains at two different
times for the three load scenarios are shown in Fig. 7. The case with
negative eccentricity is characterized by negative instantaneous
stresses and strains, which then become positive at the bottom
face of the beam due to the creep deformations and the geometric
nonlinearity. The instantaneous strains in the centric loading
(e = 0) are almost uniform, with slight difference between the top
and bottom strains due to the effect of the geometric nonlinearity.
It can be seen that the strain distribution significantly changes
with time, along with a considerable nonlinear redistribution of
the stresses. The distributions of stresses and strains with positive
3050 E. Hamed / Composite Structures 94 (2012) 3043–3051

____
Fig. 7. Distribution of normal stresses and strains at two different times with different loading eccentricities. (Legend: t = 0, - - - t = 50 h).

eccentricity show that at t = 0, larger compressive stresses and

strains develop at the bottom face, which yield negative deflections
of the beam-column. With time, the compressive stresses and
strains at the top face are increased (in absolute values), while
those at the bottom face are decreased as a result of the transition
of the beam-column from negative deflections to positive ones (as
observed in Fig. 6).
As creep buckling of beam-columns is mainly influenced by the
combination of both the magnitude of the axial load and its eccen-
tricity, these two parameters are examined in Fig. 8. The figure
shows the dependence of the critical time to cause creep buckling
on the level of the load for the three load eccentricities described
earlier. For the first and second load scenarios, it is seen that the
critical time significantly decreases with the increase of the applied
load. It can also be seen that in these two cases the Logarithm of Fig. 9. Time variation of the deflection at three load levels with e = 5.67 mm.

the critical time is almost linearly proportional to the level of the

applied load for 0.35 < Nxx/Ncr. The minimum sustained loads that
this scenario also occurs for 0.8 6 Nxx/Ncr. The reason for this
cause creep buckling are 0.275Ncr for e = 5.67 mm and 0.3Ncr for
behavior is owned to the combined but opposite effects of the
e = 0. For 0.7 < Nxx/Ncr, creep buckling occurs very rapidly and it is
shifting of the neutral axis downwards, and the increased geomet-
not presented here. These observations are important for the de-
rically nonlinear effects (P  D) with time due to creep. The former
sign and the safety assessment of FGBCs, as creep can significantly
effect tends to decrease the effective load eccentricity and to bend
decrease the buckling capacity and reduce the design life of such
the structure downwards, while the latter effect tends to increase
structures.
the negative instantaneous deflection (or the positive one if transi-
However, the third loading scenario (e = 5.67 mm) exhibits a
tion occurs). At some load levels, this combination leads to creep
different and unique behavior. It can be seen that increasing the
buckling, while in others, the response becomes stable in time
load level may not necessarily increase the structural susceptibility
and may lead to final positive (for Nxx/Ncr < 0.4) or negative deflec-
to creep buckling failure, which was observed only in the range
tion (for 0.6 < Nxx/Ncr < 0.8). Fig. 9 shows the time-dependent
0.4 6 Nxx/Ncr 6 0.6. Not shown in the figure, but creep buckling in
deflection of the beam-column under three different load levels,
which exhibit the complex and unique structural response. This
behavior actually reveals some of the benefits in using functionally
graded structures, as they can be optimized to prevent creep buck-
ling failures.

4. Conclusions

The challenges associated with the viscoelastic modeling and

Fig. 8. Critical time for creep buckling vs. level of applied load.
analysis of functionally graded beams and beam-columns have
been discussed, and a general theoretical model for their nonlinear
analysis has been developed. A rheological viscoelastic model that
is based on a differential-form of Boltzmann’s principle after the
expansion of the relaxation function into a Dirichlet series has been
adopted. In addition, due to the potential strong nonlinearity asso-
ciated with different viscoelastic characteristics of the constituent
 E. Hamed / Composite Structures 94 (2012) 3043–3051
materials, it has been proposed to evaluate the spring and dashpot
constants in the rheological model using the least squares method
at a number of selected points through the depth of the beam with
interpolation function. The capabilities of the model to deal with
different viscoelastic materials, geometric nonlinearity, and differ-
ent loading schemes, have been demonstrated through numerical
examples.
The results of the numerical study have shown that the creep
response of functionally graded beams under bending is associated
with shifting of the neutral axis, which leads to a significant redis-
tribution of the strains and stresses over time. It also leads to dif-
ferent creep amplification of the strains and stresses with time,
which require special attention in the design and analysis of func-
tionally graded beams. Moreover, it has been shown that unlike
structures made from homogenous materials, the creep buckling
response of functionally graded beam-columns may strongly de-
pend on the direction of the eccentricity of the compressive load
with respect to the instantaneous neutral axis. Thus, functionally
graded beam-columns can be effectively designed to avoid creep
buckling failures when the load eccentricity is known. The loga-
rithm of the time to cause creep buckling failure has been found
to be almost bi-linearly proportional to the level of the applied
load.
In conclusion, the analytical model developed here sets a theo-
retical platform for the viscoelastic analysis of functionally graded
beam-columns and other functionally graded structures. It also
clarifies aspects of the structural behavior and sheds light on the
creep characteristics of functionally graded materials. Yet, further
aspects of the behavior that include material nonlinearity, temper-
ature and other effects, need to be investigated, which can be built
on the modeling concepts and outcomes of this paper.
References
[1] Miyamoto Y, Kaysser WA, Rabin BH, Kawasaki A, Ford RG. Functionally graded
materials: design, processing and application. London: Kluwer Academic
Publishes; 1999.
[2] Shen HS. Functionally graded materials: nonlinear analysis of plates and
shells. New York: CRC Press; 2009.
[3] Birman V, Byrd LW. Vibration of damaged cantilevered beams manufactured
from functionally graded materials. AIAA J 2007;45(11):2747–57.
[4] Roesler J, Paulino G, Gaedicke C, Bordelon A, Park K. Fracture behavior of
functionally graded materials for rigid pavements. Transport Res Record: J
Transport Res Board 2007;2037:40–9.
[5] Venkataraman S, Sankar BV. Elasticity solution for stresses in a sandwich beam
with functionally graded core. AIAA J 2003;41(12):2501–5.
[6] Hilton HH. Optimum linear and nonlinear viscoelastic designer functionally
graded materials – characterization and analysis. Compos Part A: Appl Sci
Manuf 2005;36(10):1329–34.
[7] Shen B, Hubler M, Paulino GH, Struble LJ. Functionally-graded fiber-reinforced
cement composites: processing, microstructure, and properties. Cem Concr
Compos 2008;30(8):663–73.
[8] Librescu L, Oh SY, Song O. Spinning thin-walled beams made of functionally
graded materials: modeling, vibration and instability. Eur J Mech A/Solids
2004;23(3):499–515.
[9] Simsek M. Vibration analysis of a functionally graded beam under a moving
mass by using different beam theories. Compos Struct 2010;92(4):904–17.
[10] Ke LL, Yang J, Kitipornchai S, Xiang Y. Flexural vibration and elastic buckling of
a cracked Timoshenko beam made of functionally graded materials. Mech Adv
Mater Struct 2009;16(6):488–502.
3051
[11] Cai H, Bao G. Crack bridging in functionally graded coatings. Int J Solids Struct
1998;35(7–8):701–17.
[12] Emam SA. A static and dynamic analysis of the postbuckling of geometrically
imperfect composite beams. Compos Struct 2009;90(2):247–53.
[13] Ke L-L, Yang J, Kitipornchai S. Postbuckling analysis of edge cracked
functionally graded Timoshenko beams under end shortening. Compos
Struct 2009;90(2):152–60.
[14] Hamed E. Creep response of sandwich beams with a viscoelastic core and
composite laminates. In: Kollar L, Czigany T, Karger-Kocsis J, editors. 14th
European conference on composite materials, 7–10 June, Budapest, Hungary;
2010 [Cd version].
[15] Hamed E, Bradford MA. Creep in concrete beams strengthened with composite
materials. Eur J Mech – A/Solids 2010;29(6):951–65.
[16] Hamed E, Rabinovitch O. Modeling and dynamics of sandwich beams with a
viscoelastic soft core. AIAA J 2009;47(9):2194–211.
[17] Paulino GH, Jin ZH. Correspondence principle in viscoelastic functionally
graded materials. J Appl Mech 2001;68(1):129–32.
[18] Paulino GH, Jin ZH. Viscoelastic functionally graded materials subjected to
antiplane shear fracture. J Appl Mech 2001;68(2):284–93.
[19] Mukherjee S, Paulino GH. The elastic–viscoelastic correspondence principle for
functionally graded materials, revisited. J Appl Mech 2003;70(3):359–63.
[20] Jin ZH. Some notes on the linear viscoelasticity of functionally graded
materials. Math Mech Solids 2006;11(2):216–24.
[21] Hilton HH, Lee DH, El Fouly ARA. Generalized viscoelastic designer functionally
graded auxetic materials engineered/tailored for specific task performances.
Mech Time-Dependent Mater 2008;12(2):151–78.
[22] Pan F, Li W, Wang B, Zhang X. Viscoelastic fracture of multiple cracks in
functionally graded materials. Comput Methods Appl Mech Eng 2009;198(33–
36):2643–9.
[23] Khan KA, Muliana AH. A multi-scale model for coupled heat conduction and
deformations of viscoelastic functionally graded materials. Composites: Part B
2009;40(6):511–21.
[24] Chen J, Xuan F, Wang Z, Tu S-T. Creep behavior of functionally graded material
under in-plane bending moment. Key Eng Mater 2007;335–358:449–52.
[25] Hilton HH. Tailored designer functionally graded materials for minimizing
probabilistic creep buckling failures in linear viscoelastic columns with large
deformations and follower loads. In: 47th AIAA/ASME/ASCE/AHS/ASC
structures, structural dynamics, and materials conference; 1–4 May 2006,
Newport, Rhode Island; 2006.
[26] Mao YQ, Fu YM, Dai HL. Creep buckling and post-buckling analysis of the
laminated piezoelectric viscoelastic functionally graded plates. Eur J
Mech – A/Solids 2011;30(4):547–58.
[27] Findley WN, Lai JS, Onran K. Creep and relaxation of nonlinear viscoelastic
materials (with an introduction to linear viscoelasticity). Amsterdam: North-
Holland Publishing Company; 1976.
[28] Sankar VB. An elasticity solution for functionally graded beams. Compos Sci
Technol 2001;61(5):689–96.
[29] Zienkiewicz OC, Watson M. Some creep effects in stress analysis with
particular references to concrete pressure vessels. Nucl Eng Des
1966;4:406–12.
[30] Taylor RL, Pister KS, Goudreau GL. Thermomechanical analysis of viscoelastic
solids. Int J Numer Methods Eng 1970;2:45–9.
[31] Bazant ZP, Wu ST. Rate-type creep law of aging concrete based on maxwell
chain. Mater Struct 1974;7(1):45–60.
[32] Jendele L, Philips DV. Finite element software for creep and shrinkage in
concrete. Comput Struct 1992;45(1):113–26.
[33] Nielsen LF. Composite materials: properties as influenced by phase
geometry. Berlin: Springer-Verlag; 2005.
[34] Bert CW. Simplified analysis of static shear factors for beams of
nonhomogeneous cross section. J Compos Mater 1973;7(4):525–9.
[35] Hilton HH. Viscoelastic Timoshenko beam theory. Mech Time-Dependent
Mater 2009;13(1):1–10.
[36] Stoer J, Bulirsch R. Introduction to numerical analysis. New York: Springer-
Verlag; 2002.
[37] Hoff NJ. A survey of the theories of creep buckling. In: Proceedings of 3rd US
national congress of applied mechanics. New York, United States: American
Society of Mechanical Engineers; 1958. p. 29–49.
[38] Hamed E, Bradford MA, Gilbert RI. Creep buckling of imperfect thin-walled
shallow concrete domes. J Mech Mater Struct 2010;5(1):107–28.
[39] Vinson JR, Sierakowski RL. The behavior of structures composed of composite
materials. Dordrecht, The Netherlands: Martinus-Nijhoff, Inc.; 1987.