Beruflich Dokumente
Kultur Dokumente
136]
On: 12 February 2014, At: 04:49
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
To cite this article: Romildo S. Escarpini Filho , Severino P. C. Marques & G. J. Creus (2014) A Model for Viscoelastic
Heterogeneous Materials Based on the Finite-Volume Theory, Mechanics of Advanced Materials and Structures, 21:5, 349-361,
DOI: 10.1080/15376494.2012.680800
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Mechanics of Advanced Materials and Structures (2014) 21, 349–361
Copyright
C Taylor & Francis Group, LLC
This article presents a new numerical model for the analysis of structures of heterogeneous materials with linear viscoelastic con-
stituents. The model is based on the recently developed parametric finite-volume theory that has produced a paradigm shift in the
finite-volume theory’s development. This parametric formulation is here extended to model linear viscoelastic behavior. The present
model employs the state variables approach for the computation of the time-dependent strains. Several examples, including both ho-
mogeneous and heterogeneous situations, are analyzed. Comparison between the numerical and analytical results shows the excellent
performance of the proposed model.
Keywords: viscoelasticity, heterogeneous materials, parametric formulation, finite-volume theory, state variable approach
to merge the micromechanical details directly into the macro- by using the correspondence principle. The comparison be-
level analysis. To have a more complete understanding, for tween the numerical and analytical results shows an excellent
instance, on the viscoelastic behavior of a composite, the in- performance of the proposed model.
fluence of the heterogeneous nature at the microscale on the The article is organized as follows. In Section 2, the para-
macroscale response must be examined in details. metric viscoelastic formulation of the finite volume theory
Lately, computational tools other than finite elements have is described for the cases of plane stress and plane strain.
been proposed to model heterogeneities and their effects un- Section 3 describes the procedures employed to evaluate the
der a variety of thermo-mechanical conditions [16–18]. An viscoelastic strain using the state variables formulation. Sec-
attractive alternative technique is the recently developed para- tion 4 presents the results obtained for several examples and
metric formulation of the finite-volume theory [19], which comparisons of them with analytical and numerical solutions.
incorporated a parametric mapping capability into the FV-
DAM (Finite-Volume Direct Averaging Method) [18]. In the
standard version of FVDAM, the material microstructure 2. Parametric Viscoelastic Formulation of the
is discretized into a grid of rectangular subvolumes. Due
Finite-Volume Theory–Plane Stress and Plane Strain
to the rectangular geometry of the subvolumes, the model-
Problems
ing of microstructures and geometries with curved features
using that standard version, in general, requires extensive
2.1. Local Incremental Equilibrium Equation
discretization in order to reduce the artificial stress concen-
Downloaded by [200.17.114.136] at 04:49 12 February 2014
trations at the boundaries and material interfaces. However, In the parametric finite-volume theory, the whole domain
these limitations have been overcome by the parametric for- occupied by the heterogeneous material is discretized into
mulation of the finite-volume theory, which enables the use quadrilateral subvolumes [19]. The formulation is based on the
of quadrilateral subvolumes in approximating the heteroge- mapping of a reference square subvolume onto each quadri-
neous microstructures, allowing a more efficient modeling of lateral subvolume, as shown in Figure 1. The mapping of the
microstructural details with curvilinear geometry [19]. This point (, ) in the reference subvolume to the corresponding
capability of the parametric formulation for modeling curved point (x, y) in the actual subvolume is given by:
boundaries and interfaces in material with heterogeneous mi-
crostructures has been already demonstrated in several recent x (, ) = N1 (, ) x1 + N2 (, ) x2 + N3 (, ) x3
publications [19–24]. + N4 (, ) x4 ,
This parametric formulation of the finite-volume theory, y (, ) = N1 (, ) y1 + N2 (, ) y2 + N3 (, ) y3
initially formulated for thermo-mechanical analysis of linear
+ N4 (, ) y4 , (1)
elastic composite materials [19], has been extended to include
plastic effects [22] and also used as framework for the de-
velopment of an interesting homogenization technique ap- where Ni are the shape functions given by:
plied to materials with periodic heterogeneous microstructures
1 1
[23, 24]. N1 (, ) = (1−) (1 − ) , N2 (, ) = (1+) (1 − ) ,
A systematic comparison of the convergence characteris- 4 4
tics and processing times of the parametric formulation of 1 1
N3 (, ) = (1+) (1 + ) , N4 (, ) = (1 − ) (1 + ) .
the finite-volume theory and the finite-element method is ad- 4 4
dressed in [21]. To have a meaningful comparison of the execu- (2)
tion times consumed by the two methods in the analyses, the
study was developed using similar conditions of computing The displacement field at time t is approximated using a
environmental, machine, matrix assemblage and solver, etc. second-order expansion in the reference square subvolume
For the problems analyzed in this comparative investigation, coordinates as follows:
the parametric formulation of finite-volume theory showed
1
an excellent performance with respect to the convergence and, u 1 (t) = U1(00)
t
+ U1(10)
t
+ U1(01)
t
+ (32 − 1)U1(20)
t
particularly, in processing times. Another detailed and direct 2
comparative study on convergence of the parametric FVDAM 1
+ (3 2 − 1)U1(02)
t
,
and finite–element predictions is found in [24]. 2
Considering the previous and promising results provided
by the parametric formulation of the finite-volume theory, the (-1,1) (1,1) (x3,y3)
authors developed the present novel numerical model for the F3 (x2,y2)
analysis of heterogeneous materials with linear viscoelastic F4
constituents. The model has been derived on the parametric y (x4,y4)
F2
finite-volume theory framework and employs a state variables F1
(-1,-1) (1,-1) (x1,y1)
formulation for the computation of the viscoelastic strains x
[25].
Several examples, including both homogenous and hetero- Fig. 1. Mapping of the reference square subvolume in the −
geneous materials, are analyzed and their results are compared plane onto a quadrilateral subvolume in the x − y plane of the
with analytical solutions found in the literature or are obtained actual microstructure.
Model for Materials with Viscoelastic Constituents 351
⎧ ⎫
1 ⎪ ∂u 1 ⎪
u 2 (t) = t
U2(00) + U2(10)
t
+ U2(01)
t
+ (32 − 1)U2(20)
t ⎪
⎪ ⎪
⎪
2 ⎪
⎪ ∂ ⎪
⎪
⎪
⎪ ⎪
⎪
1 ⎪
⎪ ⎪
⎪ ⎡ ⎤
+ (3 2 − 1)U2(02)
t
, (3) ⎪
⎪ ∂u 1 ⎪
⎪ 1 0 0 0 0 0 0 0
2 ⎪
⎪ ⎪
⎪
⎨ ∂ ⎬ ⎢0 1 0 ∓3 0 0 ⎥
⎢ 0 0 ⎥
=⎢ ⎥
where Uit(kl) are the polynomial coefficients. Similarly, the in- ⎪
⎪ ∂u 2 ⎪
⎪ ⎣0 0 0 0 1 0 0 0 ⎦
⎪
⎪ ⎪
⎪
cremental displacement field, corresponding to time interval ⎪
⎪ ∂ ⎪
⎪ 0 0 0 0 0 1 0 ∓3
⎪
⎪ ⎪
⎪
[t, t + t] is given by the following second-order approxima- ⎪
⎪ ⎪
⎪
⎪
⎪ ∂u 2 ⎪
⎪
tions: ⎪
⎩ ⎭
⎪
∂ k=1,3
⎧ ⎫
1 ⎪ U1(10) ⎪
u 1 = U1(00) + U1(10) + U1(01) + (32 − 1)U1(20) ⎪
⎪ ⎪
⎪
2 ⎪
⎪ U1(01) ⎪
⎪
⎪
⎪ ⎪
⎪
1 ⎪
⎪ U ⎪
⎪
+ (3 2 − 1)U1(02) , ⎪
⎪ 1(20) ⎪
⎪
2 ⎪
⎨ U ⎪
⎬
1 1(02)
u 2 = U2(00) + U2(10) + U2(01) + (32 − 1)U2(20) × . (6)
⎪
⎪ U2(10) ⎪
⎪
2 ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
1
+ (3 2 − 1)U2(02) . (4) ⎪ U2(01) ⎪
⎪ ⎪
⎪
⎪ ⎪
Downloaded by [200.17.114.136] at 04:49 12 February 2014
2 ⎪
⎪ U2(20) ⎪
⎪
⎪
⎩ ⎪
⎭
U2(02)
The surface-averaged partial derivatives of the incremental
displacement components on each Fk (k = 1, 2, 3, 4) of the
quadrilateral subvolume (Figure 1) are defined in the form:
The surface-averaged partial derivatives of the incremen-
tal displacement field with respect to the Cartesian coordi-
∂u i 1 1 ∂u i
= d , nates are related to those derivatives given in Eq. (6) by the
∂ k=1,3 2 −1 ∂ expressions:
∂u i 1 1 ∂u i
= d,
∂ k=1,3 2 −1 ∂ ⎧ ⎫ ⎧ ⎫
⎪ ∂u 1 ⎪ ⎪ ∂u 1 ⎪
∂u i 1 1 ∂u i ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
= d, ⎪
⎪ ∂x ⎪
⎪ ⎪
⎪ ∂ ⎪
⎪
∂ k=2,4 2 −1 ∂ ⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ∂u 1 ⎪
⎪ ⎪
⎪ ∂u 1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
∂u i 1 1 ∂u i ⎪
⎨ ∂y ⎪
⎬ ⎪ ⎪
= d. (5) Jˆ 0 ⎨ ∂ ⎬
∂ k=2,4 2 −1 ∂ = ,
⎪
⎪ ∂u 2 ⎪
⎪ 0 ⎪ ∂u 2
Jˆ ⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ∂x ⎪
⎪ ⎪
⎪ ∂ ⎪
⎪
Considering Eqs. (4) and (5), the following matrix relations ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
can be readily derived: ⎪
⎪ ∂u 2 ⎪
⎪ ⎪
⎪ ∂u 2 ⎪
⎪
⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
∂y ∂
⎧ ⎫ k=2,4 k=2,4
⎪ ∂u 1 ⎪ ⎧ ⎫ ⎧ ⎫
⎪
⎪ ⎪
⎪
⎪
⎪ ∂ ⎪
⎪ ⎪
⎪ ∂u 1 ⎪
⎪ ⎪
⎪ ∂u 1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎡ ⎤ ⎪
⎪ ∂x ⎪
⎪ ⎪
⎪ ∂ ⎪
⎪
⎪
⎪ ∂u 1 ⎪
⎪ 1 0 ±3 0 0 0 0 0 ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎨ ∂ ⎬
⎢0
⎢ 1 0 0 0 0 0 0⎥⎥ ⎪ ∂u 1
⎪
⎪
⎪
⎪
⎪ ⎪ ∂u 1
⎪
⎪
⎪
⎪
⎪
=⎢ ⎥ ⎨ ∂y ⎬ Jˆ 0 ⎨ ∂ ⎬
⎪ ∂u 2 ⎪ ⎣0 0 0 0 1 0 ±3 0⎦
⎪
⎪ ⎪
⎪ = . (7)
⎪
⎪ ⎪
⎪ ⎪
⎪ ∂u 2 ⎪
⎪ 0 ⎪ ∂u 2
Jˆ ⎪ ⎪
⎪
⎪
⎪ ∂ ⎪
⎪ 0 0 0 0 0 1 0 0 ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ∂x ⎪
⎪ ⎪
⎪ ∂ ⎪
⎪
⎪
⎪ ∂u 2 ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪ ⎪ ⎪ ⎪
⎩ ⎭ ⎪
⎪
⎪ ∂u 2 ⎪
⎪
⎪ ⎪
⎪
⎪ ∂u 2 ⎪
⎪
⎪
∂ ⎪
⎩ ⎪
⎭ ⎪
⎩ ⎪
⎭
k=2,4
⎧ ⎫ ∂y ∂
⎪ U1(10) ⎪ k=1,3 k=1,3
⎪
⎪ ⎪
⎪
⎪
⎪ U1(01) ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ U ⎪
⎪
⎪
⎪ 1(20) ⎪
⎪
⎪
⎨ U ⎪
⎬ where Jˆ stands for the inverse of the volume-averaged Jacobian
×
1(02)
, J¯ defined by:
⎪ U2(10) ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ U2(01) ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ U2(20) ⎪
⎪ 1 +1 +1
⎩ ⎭ J¯ = Jdd, (8)
U2(02) 4 −1 −1
352 R. S. Escarpini Filho et al.
1
being 1
u 1 F4 = u 1 |k=4 = u 1 d = U1(00) − U1(10)
⎡ ⎤ 2 −1
∂x ∂y + U1(20) ,
⎢ ∂ ∂ ⎥ 1
⎢ ⎥ 1
J=⎢ ⎥. (9) u 2 F1 = u 2 |k=1 = u 2 d = U2(00) − U2(01)
⎣ ∂x ∂y ⎦ 2 −1
∂ ∂ + U2(02) ,
1 1
u 2 F2 = u 2 |k=2 = u 2 d = U2(00) + U2(10)
Using the strain-displacement relations, the surface-averaged 2 −1
incremental strains on the faces of the quadrilateral subvolume + U2(20) ,
are given by:
1 1
u 2 F3 = u 2 |k=3 = u 2 d = U2(00) + U2(01)
⎧ ⎫ 2 −1
⎪ ∂u 1
⎪
⎪
⎪
⎪
⎪ + U2(02) ,
⎪
⎪ ∂x ⎪
⎪
⎪
⎪ ⎪
⎪ 1 1
⎪
⎪ ⎪
⎪ u 2 F4 = u 2 |k=4 = u 2 d = U2(00) − U2(10)
⎧ ⎫ ⎪
⎪ ∂u 1 ⎪
⎪
⎪ ⎪ 2 −1
⎨ ε11
⎪ ⎪
⎬ ⎪
⎨ ∂y ⎪
⎬
+ U2(20) , (12)
ε22 = Ē ,
Downloaded by [200.17.114.136] at 04:49 12 February 2014
⎪
⎩ ε ⎭
⎪ ⎪
⎪ ∂u 2 ⎪
⎪
⎪
⎪ ⎪
⎪
12 k=2,4 ⎪
⎪ ∂x ⎪
⎪
⎪
⎪ ⎪
⎪ Using Eqs. (12), it can be shown that the coefficients of the in-
⎪
⎪ ⎪
⎪
⎪ ∂u 2
⎪ ⎪
⎪ cremental displacement expansions are related to the surface-
⎪
⎩ ⎭
⎪
∂y averaged increments of displacement components as follows
⎧ ⎫k=2,4 (i = 1, 2):
⎪ ∂u 1 ⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ∂x ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ 1
⎧ ⎫ ⎪ ⎪
⎪ ∂u 1
⎪ ⎪
⎪ Ui (10) = (u i F2 − u i F4 ),
⎨ ε11
⎪ ⎪
⎬ ⎪
⎨ ∂y ⎪
⎬ 2
1
ε22 = Ē , (10) Ui (01) = (u i F3 − u i F1 ),
⎪
⎩ ε ⎭
⎪ ⎪
⎪ ∂u 2 ⎪
⎪ 2
⎪
⎪ ⎪
⎪
12 k=1,3 ⎪
⎪ ∂x ⎪
⎪ 1
⎪
⎪ ⎪
⎪ Ui (20) = (u i F2 + u i F4 ) − Ui (00) ,
⎪
⎪ ⎪
⎪ 2
⎪
⎪ ∂u 2 ⎪
⎪
⎪
⎩ ⎭
⎪ 1
∂y k=1,3
Ui (02) = (u i F1 + u i F3 ) − Ui (00) . (13)
2
being
For an isotropic linear viscoelastic material occupying the sub-
⎡ ⎤ volume, the increments of stress are related to increments of
1 0 0 0 strain, corresponding to time interval [t, t + t], through the
⎢0 0 0 1⎥ expression:
Ē = ⎢
⎣
⎥.
⎦ (11)
1 1
0 0
2 2 ⎧ ⎫ ⎛⎧ ⎫ ⎧ T ⎫ ⎧ V ⎫⎞
⎪ ε11 ⎪
⎨ 11 ⎪
⎪ ⎬ ⎜
⎪
⎨ ε11 ⎪⎬ ⎪⎨ ε11 ⎪
⎬ ⎪ ⎨ ⎪
⎬⎟
Through Eqs. (4), the following relations for the surface- 22 = C̄ ⎜
⎝⎪ ε 22 − ε T − ε V ⎟
22 ⎪⎠ . (14)
⎪
⎩ ⎪ ⎭ ⎩ ε ⎪ ⎭ ⎪⎩ 22 ⎪ ⎭ ⎪ ⎪ ⎪
averaged increments of displacement components on each face ⎩ ⎭
12 12 0 ε V 12
of the subvolume can be found:
1
1 Here the indices T and V are used to indicate the increments
u 1 F1 = u 1 |k=1 = u 1 d = U1(00) − U1(01) of thermal and viscoelastic strains, respectively, and C̄ is the
2 −1
+ U1(02) , linear elastic constitutive matrix for plane stress and plane
1 strain:
1
u 1 F2 = u 1 |k=2 = u 1 d = U1(00) + U1(10)
2 −1 ⎡ ⎤
+ U1(20) , C̄11 C̄12 0
⎢ ⎥
1 1 C̄ = ⎣ C̄12 C̄22 0 ⎦. (15)
u 1 F3 = u 1 |k=3 = u 1 d = U1(00) + U1(01)
2 −1
0 0 C̄33
+ U1(02) ,
Model for Materials with Viscoelastic Constituents 353
The surface-averaged incremental stress-strain relation for the where b1 and b2 are the increments in the volume forces
subvolume faces can be expressed in the form: in the time interval [t, t + t]. Using Eq. (14) and the rela-
tions strain-displacement in incremental form, the following
⎧ ⎫
expressions are derived for the partial derivatives appearing in
⎨ 11 ⎪
⎪ ⎬
Eqs. (19):
22
⎩ ⎭
⎪ ⎪
12
⎛⎧k=2,4 ⎫ ⎧ ⎫ ⎧ V ⎫⎞ ∂11
⎪ ε T ⎪ ⎪ ε ⎪ = 3C̄ 11 [(Ĵ11 )2 U1(20) + (Ĵ12 )2 U1(02) ]
⎪ ε11 ⎪ ⎪ ⎪ ⎪ ⎬⎟
⎪ ∂x
⎜⎨ ⎬ ⎨ 11 ⎬ ⎨ 11
⎜
= C̄ ⎝ ε22 − ε22 T − ε22 V ⎟ , + 3C̄ 12 (Ĵ11 Ĵ21 U2(20) + Ĵ12 Ĵ22 U2(02) )
⎪ ⎪ ⎠
⎩ ε ⎭ ⎪⎪
⎩
⎪
⎪
⎭ ⎩
⎪
⎪ ⎪
⎭
⎪ ∂T ∂ε11
V
∂ε22
V
12 0 ε12
V
− 1 − C̄ 11 − C̄ 12 ,
⎧ ⎫ k=2,4
∂x ∂x ∂x
⎨ 11 ⎪
⎪ ⎬ ∂22
= 3C̄ 21 (Ĵ11 Ĵ21 U1(20) + Ĵ12 Ĵ22 U1(02) )
22 ∂y
⎩ ⎪
⎪ ⎭
12 + 3C̄ 22 [(Ĵ21 )2 U2(20) + (Ĵ22 )2 U2(02) ]
⎛⎧k=1,3 ⎫ ⎧ ⎫ ⎧ V ⎫⎞
∂T ∂ε11 ∂ε22
⎪ ε11 ⎪ ε11
V V
⎪ ε11 ⎪ ⎪ T
⎪
⎪ ⎪ ⎪
⎪
⎜⎨ ⎬ ⎨ T ⎬ ⎨ V ⎬⎟ − 2
∂y
− C̄ 21
∂y
− C̄ 22
∂y
,
= C̄ ⎜
⎝⎪ ε22 − ε22 − ε22 ⎟ . (16)
Downloaded by [200.17.114.136] at 04:49 12 February 2014
⎪ ⎠
⎩ ε ⎭ ⎪⎪
⎩
⎪
⎭ ⎪
⎪ ⎪
⎩
⎪
⎪
⎭ ∂12
12 0 ε12
V = 3C̄ 33 [Ĵ11 Ĵ21 U1(20) + Ĵ12 Ĵ22 U1(02) + (Ĵ11 )2 U2(20)
k=1,3 ∂x
∂ε12 V
+ (Ĵ12 )2 U2(02) ] − C̄ 33 ,
Using the Cauchy equation, the surface-averaged increments ∂x
of tractions on the faces of the subvolume are given as: ∂12
= 3C̄ 33 [(Ĵ21 )2 U1(20) + (Ĵ22 )2 U1(02) + Ĵ11 Ĵ21 U2(20)
⎛ ⎧ ⎫⎞ ∂y
⎪ 11 ⎪ ∂ε12 V
t1 ⎜ n1 0 n2 ⎨ ⎬
⎟ + Ĵ12 Ĵ22 U2(02) ] − C̄ 33 , (20)
=⎝ 22 ⎠ , ∂y
t
2 k=1,2,3,4
0 n2 n1 ⎪
⎩ ⎭
⎪
12 k=1,2,3,4
(17) where 1 = (C̄ 11 + C̄ 12 )␣, 2 = (C̄ 12 + C̄ 22 )␣, T is the tem-
perature increment, and ␣ stands for the thermal expansion
where n 1 and n 2 are the components of the unit normal vector coefficient.
to each face Fk in the directions x and y, respectively. Through Substituting Eqs. (20) into Eqs. (19), the following relation
Eqs. (6), (7), (10), (16), and (17), the following relation for the for the zeroth order coefficients is obtained:
surface-averaged increments of traction components on the
⎧ ⎫
subvolume faces can be found: ⎪ u 1 F2 + u 1 F4 ⎪
⎪
⎪ ⎪
⎪
U1(00) ⎨ u + u 1 F3 ⎬
⎧ ⎫ ⎧ ⎫ 1 F1
⎪ t U1(10) ⎪ = Φ−1 Θ + Φ−1 Δ, (21)
⎪ 1 F1 ⎪
⎪ ⎪
⎪
⎪
⎪
⎪ ⎪
⎪ U2(00) ⎪
⎪ u + u 2 F4 ⎪
⎪
⎪
⎪ t2 F1 ⎪
⎪ ⎪
⎪ U1(01) ⎪ ⎪ ⎛⎧ ⎫ ⎧ ⎫⎞ ⎪
⎩
2 F2
⎪
⎭
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ u 2 F1 + u 2 F3
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ΔTF1 ⎪
⎪ ⎪
⎪ ΔV F1 ⎪
⎪
⎪
⎪ t1 F2 ⎪
⎪ ⎪
⎪ U 1(20) ⎪
⎪ ⎜⎪⎪ ⎪ ⎪
⎪ ⎪ ⎪
⎪⎟
⎨ t ⎪
⎪ ⎬ ⎪
⎨ U ⎪
⎬ ⎜⎪⎨ ΔT ⎪ ⎬ ⎪ ⎨ ΔV ⎪ ⎬⎟
2 F2 1(02) ⎜ F2 F2 ⎟
= Ā − DC ⎜ + ⎟, where
⎪
⎪ t1 F3 ⎪
⎪ ⎪
⎪ U2(10) ⎪ ⎪ ⎜⎪⎪ ΔTF3 ⎪
⎪ ⎪
⎪ ΔV ⎪⎟
⎪ ⎪ ⎪ ⎪ ⎝ ⎪ ⎪ ⎪ F3 ⎪
⎪ ⎠
⎪
⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎪ ⎪
⎪
⎪ t2 F3 ⎪
⎪
⎪
⎪
⎪
⎪ U2(01) ⎪ ⎪
⎪
⎪
⎩ T ⎭
⎪
⎩ V ⎭
⎪ ⎧ ⎫
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ Δ Δ ⎪ ∂T ∂ε11
V
∂ε22
V
⎪
⎪
⎪ t ⎪ ⎪ ⎪
⎪ U2(20) ⎪ ⎪
F4 F4
⎪
⎪ b − − C̄ − C̄ ⎪
⎪
⎩ 1 F4 ⎪
⎪ ⎭ ⎪
⎩ ⎪
⎭ ⎪
⎪
1 1
∂x
11
∂x
12
∂x ⎪
⎪
t2 F4 U2(02) ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ∂ε12 V
⎪
⎪
(18) ⎪ − C̄ 33
⎪ ⎪
⎪
where Ā = DC E B A. The matrices appearing in the defini- 1 ⎨ ∂x ⎬
tion of Ā are presented explicitly in Appendix A. In Eq. (18), Δ = ,
3⎪
⎪ ∂T ∂ε11
V
∂ε22
V ⎪
⎪
ΔTF k and ΔVFk represent the vectors whose components are ⎪
⎪ ⎪
⎪
⎪
⎪ b2 − 2 − C̄ 21 − C̄ 22 ⎪
⎪
the surface-averaged increments of thermal and viscoelastic ⎪
⎪ ∂y ∂y ∂y ⎪
⎪
⎪
⎪ ⎪
⎪
strains on Fk , respectively. The incremental equilibrium equa- ⎪
⎪ ∂ε12V ⎪
⎪
⎪
⎩ − C̄ 33 ⎪
⎭
tions can be written in the form: ∂y
(22)
∂11 ∂12 and the matrices Φ and Θ are presented in the appendix.
+ + b1 = 0,
∂x ∂y Introducing Eq. (21) into Eqs. (13), the first- and second-
∂21 ∂22 order coefficients of the displacement increments are related to
+ + b2 = 0, (19)
∂x ∂y the surface-averaged increments of displacement components
354 R. S. Escarpini Filho et al.
as follows: thermal and viscoelastic effects. The vectors Δt̄ and Δt̄ o are
⎧ ⎫ ⎧ ⎫ known at the beginning of each time incremental step. Due
⎪ U1(10) ⎪ ⎪ u 1 F1 ⎪ to the interfacial traction continuity conditions on the com-
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ U1(01) ⎪
⎪ ⎪
⎪ u 2 F1 ⎪
⎪ mon interface of adjacent subvolumes, Δt̄ is sparse, with its
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪ U
⎪ ⎪
⎪ ⎪
⎪ u ⎪
⎪ nonzero terms being the increments of surface-averaged trac-
⎪
⎪ 1(20) ⎪
⎪ ⎪
⎪ 1 ⎪
⎪
⎪
⎨ ⎪
⎬ ⎪
⎨
F2
⎪ tions on the discretized boundary regions.
U1(02) u 2 F2 ⎬
= B̄ − NΦ−1 Δ, (23)
⎪
⎪ U ⎪
⎪ ⎪
⎪ u ⎪
F3 ⎪
⎪
⎪
2(10)
⎪ ⎪ 1
⎪
⎪ U2(01) ⎪
⎪ ⎪
⎪
⎪
⎪
⎪
⎪
u 2 F3 ⎪
⎪ 3. Evaluation of the Viscoelastic Strains
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ U2(20) ⎪
⎪ ⎪
⎪ u 1 F4 ⎪
⎪
⎩ ⎭ ⎩ ⎭ 3.1. Linear Viscoelastic Constitutive Relation
U2(02) u 2 F4
In the present model, the time-dependent strain components
where B̄ = P − NΦ−1 ΘM and M, N, and P are given in the are determined incrementally using the state variable approach
appendix. [3, 25] that avoids the need to carry on the whole past history
Substituting Eq. (23) into Eq. (18), the incremental equi- of the deformation.
librium equation of the subvolume is obtained as follows: The most general representation for the constitutive rela-
tion of a linear non-aging viscoelastic material is given by:
⎧ ⎫ ⎧ ⎫
Downloaded by [200.17.114.136] at 04:49 12 February 2014
⎪ t1 F1 ⎪ ⎪ u 1 F1 ⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ t
∂ Di j (t − )
⎪
⎪ t2 F1 ⎪
⎪ ⎪
⎪ u 2 F1 ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ εi (t) = Di j (0) j (t) − j ( ) d , (27)
⎪
⎪ t ⎪
⎪ ⎪
⎪ u ⎪
⎪ ∂
⎪
⎪ 1 ⎪
⎪ ⎪
⎪ 1 ⎪
⎪
0
⎪
⎨ t ⎬
F2
⎪ ⎪
⎨ u ⎪
F2
⎬
2 F2 2 F2 where Di j (t) are the material creep functions. Each one of
=K − Δt 0 , (24)
⎪ ⎪
⎪ t1 F3 ⎪ ⎪ u 1 F3 ⎪
⎪ ⎪ these functions can be approximated as well as needed by a
⎪
⎪ ⎪ ⎪ ⎪
⎪ t2 ⎪
⎪ ⎪
⎪
⎪
⎪
⎪
⎪
u 2 F3 ⎪
⎪ Dirichlet-Prony series:
⎪
⎪ F3 ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
⎪ t1 ⎪
F4 ⎪ ⎪
⎪ u 1 F4 ⎪
⎪ # $
⎩ ⎭ ⎩ ⎭ "
n
t−
t2 F4 u 2 F4 Di j (t − ) = Di0j +
p
Di j 1 − exp − p , (28)
p=1
i j
where K = ĀB̄ is the incremental stiffness matrix of the sub-
volume and Δt 0 is an incremental initial traction vector given p p
where Di0j , Di j , and i j are material constants. For the uniaxial
by: case, Eq. (28) corresponds to the creep function of a general-
! ized Kelvin model with n + 1 springs of constants E p = 1/D p
Δt 0 = ĀNΦ−1 Δ + DC ΔT + ΔV . (25) and dashpots of constants p = p .D p ( p = 0, 1 . . . .n), be-
p
ing 0 = 0 (Figure 2). The parameters i j are the retardation
In this last equation, the vectors ΔT and ΔV contain times.
the surface-averaged increments of thermal and viscoelastic Substituting Eq. (28) into Eq. (27), the following expression
strains on the four faces of the subvolume, respectively. is derived:
"
n "
m
2.2. Global Incremental Equilibrium Equation εi (t) = Di j (0) j (t) +
p
i j (t), (29)
The global incremental stiffness matrix KG is obtained us- p=1 j =1
ing the same procedures employed in [19]. The local stiffness
matrices are assembled into a global system of equations by with the state variables defined by:
applying continuity conditions for the increments of surface- # $
averaged interfacial tractions and displacements, followed by p
p Di j t
t−
the specified boundary conditions. The system of equations re- i j (t) = p exp − p j ( ) d , (30)
i j 0 i j
sulting from this approach relates the increments of unknown
interfacial and boundary surface-averaged displacements to
the increments of surface-averaged tractions, as follows: where no summation is implied over the repeated indices i and
j . In Eq. (29), m stands for the dimension of the stress vector.
K G ΔŪ = Δt̄ + Δt̄ o . (26)
Thus, the total elastic and viscoelastic strain components at 3.2. Numerical Evaluation of the State Variables
time t are given, respectively, by:
Considering a small time interval, t, the following equation
for the state variables defined by Eq. (30) can be written:
εie (t) = Di j (0) j (t) ,
"n " m
p # $
εiV (t) =
p
i j (t) . (31) p Di j t+t
t + t −
i j (t + t) = p exp − p j ( ) d .
p=1 j =1 i j 0 i j
(36)
For an isotropic material and considering the volumetric and
distortional effects separately, the constitutive viscoelastic re-
Assuming j ( ) as constant along the interval t and equal to
lations can be written in the form:
j (t), it can be easily derived from Eq. (36) the expression that
p
t allows obtaining the variables i j at time t + t as functions
∂ DK (t − )
εo (t) = DK (0) m (t) − m ( ) d , of their values at time t:
∂
t0
# $
∂ DG (t − )
ei (t) = DG (0) si (t) − si ( ) d , (32) p p t
0 ∂ i j (t + t) = i j (t) exp − p
i j
# $
Downloaded by [200.17.114.136] at 04:49 12 February 2014
"
n
εoe (t) = DK (0) m (t) , εoV (t) =
p
K (t), or, in compact form,
p=1
"
n "
n
p p
In Eqs. (34), K and Gi indicate the state variables corre- where
sponding to the volumetric and distortional behaviors, which # $ # $
are defined by: t t
p p p p
Si j = i j (t) exp − p , Ri j = Di j 1 − exp − p .
p % & i j i j
p DK t
t−
K (t) = p exp − p m ( ) d , (40)
K 0 K
p t % &
DG t−
p
Gi (t) = p exp − p si ( ) d . (35) Hence, the vector ΔV of surface-averaged increments of
G 0 G viscoelastic strain on the faces Fk of the subvolume is
356 R. S. Escarpini Filho et al.
given by:
⎧ ⎫ ⎛⎡ ⎤⎧ ⎫
⎪
⎪ ΔV
F1 ⎪
⎪
p
SF1 0 0 0⎪ 1⎪
⎪
⎪ ⎪
⎪ ⎪ ⎪
⎪ ⎪
⎪
⎨ ΔV ⎪⎬ n ⎜⎢ ⎥⎪ ⎪
SF2 0 0 ⎥ ⎨ 1 ⎬
p
" ⎜⎢ 0
F2
= ⎜⎢ ⎥
⎪ ⎪ ⎜⎢ p ⎥⎪ ⎪
⎪
⎪ ΔV
⎪
F3 ⎪ p=1 ⎝⎣ 0 0 SF3 0 ⎦ ⎪ ⎪ 1⎪⎪
⎪
⎪ ⎪
⎪ ⎩ ⎪
⎪ ⎭
⎩ ⎭ p
ΔV 0 0 0 SF4 1
⎫⎞ ⎧ V ⎫
F4
⎡ p ⎤⎧ F1 ⎪
R 0 0 0 ⎪ F1 ⎪ ⎪
⎪ ⎪
⎪ ⎪ ⎪ ⎪
⎢ ⎥⎪⎪ ⎪
⎪ ⎟ ⎪
⎪ ⎪
⎪ Fig. 3. Viscoelastic block in a rigid die.
⎢ 0 Rp 0 0 ⎥ ⎨ F2 ⎬⎟ ⎨ V ⎬
+⎢ ⎥ ⎟− F2
⎢ ⎥ ⎟ , (41)
⎣ 0 0 Rp 0 ⎦ ⎪⎪ F3 ⎪
⎪ ⎪⎠ ⎪
⎪ ⎪
⎪ V ⎪
F3 ⎪
⎪
⎪
⎩ ⎪
⎭ ⎪
⎪ ⎪
0 Rp F4 ⎩ V ⎪ ⎭
0 0 F4 friction effect between block and container is neglected. The
value of the normal stress in the y-direction is the unknown.
where Fk , V
p The viscoelastic behavior was modeled by a Maxwell el-
Fk , and SFk indicate the surface-averaged stress
vector, surface-averaged viscoelastic strain vector, and surface- ement with a spring of constant G = 3, 846.154 MPa and a
averaged matrix Sp , on the face Fk and at time t, respectively. dashpot with G = 400 MPa.s. The adopted elastic bulk mod-
Downloaded by [200.17.114.136] at 04:49 12 February 2014
When the material creep functions corresponding to the ulus was K = 8333.333 MPa. As the formulation employs a
volumetric and distortional behaviors are considered, the hy- generalized Kelvin model, the Maxwell model was simulated
drostatic and deviator viscoelastic strains at time t + t, re- by a standard model with a very small spring constant for the
sulting from Eqs. (34) and (35), are given by Kelvin element.
The analytical solution for this example is given by [28]:
"
n
εoV (t + t) = [Sp + Rp m (t)] ,
p=1 6G 3Kt
⎧ V ⎫ y (t) = −o 1 − exp − ,
⎪ e11 (t + t) ⎪ 3K + 4G (3K + 4G)
⎪
⎨ ⎪
⎬
e22 (t + t)
V
⎪
⎪ ⎪
⎪
⎩ V ⎭ where = G /G. Figure 4 presents both the numerical and an-
e12 (t + t)
⎛ ⎧ ⎫ ⎡ p ⎤⎧ ⎫⎞ alytical results for the horizontal compression stress as func-
⎪1⎪ RG 0 0 ⎪ s11 (t) ⎪
"n
⎜ p⎨ ⎬ ⎢ ⎥ ⎨ ⎬⎟ tion of time t. It is observed that the proposed formulation
= ⎜S ⎢ p
0 ⎥ ⎟
⎝ G ⎪1⎪ + ⎣ 0 ⎦ ⎪ 22 ⎪⎠ , (42)
RG s (t) provides results nearly identical with the exact analytical so-
p=1 ⎩ ⎭ p ⎩ ⎭ lution. The absolute maximum errors were about 1.75% for a
1 0 0 RG s12 (t) time increment t = 0.02 s and 0.08% for t = 0.001s. In this
example, a uniform mesh discretization of 5 × 5 subvolumes
where was adopted.
% & % &
p t p t
S =
p
(t) exp − p ,
K R =
p
DK1 − exp − p ,
% K & % K &
p p t p p t
SG = G (t) exp − p , RG = DG 1 − exp − p .
G G
(43)
4. Numerical Examples
where Ne and Nv stand for the number of elastic and viscoelas- e (t) = Eε0 and
% & % % &&
tic layers, respectively. m and s indicate the hydrostatic and t E1 E2 t
normal deviator stress components in the axial direction. v (t) = E1 exp − + 1 − exp − ε0 ,
E1 + E2
Figure 6 shows the numerical and analytical results of the
axial displacements in the panel’s right end as function of time.
The subvolume mesh employed is shown in Figure 5. It is seen being = 2 /E2 and εo = u o /L. The analytical values of the
that the values found by the present formulation present an mean normal axial stress was defined by:
excellent agreement with those corresponding to the analytical
solution. e (t) h e + v (t) (B − h e )
¯ (t) = .
B
4.3. Viscoelastic Bar with an Axial Elastic Reinforcement
In this example, a linear viscoelastic bar, reinforced by a cen-
tered linear elastic fiber and fixed in its left end (Figure 7),
is subjected to two cases of loading: (a) a constant hori-
zontal displacement u 0 = 0.5 cm, which is imposed instan-
taneously on the right free end of the bar, and (b) a uni-
form and instantaneous cooling of T = −153◦ C. The bar
has a length L = 100 cm, total height B = 21.5 cm, and thick-
ness e = 1 cm. The fiber transversal dimension is h e = 1.5 cm. Fig. 7. Viscoelastic bar reinforced by an elastic fiber.
358 R. S. Escarpini Filho et al.
Fig. 8. Relaxation curves for the axial normal stress due to the
imposed displacement u 0 = 0.5cm.
Downloaded by [200.17.114.136] at 04:49 12 February 2014
As observed in Figure 8, the numerical results are in excellent Fig. 10. Axial normal stress distributions over the cross sections
agreement with the analytical solution. The maximum error defined by x = 0.5L and x = 0.7L.
was of 0.046%.
The results obtained for the second loading condition (i.e.,
composite is square with sides of 100 mm. Plane stress and
the instantaneous cooling) are illustrated in Figures 9 and 10.
plane strain conditions were considered. Due to the sym-
For generation of these results, the following thermal expan-
metry of the problem, only one-quarter of the domain was
sion coefficients were adopted for the elastic and viscoelastic
modeled.
material: ␣e = 2 × 10−5 /◦ C and ␣v = 16 × 10−5 /◦ C, respec-
tively. Figure 9 shows the variation of the axial displacement
in the center of the right cross section of the reinforced bar (a)
and non-reinforced bar. Figure 10 illustrates the axial nor-
mal stress distributions over the cross sections defined by the
coordinates x = 0.5L and x = 0.7L.
Fig. 9. Variation of the axial displacements in the panel’s right Fig. 11. Composite with an elastic circular inclusion and mesh
end in function of the time for the instantaneous cooling. discretizations.
Model for Materials with Viscoelastic Constituents 359
5. Conclusions
The matrices Φ and Θ appearing in Eq. (21) are given by: where
W11 + W12 W13 + W14 W11 = C̄11 (Ĵ11 )2 + C̄33 (Ĵ21 )2 , W12 = C̄11 (Ĵ12 )2 + C̄33 (Ĵ22 )2 ,
Φ= ,
W21 + W22 W23 + W24 W13 = (C̄12 + C̄33 )Ĵ11 Ĵ21 , W14 = (C̄12 + C̄33 )Ĵ12 Ĵ22 ,
1 W11 W12 W13 W14 W21 = (C̄33 + C̄21 )Ĵ11 Ĵ21 , W22 = (C̄33 + C̄21 )Ĵ12 Ĵ22 ,
Θ= , (A.8)
2 W21 W22 W23 W24 W23 = C̄33 (Ĵ11 )2 + C̄22 (Ĵ21 )2 , W24 = C̄33 (Ĵ12 )2 + C̄22 (Ĵ22 )2 .
(A.9)
Downloaded by [200.17.114.136] at 04:49 12 February 2014