Sie sind auf Seite 1von 19

journal homepage: www.elsevier.

com/locate/acme
Available online at www.sciencedirect.com
Original Research Article
Linear and Non-linear Creep models for a
multi-layered concrete composite
R. Balevicius
n
, G. Marciukaitis
Dept. of Reinforced Concrete and Masonry Structures, Vilnius Gediminas Technical University, Saultekio av. 11, Vilnius,
Lithuania
a r t i c l e i n f o
Article history:
Received 8 July 2012
Accepted 2 April 2013
Available online 10 April 2013
Keywords:
Non-linear creep
Volterra equation
Multi-layered quasi-brittle
composite
Long-term strength
a b s t r a c t
One- and two-dimensional linear and nonlinear creep models for predicting the time-
dependent behavior of a concrete composite under compression are proposed. These
models use the analytical and iterative solutions of the Volterra integral equation. The
analytical approach is based on the age-adjusted effective modulus method, and the
nonlinear technique applies an iterative approach to the system of non-linear equations,
implying a generalization of the principle of superposition. Both models are validated in
this study.
It has been recently found that negative values of the aging coefcient can emerge in
early age multi-layered composites when the stress redistribution between the layers is
governed by the combination of considerably different creep strains and aging of the
layers.
In the plane-strain state, the two-dimensional creep analysis of multi-layered compo-
sites yields the same vertical stress-time history as that in a one-dimensional case if the
Poisson ratios of the layers are equal. This is valid even though the average value of the
vertical stress used to calculate the Volterra integral term is dependent on the Poisson ratio
of the layers. In particular, the evolution of vertical stress with time is dependent only on
the vertical strain and compatibility conditions in a direction parallel to the lamination.
A fracture mechanics approach is also introduced to predict the gradual degradation of
long-term strength for a multi-layered composite under a sustained compressive load. The
results show that the stress redistribution near the crack-tip under the nal period of a
high level of sustained loading may lead to an additional required compressive stress for
complete failure of the composite. Long-term failure primarily begins with the less
deformable (stiffer) layers because the more-deformable layers can relieve the initial
stresses. Thus, the long-term strength of the composite can exceed its instantaneous
strength for early age composites or for composites composed of layers that possess
considerably different creep and aging properties.
& 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights
reserved.
1644-9665/$ - see front matter & 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved.
http://dx.doi.org/10.1016/j.acme.2013.04.002
n
Corresponding author. Fax: +370 527445225.
E-mail addresses: robertas.balevicius@vgtu.lt, gelz@vgtu.lt (R. Baleviius).
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0
1. Introduction
Recently, composite structures have been widely used in
structural engineering applications. These applications are
primarily governed by economics; however, the thermal
insulating properties and the added strengthening of the
existing structures are still necessary. The strengthening is
important for situations in which the structural elements are
damaged or an increase in load carrying capacity is required.
Traditionally, building composite structures are multi-
layered elements composed of completely different materi-
als, providing a lightweight and economic alternative for
traditional building materials. Among others, the most pop-
ular layered members are the three-layered wall structures
with external layers of a quasi-brittle composite, such as
concrete-based materials, and an internal layer designed
according to the purpose of the structure. Currently, one of
the greatest advantages of a composite structure is a short
production time at the stage when the concrete-based mate-
rials for the layers are applied. The production of these
structures is performed in a single technological cycle rather
than casting one layer after another.
During the life of a structure, sustained loading is pre-
dominant. Thus, a difference in material properties, espe-
cially a time-variation in the physical characteristics of the
concrete layer, can affect the behavior of the composite as a
whole. High-efciency multi-layered composite structures
are designed to account for the time-dependent stress redis-
tribution between the layers. This design phase is difcult
because the materials are often age and creep dependent,
especially in the case of non-perfect bonds between the
layers. According to the studies reported in 1994 for the
industry in Great Britain [1], the cost of destruction resulting
from material plasticization was evaluated to be approxi-
mately 300 million per year, from which 10% is attributed to
destruction induced by the creep process and especially to
the stress-relaxation effects.
When the compressive stress of a layer exceeds a certain
limit, the layer does not exhibit perfect linear creep behavior
and the creep strain is no longer proportional to the applied
stress. This condition can trigger an instability or even
element failure [14].
It is also well known that, in the presence of creep and a
perfect bond between the layers, a composite structure
perfectly redistributes the initial stress over the layers to
maintain the sustained loading [5]. In particular, the higher-
strain layer produces the relaxation, while the lower-strain
layer experiences a stress increase [5]. The delayed failure
would thus occur in the lower-strain material if the creep
strain ceases to be proportional to the acting stress. For
example, in a composite frame of steel columns and compo-
site beams, a signicant amount of moment redistribution
from cracking, creep and shrinkage of concrete occurs [6].
Experience combined with monitoring of three-layered con-
crete composite structures has indicated that the cracking of
the layers may be because there is insufcient bond strength
to accommodate the deformation of the layer [7], [8].
It is analytically difcult to predict creep behavior in
a quasi-brittle multi-layered composite, which possesses
age-variable properties in each of the layers. The best t of
the creep strain to the experimental measurements requires
adopting mathematically complex functions that conse-
quently complicate the derivation of an explicit solution for
a time-varying stress history of the layers (cf. [9]-[12]). The
implications of empirically based approaches are usually
restricted because of the lack of generality and the limitations
of the data.
Numerous methods [13] have been developed to predict
the instantaneous elastic properties of composite materials.
However, fewer methods have been developed that enable
one to evaluate the viscoelastic properties of composite
materials with aging.
Jones [14] modeled the viscoelastic properties of the
composite using an operational approach. In this case, the
stress-strain relations of the viscoelastic composite are
retrieved from the elastic analysis as mappings. When the
creep strain is described by mathematically sophisticated
expressions, derivation of the original stress-strain expres-
sions from the obtained mappings appears to be extremely
complicated.
The effective modulus method, proposed by Faber [15], is
the oldest and simplest approach that is applied to model
concrete-based structures and that overcomes the aforemen-
tioned difculties. This method is also well known, and it is
suitable when the stress tends to be constant in time.
However, for aging materials, this approach results in sig-
nicant errors.
Baant [16] proposed the age-adjusted effective modulus
method as an attractive renement of the latter approach,
introducing the coefcient of aging. This method is theoreti-
cally exact for any problem if the strains vary proportionally
to the creep coefcient. To nd the exact solution, the aging
coefcient must be computed in advance through the inver-
sion of the Volterra integral equation for a given creep
coefcient [10]. This technique is generally used to model
the time-dependent behavior of reinforced concrete struc-
tures [17]. Additionally, in an approximate manner, this
approach was applied to model the creep behavior of CFRP-
strengthened reinforced concrete beams in [18].
Recently, an average stress-strain approach to creep ana-
lysis of reinforced concrete elements has been proposed by
Baleviius [10]. This approach produces the same values of
the time-dependent stress-strain state as those determined
using the well-known age-adjusted effective modulus
method; however, the proposed approach does not require
the introduction of the ctitious incremental restraining
actions, thus dramatically simplifying the computation.
A viscoelastic creep model for polymeric materials, such
as the adhesive Epidian 53: PAC100:80 at ambient tempera-
ture, for different levels of uniaxial stress has successfully
been developed using statistically estimated coefcients [11].
In this case, because of all of the coefcients of the modied
BaileyNorton model were statistically signicant, the num-
ber of statistically estimated parameters in the modied
Burger's model was reduced.
There are very few investigations of non-linear creep
predictions of concrete multi-layered composite with layers
exhibiting age-variable properties found in the literature. The
problems arising from the incompatibility of the layers
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 473
deformational properties are of primary concern in high rise
buildings or concrete layer retrots [19]. Theoretical studies
have examined the nonlinear creep problems in reinforced
concrete. In this case, the reinforced concrete may be treated
as a two-layered composite, but the reinforcement usually
behaves elastically and the direct creep analysis of these
elements does not correspond to concrete multilayer compo-
site behavior.
A comprehensive prediction of the time-dependent stress-
strain state of a multi-layered composite under a triaxial
loading is missing in the literature because of the complexity
of the problem. Some nonlinear multi-axial creep models for
concrete have been developed by Baant and Kim [20],
Benboudjema et al. [21] and Pichler et al. [22].
Yamada et al. [23] were among the rst researchers to
observe the benecial effect of a two-layered composite
structure composed of layers of high-strength and normal-
strength concrete. Later, apko et al. [24] and Sadowska-
Buraczewska [25] conducted short-term experimental and
numerical investigations of beam samples constructed with
a two-layered composite. The composite comprised a high-
performance concrete as the top layer and a normal strength
reinforced concrete as the bottom layer. Tests of this compo-
site revealed a signicant reduction in the compressive
strains and measured deections. The authors concluded
that the observed positive effects resulted from the redis-
tribution of stress between the concrete layers.
In the present study, one- and two-dimensional linear and
nonlinear creep models of a multi-layered concrete compo-
site are investigated. The analytical approach uses the age-
adjusted effective modulus method, and the numerical tech-
nique considers a generalization of the principle of super-
position for the non-linear creep phenomenon. A fracture
mechanics approach is also introduced to predict the gradual
degradation of the long-term strength of the multi-layered
composite under high levels of sustained loading.
2. Model formulation, basic assumptions and
range of application
Consider a time-dependent stress and strain state of the
multi-layered composite element (Fig. 1). The element is
subjected to a sustained loading. Introduce the following
assumptions:

A perfect bond exists between the layers until failure,

There is no connement of the layers from Poisson's


effect,

The stress state is uniaxial,

The second-order effects are negligible,

Each layer obeys the linear/non-linear creep laws of


material aging,

The instantaneous stress-strain state is linear and

Post-failure conditions are ignored.


We also conne the model to creep analysis only. In the
case of linear analysis, the shrinkage (from drying of the
layers or from endogenous shrinkage) may be easily
evaluated separately, relying on the principle of superposi-
tion. For the non-linear creep analysis, the principle of
superposition accounting for the shrinkage is violated and
shrinkage strains should be incorporated into the strain
compatibility equations. In some cases, even endogenous
drying can signicantly change the time-dependent state of
stress for high-strength cement-based materials [26].
It is important to consider the range for the application of
the model (Fig. 1). The bond conditions between the layers are
of primary importance in the performance of the entire
structure under sustained loading. When the bond strength
is sufciently high, the composite structure behaves mono-
lithically, effectively mobilizing all layer strengths, and the
ultimate capacity of the entire composite increases. This is a
necessary condition for a reliable repair of concrete-based
structures, and it is also a necessary condition for the
proposed model.
The common practice to obtain sufcient bond strength is
to rst increase the roughness of the base layer by applying a
bonding agent or steel connectors (transverse reinforcement,
staples, anchors, dispersed reinforcement, etc.), if required
[7], [8], [27], [28].
In this model, a perfect bond between layers can be
achieved when the production of the multi-layered
concrete-based composite involves a single technological
cycle, i.e., all layers are cast simultaneously. The result is
that the age for all the layers is the same, but the material
properties and cross-sectional parameters of the layers can
be different. A single technological cycle for the production of
the composite structure ensures strain compatibility between
the entire composite and each layer [24], and the bond
strength at the interface is higher than the strength of the
adjacent layers. Thus, longitudinal cracks will develop within
the layers rather than at the interfaces [27], [29]. The second
Fig. 1 Schematic of a multi-layered composite.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 474
case of monolithic behavior occurs when the age of the layers
is different but the previously concreted layer is not com-
pletely hardened during the production process. Then, a
mixed chemical and intermixture bond develops between
the previously concreted and new layers. In the contact zone
between the layers, a seam with a unique chemical composi-
tion and structure develops [30]. The elasticity modulus of
this seam may be up to 5 times higher than that of the layers.
To ensure strain compatibility conditions between the
layers when the deformability characteristics of the layers
are completely different, transverse reinforcement, staples,
anchors and dispersed reinforcement can also be utilized
during the casting cycle (Fig. 2). Recently, an innovative shear
connection with composite dowels has been studied by
Lorenc et al. [28].
The above model (Fig. 1) can also be effectively applied for
strengthening and retrotting the structures. As a simple
example of such an application, a new layer can be wrapped
around a concrete column cross-section. This structural
technique can improve the strength, ductility and long-term
performance of the strengthened structure. Following this
structural practice, wrapping a new layer around the old one
produces the monolithic behavior between the old and new
layers. The bond between the cement mortar and concrete
layers is strong when the previously formed layer is hydro-
philic because the layer is moistened well with water and
grout. To ensure compatibility between the strains, the sur-
face must be porous, with knobs and caves. The caves and
knobs can be simply and effectively formed by embrocating
the surface of the strengthening structure with sulfuric acid.
Various investigations have shown that the strength of the
bond between the layers depends on the shape, size and
quantity of the roughness. It has been demonstrated that a
bond formed with cement mortar or concrete is stronger
when the surface knobs and caves have an irregular or
conical shape, and the bond is weaker when the knobs and
caves have a round shape [31], [32].
In most repairing or strengthening cases, the difference in
the layer ages usually prevails and the long-term strength of
the old structure compared with the long-term strength of
the strengthened composite is unknown. Evaluation of these
problems will be discussed below.
Finally, the model does not involve the evaluation of
second-order effects, i.e., geometrical non-linearity is
excluded from the current analysis. Therefore, the multi-
layered composite should be non-slender, i.e.,
c

c;lim
(where

c
and
c;lim
are the slenderness of the entire composite and a
limit value of the slenderness, respectively). A short element
of a rectangular cross-section is non-slender if its length does
not exceed approximately four times the minimal value of
the cross-sectional width. In Fig. 2, the case of a high-length
three-layered wall, but without the second-order effects,
matching the above model (cf. Fig. 1) is also illustrated. In
this illustration, the lateral walls can eliminate the second-
order behavior of the three-layered wall; however, to consider
a three-layered wall separately, the transfer connectors
should be constructed with a mild amount of reinforcement.
A buckling analysis via FEM and approximate relations to
account for the behavior of the exible ties of the three-
layered wall were given in [33]. The FEM modeling and
analysis of critical loads of a three-layered plate with a soft
core composed of foam with elastic and viscoelastic proper-
ties is highlighted and discussed in a previous work [34].
3. The governing relations: a one-dimensional
model
In view of the above assumptions, the time-dependent stress
and strain state of the n-th layered composite can be
evaluated using a system of n+1 equations and n unknowns
[35] as follows:

i n
i 1
s
i
t A
i
N

c
t
1
t 0

c
t
i
t 0

c
t
n
t 0
; i 1; ; n;
_

_
1
where s
i
(t) and
i
(t) are the stress and strain at time t,
respectively; A
i
is the cross-sectional area of the i-th layer;

c
(t) is the strain of the entire composite at time t; and n is the
total number of layers.
In the set of eq. (1), the rst equation represents the force
equilibrium between the internal forces and induced loading
N, while the other equations describe the strain compatibility
between the entire composite and each layer. Because the
creep strain increases with time, the stresses between the
layers redistribute to maintain the force equilibrium. When
the layer stress exceeds a certain limit, the system no longer
exhibits perfect linear creep behavior and the non-linear
creep strain equations should be evaluated.
As shown in [36], although non-linear behavior is consid-
ered, the principle of superposition for the creep strains still
holds. Arutyunyan [9] generalized the principle of super-
position by introducing the nonlinear creep strain. Following
this approach, the i-th layer strain at t can be determined Fig. 2 Application scheme.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 475
as follows:

i
t
s
i
t
E
i
t

_
t
t0
s
i

d
d
1
E
i

_ _
d

_
t
t0
F s
i

C
i
t;

d; i 1; ; n; 2
where s
i
(t) and E
i
(t) are the stress and elasticity modulus,
respectively, of the i-th layer at t, C t; is the specic creep, t
0
is the time instant at the initial loading, and F s
i
is the
non-linear stress function.
A non-linear dependency of creep strain and stress for the i-
th layer for concrete is dened by the following function
proposed by Bondarenko and Bondarenko [37] and later speci-
ed in NIIZHBs specications [38] with some additional factors:
F s
i
s
i
1 v
i
s
i

f
c;i

_ _
m
_ _
; 3
where the coefcient v
i
denes the increase in the creep strain
of the i-th layer during the delayed failure and depends on the
compressive strength, f
c;i
is the material strength of the i-th
layer, and m4 is the factor used to evaluate the creep strain
and stress nonlinearity. A theoretical and experimental analysis
of F s
i
for varying complexities was given in [39]. It was
found that eq. (3) provides a reasonable t to the experimental
data for plain concrete. This construction of the equation is
appropriate for use in our derivation.
Combining (3) with (2), we obtain the following:

i
t
s
i
t
E
i
t

_
t
t0
s
i

i
t;

d
..

cr
i
t;t0
v
i
_
t
t0
s
i

m1
f
c;i

m
C
i
t;

d
..

cr
i
s
i
=f
c;i
;t;t0
; 4
where
i
t;
1
E
i

C
i
t; , i 1,, n.
This relation implies that the total strain of the layer is the
sum of the instantaneous strain, 1=E
i
t, resulting from the
stress s
i
t developed at t; the linear creep strain,
cr
i
t; t
0
,
evaluating the effect of creep and aging; and the non-linear
creep strain,
cr
i
s
i
=f
c;i
; t; t
0
_ _
, reecting the afnity of the creep
corresponding to the different stress-strength levels.
By substituting expression (4) into system (1), we obtain a
set of non-linear equations with unknowns s
i
as follows:

i n
i 1
s
i
t A
i
N
s1 t
E1 t

_
t
t0
s
1

1 t;

dv
1
_
t
t0
s1
m1
f
1

m
C1 t;

d
c
t

s
i
t
E
i
t

_
t
t0
s
i

i
t;

dv
i
_
t
t0
s
i

m1
f
i

m
C
i
t;

d
c
t

sn t
En t

_
t
t0
s
n

n t;

dv
n
_
t
t0
sn
m1
f
n

m
Cn t;

d
c
t
_

_
5
where i is the layer number, and there are n layers.
Assuming that v
i
0, we arrive at the linear formulation of
(4) as follows:

i n
i 1
s
i
t A
i
N
s1 t
E1 t

_
t
t0
s
1

1 t;

d
c
t

s
i
t
E
i
t

_
t
t0
s
i

i
t;

d
c
t

sn t
En t

_
t
t0
s
n

n t;

d
c
t
_

_
6
Thus, the mathematical consideration of (1) for the multi-
layered concrete compressive composite becomes the set of
linear (6) or non-linear (5) equations.
3.1. The instantaneous solution
For the instantaneous analysis, when t t
0
, the strain of the
i-th layer is dened as follows:

i
t
0

s
i
t
0

E
i
t
0

: 7
The substitution of this equation into (1) results in the
equation for the i-th layer stress as follows:
s
i
t
0

NE
i
t
0

i n
i 1
E
i
t
0
A
i
; i 1; 2; ; n: 8
The i-th and j-th layer stresses of the composite interrelate
in terms of a piecewise constant function as follows:
s
i
t
0
s
j
t
0

E
i
t
0

E
j
t
0

; ij; i 1; 2; :::; n; 9
Finally, the entire composite instantaneous stress, strain
and elasticity modulus, induced by load N, can be predicted
using the following formulae:
s
c
t
0

N

i n
i 1
A
i
: 10

c
t
0

N

i n
i 1
E
i
t
0
A
i
: 11
E
c
t
0

i n
i 1
E
i
t
0
A
i

i n
i 1
A
i
; i 1; 2; ; n 12
3.2. Linear creep analysis: analytical approach
For practical applications, it is important to have the relation-
ships for predicting the time-dependent stress-strain state of
the multi-layered composite. Theoretically, an analytical
solution of eq. (4) for the unknown stress history may
occasionally be obtained by reducing the Volterra integral
term to a rst-order differential equation (with variable
coefcients) when a singular kernel is established as a
product of functions of t and or the series of these products
[9], [36]. However, this approach results in considerable
mathematical complications, even for the pure relaxation
test, while, for the multi-layered composite subjected to a
sustained load, it is a practically impossible task. However,
the analytical solution may be dened using the coefcient
of aging.
Suppose the coefcient of aging t; t
0
is known a priori for
a given creep function. Then, for the case of linear creep, the
i-th integral eq. (4) can be rearranged into an algebraic
equation following the Trost-Baant method [16], [40] as
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 476
follows:

i
t
s
i
t
0

E
e;i
t

s
i
t
E
e;i
t
; i 1; 2; ; n; 13
where
E
e;i
t
E
i
t
0

1
i
t; t
0

; 14
E
e;i
t
E
i
t
0

1
i
t; t
0

i
t; t
0

; 15
s
i
t s
i
t s
i
t
0
; 16
where E
e;i
t and E
e;i
t are the effective and age-adjusted
effective elasticity moduli, respectively, for the i-th layer;
s
i
t is the i-th layer stress increment between the initial
stress and the stress developed at t; and
i
t; t
0
is the creep
coefcient.
Eq. (13) describes the unknown stress increment, s
i
t ;
therefore, we rewrite the force equilibrium in (1) in the
incremental form as follows:

i n
i 1
s
i
t A
i
0; i 1; 2; ; n 17
A combination of eq. (17) with the strain compatibility
conditions results in the following equation:

i n
i 1

c
t
s
i
t
0

E
e;i
t
_ _
E
e;i
t A
i
0; 18
From eq. (18), the strain of the entire composite developed
at t can be determined as follows:

c
t

i n
i 1
s
i
t
0
A
i
E
e;i
t=E
e;i
t

i n
i 1
E
e;i
t A
i
; 19
When
c
t
i
t is dened, the combination of formula
(19) with eq. (13) and subsequently with (16) provides the
formula to predict the i-th layer stress at t as follows:
s
i
t s
i
t
0

i
t
s
i
t
0

E
e;i
t
_ _
E
e;i
t : 20
A prediction of the creep coefcient for an entire compo-
site, containing n layers, is calculated using the following
expression:

c
t; t
0


c;cr
t

c;el
t
0



c
t

c
t
0

1; 21
where
c;el
t
0

c
t
0
and
c;cr
t are the elastic, i.e., instanta-
neous, and the creep strains of the composite.
3.3. Linear creep analysis: numerical approach
To validate the above relations, a numerical solution for
system (6) should be considered. Potential solutions, such
as an exponential algorithm, may be found in the literature,
e.g., [41]. This approach would require an additional conver-
sion to the Maxwell/Kelvin chain. We chose to use a different
approach, i.e., we rewrite system (6) into a recurrent form.
For the discretized time scale (t
0
, t
1
, , t
j-1
, t
j
, , t
k
), the
stress relation s
;i
t
j
; t
j1
_ _
s
i
t
j
_ _
=2 s
j
t
j1
_ _
=2 is a middle
value for stress, for [t
j-1
, t
j
], to nd s
i
t
0
, s
i
t
1
, , s
i
t
j1
_ _
,
s
i
t
j
_ _
, , s
i
t
k
. Application of the intermediate value theorem
and the rst mean value theorem for integration yields the
following expression for the total strain of the i-th layer:

i
t
k

s
i
t
k

E
i
t
k


j k1
j 1
s
;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _

i
t
k
; t
j
_ _ _
s
;i
t
k
; t
k1
C
n
i
t
k
; t
k1
; i 1; 2; :::; n; k 1; 2; ; k k; 22
where C
n
i
t
k
; t
k1

1
E
i
t
k1

1
E
i
t
k

C
i
t
k
; t
k1
is the pure specic
creep [12], [42], [43].
Using the expression for stress, s
i
t
k
, from (22) and
substituting into system (6), we obtain an equation for the
strain of the entire composite at t t
k
as follows:

c
t
k
2
i n
i 1
A
i
2=E
i
t
k
C
n
i
t
k
; t
k1

_ _
1
N
i n
i 1

j k1
j 1
s
;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _

i
t
k
; t
j
_ _ _
s
i
t
k1
C
n
i
t
k
; t
k1

2=E
i
t
k
C
n
i
t
k
; t
k1

A
i
_
_
_
_
_
_
_
_
_
_
_
_
i 1; 2; :::; n; k 1; 2; ; k k 23
Relationship (23) represents a recurrent form for the
solution of system (6). For t
i
-t
i-1
-0, the numerical solution
approaches the explicit solution [10]. Thus, the numerical
analysis may also be treated as an exact solution of (6).
A number of variations for the numerical inversion of the
Volterra integral can also be found in [16], [17], [44], [46], [47].
To apply the age-adjusted modulus method, the predic-
tion of the coefcient of aging is required. Thus, we adopt an
explicit inversion of the Volterra equation in terms of the
aging coefcient for the i-th layer following [10] as follows:

i
t
k
; t
0

s
;i
t
k
; t
0
k t
k
s
i
t
k
1k t
k
s
i
t
0

s
i
t
k
s
i
t
0

; for E
c
t E
c
t
0
;
24
where
k
i
t
k

C
n
i
t
k
; t
0

C
i
t
k
; t
0

; 25
s
;i
t
k
; t
0

j n1
j 1
s
;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _
s
;i
t
k
; t
k1
C
n
i
t
k
; t
k1

_ _
C
n
i
t
k
; t
0

;
26

i
t
k
; t
j1
_ _

i
t
k
; t
j1
_ _

i
t
k
; t
j
_ _
; k 1; 2; ; k k 27
In these equations, an average value of stress, s
;i
t
k
; t
0
,
that satises the Volterra integral term, s
;i
t; t
0

_
t
t0
s
i

_
=
i
t; d=C
n
i
t; t
0
, in the entire time interval, t-t
0
, links
to the stress s
;i
t
j
; t
j1
_ _
s
i
t
j
_ _
=2 s
j
t
j1
_ _
=2 for the discrete
time interval [t
j1
, t
j
].
The numerical implementation of the above equations is
achieved using a procedural programming concept with the
FORTRAN 90 language and the Compaq compiler. Primarily, all
parameters dening an instantaneous state were dened
according to the relations described in Section 4. The creep
analysis was then implemented using a three-loop cycle.
The rst (outer) loop runs rst over t instant, for t(t
0
, t
1
,,t
j1
,
t
j
, , t
k
). The second (inner) loop runs for the prediction of all
parameters for the i-th layer, i(1, 2,, n), according to Section 6.
The third loop, working within the second loop, runs for the
Volterra equation variable , discretized using vector t indices
as (t
1
, t
2
, , t
j
, t
j+1
, , t
k-1
). Within this loop, the summation
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 477
of terms
j k1
j 1
in eqs. (23) and (26) was executed. When
the third and second loops stop, a prediction of terms within
the rst loop that depends on variable t
k
nishes the compu-
tation of eqs. (23) and (26) for the composite strain
c
t
k
and
for the averaged stresses satisfying the Volterra integral,
s
;i
t
k
; t
0
. Additionally, the computation of stress s
i
t
k
by eq.
(22) with condition
i
t
k

c
t
k
is also performed here. All of
the stress and strain history as well as the other required
variables, such as the coefcient of aging, etc., were stored
within the appropriate matrix of nk.
Because a small integration step and long creep period are
required for the analysis, the numerical computation always
demands high CPU time, much of which is wasted. To nd
the compromise between desirable computational perfor-
mance and low articial damping, the numerical analysis
should be performed using an increasing time step, such as
t
i

10
16
p
t
i1
[16]. Some analytical suggestions on the selec-
tion of a suitable time step for the numerical integration were
also provided in [48].
4. A two-dimensional model
In a two-dimensional multi-layered composite structure
(Fig. 1), the normal strain,
zz
, is constrained by the nearby
material along the z axis and is small compared to the cross-
sectional strains,
xx
and
yy
, if the structure length in the z
direction is substantially larger than in the other directions.
A plane strain state is an acceptable approximation requiring a
non-zero s
zz
to maintain the constraint
zz
0, thus reducing
the 3-D problem to a two-dimensional problem. Additionally,
the stiffness of such a composite structure (Fig. 1) will be
greater in the direction normal to the layers; however,
because s
zz
0, the two-dimensional structure is exposed to
a smaller vertical strain,
yy
, in the entire composite than that
in a one-dimensional structure.
Assume that the layers of a multi-layered composite
deform elastically, satisfying Hookes law. For the instanta-
neous prediction, the strain tensor for each layer can be
dened as follows:

ij
Ks
ij

1
2G
s
ij
; i; j 1; 2; 3; or i; jx; y; z 28
where 3s s
kk
, s
ij
s
ij
s
ij
is the deviatoric stress tensor,
ij
is
the Kronecker delta, K 12 =E is the bulk compliance
modulus,GE= 2 1 and E are the shear and elastic
moduli, respectively, and is the Poisson ratio of the layer.
4.1. An instantaneous solution
The imposed displacement constraints (Fig. 1), the kinematic
constraint, i.e., a uniform strain develop within the layer, the
strain compatibility, the layers strain and the entire compo-
site strain in the z and y directions are the same, and the
plane strain state are expressed mathematically as follows:

zz;i

zz;c
0;
yy;i

yy;c
;
xy;i

zy;i
0; s
xx
j
x x0
s
xx
j
x xn
0;
u
xx
j
x x0
u
yy
j
y 0
0. Introducing these conditions in (28)
together with the force equilibrium in the y direction at
t t
0
results in the following:

i n
i 1
s
yy;i
t
0
A
i
N

yy;c
t
0

yy;i
t
0

yy;c
t
0
1
2
i
_ _
s
yy;i
t
0
=E
i
t
0
0

xx;i
t
0

i
1
i
s
yy;i
t
0
=E
i
t
0

; i 1; :::; n;
_

_
29
where
yy;c
t
0
is the strain of the entire composite at time t
0
.
The solution of system (29) yields the following formulae:
s
xx;i
t
0
0; 30
s
zz;i
t
0

i
s
yy;i
t
0
; 31
s
yy;i
t
0
s
yy;j
t
0

E
i
t
0

E
j
t
0

1
2
j
_ _
1
2
i
_ _ ; ij; i 1; 2; ; n; 32
s
yy;i
t
0

NE
i
t
0

1
2
i
_ _

i n
i 1
E
i
t
0
A
i
=1
2
i

; i 1; 2; ; n 33

yy;c
t
0

N

i n
i 1
E
i
t
0
A
i
=1
2
i

: 34
The horizontal displacement of the free surface is
obtained by integration of
xx;i
t
0
along the x axis as follows:
u
xx

x xn

i n
i 1

i
1
i

_
x
i1
x
i
s
yy;i
t
0
=E
i
t
0
dx: 35
Comparison of eq. (33) for the two-dimensional model
with (8) for the one-dimensional case indicates that the
vertical stresses are different when the Poisson ratios of the
layers are not equal. When the Poisson ratios are the same,
there is no difference in the vertical stresses of the one- or
two-dimensional composite structures. The comparison of
the vertical strains for the entire composites (34) and (11)
shows that the strain in the two-dimensional model is stiffer
in the y direction because of the action of the constraining
stress s
zz;i
t
0

i
s
yy;i
t
0
along the z axis. For example, using
the Poisson ratio
i
0:2 for the uncracked plain concrete for
all layers, the strain
yy;c
t
0
of the two-dimensional composite
structure is less than that of the one-dimensional structure
by approximately 1.04 times. For different Poisson ratios of
the layers, this difference depends on the term 1
2
i
_ _
in the
summation term of (34).
4.2. Linear creep analysis
Applying the principle of superposition for the axial and
transverse strains, Arutyunyan [9] generalized Hookes law
(28) for the linear creep problem as follows:

ij
t K t s t
ij

s
ij
t
2G t

_
t
t0
s
ij

t; t; s
kk
t
ij

t;
_ _
d; 36
where K t 12 t =E t , G t E t = 2 1 t ,

t;

E

t; C t; , and and t; are the coefcients of transverse
instantaneous and creep strains (Poisson ratios), respectively.
Following the experimental investigations described in
[36], t; is slightly lower than and the effect of different
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 478
values of these coefcients on a two-dimensional stress
strain state of the concrete specimens was minimal. Thus,
it can reasonably be assumed that t; t; const,
allowing the simplication of eq. (36). In particular, this
results in

t; t; , and the strain


zz;i
t from eq. (36)
can be written in a more convenient form for the i-th layer as
follows:

zz;i
t
S
z;i
t
E
i
t

_
t
t0
S
z;i


i
t; d 0; 37
where
S
z;i
s
zz;i
s
xx;i
s
yy;i

_
: 38
Incorporating (31) and (32), we obtain s
zz;i

i
s
yy;i
,
s
xx;i
t
0
s
xx;i
0, and thus, S
z;i
0. However, a zero of
S
z;i
can be directly obtained from (37). It shows that the
vertical stress function, s
yy;i
, is independent of condition
(37) and depends only on the vertical strain,
yy;i
t , and the
compatibility conditions in a direction parallel to the
lamination.
For the vertical strain equation, we rewrite eq. (36) as
follows:

yy;i
t
S
y;i
t
E
i
t

_
t
t0
S
y;i


i
t; d
yy;c
t ; 39
where
S
y;i
s
yy;i
s
xx;i
s
zz;i

_
: 40
Incorporating (31) and (32), we obtain S
y;i
1
2
i
_ _
s
yy;i
.
Hence,

yy;i
t
s
yy;i
t
E
i
t

_
t
t0
s
yy;i


i
t; d
s
yy;i
t
E
i
t
s
;yy;i
t; t
0
C
n
i
t; t
0

yy;c
t
1
2
i
_ _ : 41
The comparison of (41) and (6) indicates that the vertical
strain at time t of the entire composite in a two-dimensional
formulation resolves similarly, as shown in relation (22)
(independent of (37)) for the one-dimensional case. Thus:

yy;c
t
k
2
i n
i 1
A
i
1
2
i
_ _
2=E
i
t
k
C
n
i
t
k
; t
k1

_ _
_ _
1
N
i n
i 1

j k1
j 1
s
;yy;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _

i
t
k
; t
j
_ _ _
s
yy;i
t
k1
C
n
i
t
k
; t
k1

1
2
i
_ _
2=E
i
t
k
C
n
i
t
k
; t
k1

_ _ A
i
_
_
_
_
_
_
_
_
_
_
_
_
;
i 1; 2; :::; n; k 1; 2; ; k k: 42
where s
;yy;i
t
j
; t
j1
_ _
s
yy;i
t
j
_ _
=2 s
yy;j
t
j1
_ _
=2 is the average ver-
tical stress for the discrete time interval [t
j-1
, t
j
].
An average value of stress, s
;yy;i
t
k
; t
0
, satisfying the
Volterra integral term is determined by relation (43) as
follows:
s
;yy;i
t; t
0
1
2
i
_ _

j n1
j 1
s
;yy;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _
s
;yy;i
t
k
; t
k1
C
n
i
t
k
; t
k1

_ _
C
n
i
t
k
; t
0

43
Substituting the strain,
yy;c
t
k
, calculated by formula (42)
into eq. (41) and combining with relation (43), the vertical
stress, s
yy;i
t , at any t can be determined. When const for
all the layers, s
yy;i
t in the two-dimensional model is the
same as that in the one-dimensional model.
Finally, the horizontal strain,
xx;i
, for the i-th layer can be
integrated using the vertical stress function, s
yy;i
t , as follows:

xx;i
t
S
x;i
t
E
i
t

_
t
t0
S
x;i


i
t; d
S
x;i
t
k

E
i
t
k

1
2

j n
j 1
S
x;i
t
j
_ _
S
x;i
t
j1
_ _ _ _

i
t
k
; t
j1
_ _
; 44
where
S
x;i
s
xx;i
s
yy;i
s
zz;i

_

i
1
i
s
yy;i
45
The stress-strain relations involving the nonlinear creep
phenomena were presented in [45] for the two-dimensional
state. These relations can be applied for the current multi-
layered composite structure in a similar way. The analysis
below applies to the modeling of one-dimensional
nonlinear creep.
5. Non-linear creep analysis: a one-
dimensional model
An explicit solution of the system of nonlinear eq. (5) for the
layer stresses at t is a complicated task. However, an analy-
tical solution is available for some cases if the parallel (rather
than afne) creep curve assumption for material aging is
adopted. Otherwise, explicit solutions for polynomials with
degrees greater than four do not exist. Finally, complications
solving system (5) arise as the total number of layers
increases.
In this study, we adopt a numerical solution for the i-th
layer with an unknown stress history, s
i
t
0
, s
i
t
1
, , s
i
t
j1
_ _
,
s
i
t
j
_ _
, , s
i
t
k
. Then, the application of an intermediate value
theorem and the rst mean value theorem for the integration
of (4) results in the following:

i
t
k

s
i
t
k

E
i
t
k


j k1
j 1
s
;i
t
j
; t
j1
_ _

i
t
k
; t
j1
_ _

s
i
t
k
s
i
t
k1

2
C
n
i
t
k
; t
k1

v
i

j k1
j 1
s
;i
t
j
; t
j1
_ _
m1
f
;i
t
j
; t
j1
_ _
m
C
i
t
k
; t
j1
_ _

1
2
v
i
s
i
t
k
s
i
t
k1

m1
f
c;i
t
k
f
c;i
t
k1

_ _
m
C
i
t
k
; t
k1
46
(for i 1,, n; k1, , kk )
where
C
i
t
k
; t
j1
_ _
C
i
t
k
; t
j1
_ _
C
i
t
k
; t
j
_ _
; 47
f
;i
t
j
; t
j1
_ _
f
c;i
t
j
_ _
=2 f
c;i
t
j1
_ _
=2; 48
C
n
i
t
k
; t
k1

1
E
i
t
k1

1
E
i
t
k

C
i
t
k
; t
k1
: 49
By substituting
i
t
k
for
i
t in (1), system (5) can be
rearranged into the following set of non-linear functions,
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 479
f
i
, as follows:
f
1
s
1
t
k
; s
2
t
k
; :::; s
n
t
k
0
f
i
s
i1
t
k
; s
i
t
k
0

f
n
s
n1
t
k
; s
n
t
k
0
; i 2; n;
_

_
50
where:
f
1
s
1
t
k
; s
2
t
k
; ; s
n
t
k

i n
i 1
s
i
t A
i
N; 51
f
i
s
i1
t
k
; s
i
t
k

i1
s
i1
; t
k

i
s
i
; t
k
; 52
where
i
s
i
; t
k

i
t
k
, using (46).
Generally, system (50) involves n non-linear equations
with n unknowns. Strictly, system (50) with respect to
eq. (46) represents a set of non-linear recurrence relations
dening the successive terms of a sequence s
1
t
k
; s
2
t
k
;
_
:::;
s
n
t
k
g
T
R
n
as nonlinear functions of the preceding terms.
We can rearrange the nonlinear system (50) into a classical
Newtons type of system of linear equations over all t
0
, t
1
, , t
k
1
, t
k
in t
0
t
k
. Assume that s
i
: t; t
0
-R (for any i1, n)
is a continuous differentiable function in [t, t
0
]. Assume that
f
i
s
1
; :::; s
n
: s
i
s
i
t
0
; s
i
t R
n
(i 1, n) are also contin-
uous differentiable functions. Let f
1
=s
1
; :::; f
1
=s
n
, ,
f
n
=s
1
; :::; f
n
=s
n
be the derivatives of f
1
s
1
; :::; s
n
,
f
n
s
1
; :::; s
n
at the points f
1
s
1
t
k
; :::; s
n
t
k
,
f
n
s
1
t
k
; :::; s
n
t
k
, while ds
1
=d, , ds
n
=d are the
derivatives (i.e., the stress ux) over s
1
, , s
n
at any point
t
k
.
Then, derivatives of the complex functions exist such that
df
1
d

f
1
s1
ds1
d
; ;
f
1
sn
dsn
d

df
n
d

f
n
s1
ds1
d
; ;
f
n
sn
dsn
d
_

_
53
or,
df
i
d

j n
j 1
f
i
s
j
ds
j

d
; i 1; n 54
In mathematical terms, from (54), a differential for func-
tions f
i
s
1
; :::; s
n
is independent of d, resulting in the
following:
df
i

j n
j 1
f
i
s
j
ds
j
; i 1; n: 55
The increment of f
i
s
1
; :::; s
n
can be dened as follows:
f
i
f
i
s
1
t
k
s
1
t
k
; :::; s
n
t
k
s
n
t
k

f
i
s
1
t
k
; :::; s
n
t
k
; i 1; n: 56
Assuming df
i
f
i
consider t
k
. Then, ds
i
t
k
s
i
t
k
,
s
i
s
i
t
k
, and from (55) and (56), we express the following
approximate relationship for any f
i
:
f
i
s
1
t
k
s
1
t
k
; :::; s
n
t
k
s
n
t
k

f
i
s
1
t
k
; :::; s
n
t
k

j n
j 1
f
i
s
j
t
k

s
j
t
k
; i 1; n 57
Relation (57) traditionally approximates the non-linear i-th
function using a linear equation involving only a value of this
function and the sum of its rst derivative at the points. The
application of a Taylor series has been comprehensively studied
in [46], [47] for creep analysis of one-layered pre-stressed
cellular concrete wall panels. In these investigations, the con-
cept of a Taylor expansion was implemented for obtaining the
solution of the linear Volterra integral for an unknown stress
function and its derivatives as a set of linear algebraic equa-
tions. Those studies found that to maintain a sufcient accu-
racy in the approximation, when t-t
0
increases, a minimum of
ten Taylor series terms should be kept in the computation.
Generally, instead of rst-order Taylor series terms in eq. (57),
the implementation of the ten terms for each layer could be
adopted because the high-order derivatives for f
i
s
1
; :::; s
n
can
be dened explicitly. However, this implementation would not
be reasonable in engineering computations because of the
conversion of the linear equations into nonlinear ones at each
time step for unknown s
i
t
k
.
Thus, because the nonlinear problem is replaced by the
linear eq. (57), errors occur during the replacement. Hence,
the solution is not exact, but it obviously may be dened
within an allowed tolerance of tol units.
Let
q1
s
1
t
k
; :::;
q1
s
n
t
k
be a new approximation for the
layer stress, and let
q
s
1
t
k
; :::;
q
s
n
t
k
be the current approx-
imation. The new approximation is calculated as follows:
q1
s
i
t
k

q
s
i
t
k

q
s
i
t
k
; i 1; n; 58
where
q
s
1
t
k
; :::;
q
s
n
t
k
is the set of errors for the q-th
approximation or the unknown layers stress increments at
time t
k
and iteration q.
Inserting (58) into (57) and replacing s
1
t
k
; :::; s
n
t
k
by
q
s
1
t
k
; :::;
q
s
n
t
k
, then substituting
q
s
1
t
k
; :::;
q
s
n
t
k
for
s
1
t
k
; :::; s
n
t
k
, we obtain the following:
f
i
q1
s
1
t
k
; :::;
q1
s
n
t
k

_ _
f
i
q
s
1
t
k
; :::;
q
s
n
t
k

_ _

j n
j 1
f
i
s
j
t
k

q
s
j
t
k
; i 1; n 59
When
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
-0, (59) becomes f
i
q1
s
1
t
k
;
_
:::;
q1
s
n
t
k
-f
i
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
, and then the stresses at
the q+1-th approximation are equal to the stresses at the q-th
approximation, i.e., the solution converges.
From (50), function f
i
q1
s
1
t
k
; :::;
q1
s
n
t
k

_ _
0. Summing
over n in (59), we classically arrive to the Newtonian method
for the numerical solution of the system of non-linear
equations as follows:
f
1
s1 t
k

q
s
1
t
k
:::
f
1
sn t
k

q
s
n
t
k
f
1
q
s
1
t
k
; :::
q
s
n
t
k

_ _

f
n
s1 t
k

q
s
1
t
k
:::
f
n
sn t
k

q
s
n
t
k
f
n
q
s
1
t
k
; :::
q
s
n
t
k

_ _
_

_
60
In a classical way, system (60) may be conveniently
rewritten into a matrix of linear equations with an unknown
vector of the layer stress increments
q
r t
k

_ _
at iteration q
and any t
k
as follows:
J
q
s t
k

_ _ _
q
r t
k

_ _
f
q
s t
k

_ _ _ _
; 61
where
J
q
s t
k

_ _ _

f
1
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
T

f
n
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
T
_

_
_

_
62
q
r t
k

_ _

q
s
1
t
k
; :::;
q
s
n
t
k

_ _
T
63
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 480
f
q
s t
k

_ _ _ _

f
1
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
T

f
n
q
s
1
t
k
; :::;
q
s
n
t
k

_ _
T
_

_
_

_
64
where J
q
s t
k

_ _ _
is an n-by-n Jacobian matrix involving a
gradient of vector f
q
s t
k

_ _
at points
q
s
1
t
k
; :::;
q
s
n
t
k
for the
iteration q at any t
k
.
The Jacobian n-by-n matrix with respect to eqs. (5) and (50)
for a multi-layered composite can thus be written as follows:
J
q
s t
k

_ _ _

A
1
A
2
A
3
A
n1
A
n
f
2
s1 t
k

f
2
s2 t
k

0 0 0
0
f
3
s2 t
k

f
3
s3 t
k

0 0

0 0 0
f
n
sn1 t
k

f
n
sn t
k

_

_
_

_
65
The major difculty in this formulation is the explicit
denition required for all functions included in the Jacobian.
The formulation of (50) with respect to (46) allows explicit
differentiating over s
1
t
k
; :::; s
n
t
k
. Thus, all terms in the
Jacobian are generalized using the following formulae:
f
i
s
j
t
k


_
_
_
1
E
j
t
k


1
2
C
n
j
t
k
; t
k1

m 1
2
v
j
q
s
j
t
k

q
s
j
t
k1

_ _
m
f
c;i
t
k
f
c;i
t
k1

_ _
m
C
j
t
k
; t
k1

_
_
_
f
i
=s
z
t
k
0, for zi & zj
_ _
; where z1,n; i 2,n, j i1
,i;
where 1
ij1
is provided for the sign conversion.
Thus, it has been demonstrated that the nonlinear creep
problem discretized over all t
0
, t
1
, , t
k-1
, t
k
in t
0
-t
k
may be
rearranged for unknowns
q
s
1
t
k
; :::;
q
s
n
t
k
developed at
the current time instant t
k
. The layer stresses accumulated
from any preceding time t
k1
may be treated as independent
parameters for the current time instant t
k
, where the explicit
Jacobian functions depends only on two stresses, s
i
t
k1
and
s
i
t
k
.
The adopted solution algorithm is as follows. At time t
k
,
we replace the system of nonlinear eqs. (5052) with the
linear one (61) and determine the unknowns
q
s
1
t
k
; :::;
q
s
n
t
k
at iteration q using the following formula:
J
q
s t
k

_ _ _
1
f
q
s t
k

_ _ _ _ _ _

q
r t
k

_ _
: 66
Then, the solution obtained for
q
r t
k

_ _
is employed to
obtain a better approximation for q+1 as follows:
q1
r t
k

_ _

q
r t
k

_ _

q
r t
k

_ _
67
When
q1
r t
k

_ _

q
r t
k

_ _
otol is satised, the iteration
process is terminated, achieving convergence between itera-
tions q and q+1. The iterative computation is repeated until
the last time instant in the time series. The stress vector
0
r t
k1

_ _
is used as the rst guess.
As an illustration, snapshots from the nonlinear creep
analysis of system (50), which are described in the example
below, are shown in Fig. 3. Eq. (51), describing static equili-
brium, is represented by the plane (Fig. 3a). The strain
compatibility eq. (52) (Fig. 3b) yields a nonlinear surface. In
Fig. 3c, the solution for r t
_ _
at the intersection of the
surfaces is demonstrated.
The iterative process runs well and converges at all times
considered. However, at time intervals near t
0
, instabilities in
the layer stress eld began to appear. To overcome these
instabilities, a small initial time step, t
1
-t
0
equal to 0.0001 day,
provided a fairly accurate result.
6. Application and results: a one-dimensional
model
6.1. The linear analysis and its numerical verication
The proposed linear approach represents an explicit inver-
sion of the Volterra integral equation without using any
empirical factors or simplications. Therefore, comparing
the results obtained using the proposed method to those
measured experimentally would be a verication of the
fundamental Volterra equation for modeling a time-
variable stress-strain state of concrete, but this comparison
is not a verication of the method. Errors arising from the
application of the Volterra superposition and those emerging
from the best t of creep curves to the experimental mea-
surements are well-known [9], [12], [36]. Thus, a direct
calculation of the time-dependent stress-strain values using
the proposed approach and their comparison to those pro-
duced using the numerical technique is the most efcient
way to verify the analytical method.
Consider a three-layered composite wall structure (Fig. 4).
Assume that the concrete composite structure is made of
concrete with creep properties described by code EN 1992-1-1
2004 [49]. Assume that the outer layers have a high-strength
class of compressive concrete, C90/105, and that the inner
layer is made of a low-strength compressive class concrete,
C12/15. Hereafter, it will be denoted as a high/low/high
strength layered composite structure.
The cylinder mean compressive strength for each layer is
equal to f
cm;1
f
cm;3
98 MPa and f
cm;2
20 MPa. The devel-
opment of the elasticity and strength moduli in time are also
determined by [49]: E
i
t

cc;i
t
_
E
i
28 and f
cm;i
t
cc;i
t
f
cm;i
28 (where
cc;i
t is the coefcient evaluating the age of
the concrete for i-th layer). Additionally, the elasticity moduli
of the layers are related to the cylinder mean strength, f
cm;i
,
by the formula E
i
28 2:1510
4

f
cm;i
=10
3
_
.
The values for the creep coefcient, t; t
0
, determined by
code [49] should be additionally multiplied by the factor
cc
t
because of the code-based methodological peculiarities (see,
Ghali et al. [50]).
Graphs showing the creep coefcient, t; t
0
, used in the
current analysis for the inner and outer layers are shown in
Fig. 5. The creep coefcient and specic creep, which are
required for the above model, are interrelated as C
i
(t, t
0
)
(t,t
0
)/E
i
(t
0
).
As shown in Fig. 5, the creep strain of the inner layer is
approximately 2.6 times higher than that of the outer one, cf.

2
t; t
0
2:023 and
1;3
t; t
0
0:767 (at t
0
1, t30000 days).
In this example, the values of
i
t; t
0
are given for an effective
thickness of the layers, h
1,3
90.9 mm, h
2
166.7 mm. The rela-
tive humidity is selected to be RH95%. For this selection,
the examined parameters are considerably different to
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 481
demonstrate the important peculiarities that may arise in the
design of this type of structure.
The thickness of the outer layers is 0.1 m (A
1
A
3
0.1 m
2
),
and that of the inner layer is 0.2 m (A
2
0.2 m
2
). For the linear
analysis, the composite wall is subjected to a sufciently
low sustained loading, equal to N1.0 MN, induced at
t
0
1, 10, 28, 90 and 730 days. It represents an early-,
normal-, and late-age concrete used to construct the layered
wall structure.
The stress-strain state of the layers and the entire com-
posite at any time t is shown in Fig. 5. The coefcient of aging
of the layers (Fig. 5b) is computed by eq. (24), evaluating the
formulae (2527). The stress history (Fig. 5c) is derived from
eq. (22), expressing s
i
t
k
for
i
t
k

c
t
k
at any t
k
, for i 1, 2, 3.
Meanwhile, the strain of the entire composite
c
t
k
(Fig. 5d) is
determined by formula (23).
The following conclusions can be drawn from the results. In
Fig. 5c, the layer possessing the higher strain tends to produce a
relaxation, whereas the layers with a lower strain experience the
stress increase to maintain the equilibrium between internal and
external forces. If non-linear creep occurs, failure could primarily
originate from the layer that possesses a lower strain.
As shown in Fig. 5b, the variation in the aging coefcient of
the layers is different. This variation depends on the stress
history and the aging of the layers. The descending branches
of the aging coefcient are governed by the material aging;
however, for old concrete, the aging process is minimal, resulting
in
i
t; t
0
-1.
The values of the creep coefcient dened by (21) for the
entire composite (denoted by bold lines) fall mainly within
the values of the layers creep coefcients (Fig. 5a). In addition,
the nal values of the entire composite,
c
30000; 1 1:156 and

c
30000; 10 1:126, are virtually negligible, indicating that the
difference in the layer age is mild.
This simulation can be used to validate the proposed
analytical approach. To this end, we select the loading time
t
0
1 day and the time of consideration t 310
4
days. In this
case, from Fig. 5(a, b), the creep and aging coefcients
are
1
t; t
0

3
t; t
0
0:767 and
1
t; t
0

3
t; t
0
0:014,
respectively, for the outer layers. The values for these para-
meters for the inner layer are t; t
0
2:147and
2
t; t
0

0:305, respectively. The elasticity moduli at the time of
loading are as follows: E
1
t
0
E
3
t
0
26907:61 MPa, E
2
t
0

15841:99 MPa.
Fig. 3 Illustration of the solution of the system of non-linear eq. (50) for a three-layered structure: (a) plane of static equilibrium
equation; (b) non-linear surface of strain compatibility equations; and (c) intersection of plane and surface as a solution for
unknown layer stresses r
1
and r
2
.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 482
An instantaneous stress-strain state. The axial stiffness of the
three-layered wall structure is as follows:

i n
i 1
E
i
t
0
A
i
226907:610:1 15841:990:2 8549:80 MN:
Thus, the layers stresses at t
0
are calculated using (8) as
follows:
s
1
t
0
s
3
t
0

126907:61
8549:80
3:147 MPa;
s
2
t
0

115841:99
8549:80
1:853 MPa:
Using (1012), the instantaneous stress and elasticity
modulus for the entire composite are as follows:
s
c
t
0

1
20:1 0:2
2:5 MPa;
E
c
t
0

8549:80
20:1 0:2
21374:5 MPa:
The values for the stress and elasticity modulus for the
three-layered wall fall between their values for the layers.
From the assumption of the perfect bond between the
layers, the strains from eq. (11) and those from (7) should be
equal. Thus:

1
t
0

3
t
0

3:147
26907:61
1:1710
4
;

c
t
0

1
8549:80
1:1710
4
:
Linear creep analysis. To perform the creep analysis, the
coefcient of aging should be known in advance. Thus, when

i
t; t
0
is determined, accounting for E
i
t E
i
t
0
, then the
analytical predictions of the stress-strain state can be based
only on the initial value of the elastic modulus, E
i
t
0
. This
relationship was proven in [10].
Using (1415), we can calculate the effective and the age-
adjusted effective elasticity moduli for i-th layer. Thus:
E
e;1
t E
e;3
t
26907:61
1 0:767
15227:85 MPa;
E
e;2
t
15841:99
1 2:147
5034:0 MPa;
E
e;1
t E
e;3
t
26907:61
1 0:0140:767
27199:68 MPa;
E
e;2
t
15841:99
1 0:3052:147
9573:154 MPa:
The composite strain developed at time t is determined
from formula (19) as follows:

c
t
23:1470:127199:68=15227:85 1:8530:29573:154=5034
227199:680:1 95730:2
;

c
t 2:486910
4
:
The composite strain obtained using the numerical tech-
nique,
c
t 0:002487(see Fig. 5d), is coincident with the
analytical prediction.
When
c
t is known, the layer stresses may also be
determined using (20) as follows:
s
1
t s
3
t 3:147 2:486910
4

3:147
15227:85
_ _
27199:68;
s
1
t s
3
t 4:290 MPa, (cf., 4.289 from Fig. 5c);
s
2
t 1:853 2:486910
4

1:853
5034
_ _
9573:154;
s
2
t 0:710 MPa, (cf., 0.711 from Fig. 5c).
These results show that the outer (less deformable) layer
is further compressed under sustained loading, while the
inner layer (more deformable) produces the relaxation from
the tensional stress increment developed with t.
The creep coefcient for the three-layered wall can be
predicted by relation (21) as follows:

c
t; t
0

2:4869
1:17
1 1:126, (cf. 1.126 from Fig. 5c).
Thus, the exact value of the coefcient of aging yields the
same values of the stress-strain state of the entire composite as
those produced by the numerical inversion of the Volterra
integral.
According to these results, negative values for the coef-
cient of aging occur if the time-dependent stress redistribu-
tion between the layers is governed by the combination of the
considerably different material aging and creep strains for
the case of the early age multi-layered composite.
For example, for reinforced concrete structures (where the
aging coefcient is extensively investigated), ; t
0
can
range primarily from 0.5 to 1. The code (EN 1992-1-1 2004)
suggests an approximate nal value of ; t
0
0:8 as a
suitable case for predicting pre-stress loses due to creep. In
this case, if the designer uses the concrete layered composite
value of 0:8 in the analysis, sufcient errors occur in
the prediction of the high strength layer stress because the
actual value
1;3
1; 0:014 (Fig. 5b) is very far from that
suggested by the code (EN 1992-1-1 2004).
Fig. 4 Schematic for the analysis of a three-layered
structure.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 483
In contrast, as a well-known rule, if the loading age exceeds
approximately 90 days, the aging coefcients of the layers tend
to ()0.8. Then, for most practical purposes requiring long
creep periods, tt
0
(exceeding 360 days), an approximate value of
the layers aging coefcient equal to 0.8 should be recommended
(see, (Fig. 5, b)) for predicting the stress-strain state of the layered
composite using the proposed analytical method. Thus,
0:8 may be implemented for the analytical time-
dependent analysis of a three-layered concrete structure for
t
0
90days.
6.2. Non-linear analysis
Consider the effect of non-linear creep on the time-
dependent stress-strain state for the above three-layered
structure. Now, suppose that this structure is made of layers
of middle/low/middle strength concrete. The C25/30 strength
class of concrete is used for the outer layers, and the inner
layer is constructed of low-strength C8/10 concrete. Assume
that the initial stress strength is
2
t
0
0:94 and

1;3
t
0
0:58for the inner and for the outer layers,
respectively. The initial stress strength selected is near the
instantaneous failure state for the low-strength (inner) layer.
These levels were held constant over all t
0
considered.
Therefore, the magnitude of the sustained load should
change to keep the assumed stress strength constant from
the instantaneous prediction, which illustrates the stress
evolution at different t
0
with unied initial stress conditions.
The stress-strength ratio pertains to the mean value of the
compressive cylinder strength.
For the baseline material properties (strength, elasticity
moduli, creep strains), the code EN 1992-1-1:2004 model was
used. Following NIIZHB [49] specications, the factor m4 is
recommended. The coefcient v
i
, accounting for the
increase in the creep strain during the delayed failure, is
computed as v
i
44:47=f
pr;i
28 , for f
pr;i
28 38 MPa; other-
wise, v
i
1:22 (where f
pr;i
28 is the compressive prism
strength). Thus, the mean value of the compressive prism
strength, which was used in the approach of [49], was
divided by a factor approximately equal to 0.95 to obtain
the mean value of the compressive cylinder strength. The
graphs dening the linear creep coefcient t; t
0
used in the
0
0.5
1
1.5
2
2.5
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0

(
t
,

t
0
)
2.147 (inner layer)
0.767 (outer layer)
1.126 (composite)
0.787 (outer layer)
2.203 (inner layer)
1.157 (composite)
1.973 (inner layer)
0.705 (outer layer)
1.040 (composite)
0.596 (outer layer)
1.667 (inner layer)
0.884 (composite)
1.163 (inner layer)
0.416 (outer layer)
0.626 (composite)
0.2
0
0.2
0.4
0.6
0.8
1
1.2

i
(
t
,

t
0
)
0.305 (inner layer)
0.014 (outer layer)
0.618 (inner layer)
0.532 (outer layer)
0.715 (inner layer)
0.666 (outer layer)
0.805 (inner layer)
0.780 (outer layer)
0.914 (outer layer)
0.920 (inner layer)
0.5
1
1.5
2
2.5
3
3.5
4
4.5

i
(
t
,

t
0
)
0.711 (inner layer)
4.289 (outer layer)
1.032 (inner layer)
3.968 (outer layer)
1.136 (inner layer)
3.864 (outer layer)
1.233 (inner layer)
3.767 (outer layer)
1.373 (inner layer)
3.627 (outer layer)
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
x 10
4

c
(
t
)
0.0002487
0.0001605
0.0001395
0.0001220
0.0001006
Fig. 5 Stress-strain state at time t (measured in days) for the entire composite loaded at different ages t
0
: (a) creep coefcient for
the layers (plotted by regular curves) and the entire composite (bold curves); (b) coefcient of aging of the layers; (c) layers stress
evolution with time; and (d) total strain evolution for the entire composite.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 484
current analysis for the inner and outer layers are shown in
Fig. 6.
In Fig. 7, the stress-strain evolution for each layer is
depicted. The results accounting for non-linear creep are
shown as a dashed line, and the values specifying the linear
creep prediction are shown as solid lines. The ages at which
the loading is imposed were t
0
2, 28, 180 and 740 days.
As shown in Fig. 7a, the nonlinear creep produces an
increase in stress relaxation for the more deformable layer
(the inner layer) compared to the linear creep behavior. The
stresses in the less-deformable layer determined by the non-
linear creep analysis are greater than those found in the
linear prediction.
In structural design, the decrease in a more-deformable
layers stress from non-linear creep may be considered a
certain reserve of the linear law. However, the increase in
stress of the less-deformable layer may be a dangerous state
that leads to a delayed failure of the layer. To this end, the
next section considers the long-term strength or a delayed
failure prediction.
6.3. Long-term strength and delayed failure analysis
The most important factor for the designer is to predict the
maximum load an entire multi-layered composite can with-
stand under a long period of sustained compression imposed
on the layered composite structure. This prediction is a
complicated task because the concrete materials have
cracked, i.e., the fracture is related to the crack propagation,
which is not completely brittle and usually possesses some
degree of ductility.
Models from fracture mechanics each claim partial suc-
cess when performing the instantaneous fracture analyses
(e.g., [51], [52], [53]). In general, when a closed form solution
does not exist, an FEM analysis can be adopted to study the
sophisticated failure mechanisms of composite materials,
including the material and geometrical non-linearities com-
bined with the complex composite geometries. The FEM
analysis of a compressive shear fracture test [54] has been
performed, focusing on the near crack-tip behavior of a two-
layered composite specimen with linear elastic isotropic
properties that contains a curved interface crack when its
surfaces come into contact and are able to dissipate energy
during the formation of new crack surfaces via the sliding
friction rule. By applying the loading process to the stiffer
layer with the weaker layer supported and vice versa, a
theoretical estimation of the mode II stress intensity factors
was obtained.
In concrete materials, the fracture process is complex; if
the action of creep strain is dominant over the process of
hydration in cement paste, the cracks start to propagate in a
stable manner. Depending on the stress level induced, the
cracks can interact with each other, and beyond a certain
period after loading, the crack propagation may form an
unstable pattern. In this case, the cracks coalesce, forming a
single continuous crack of a critical length within the layer. It
is also well known that a time-dependent decrease in the
compressive strength of the quasi-brittle materials occurs
when a non-linear creep strain dominates.
Wittmann and Zaitsev [55], [56] successfully determined
the decrease in the long-term strength of concrete under a
high sustained load. Relying on classical fracture mechanics,
they analytically found the following function, which evalu-
ates the state of the material under sustained loading:
t; t
0

E t
0
t
0

E t t

E t
0

~
E t

s t
s t
0

; 68
where t
0
=t Et=Et
0
f
c
t
0
=mt; t
0
f
c
t
2
is the ratio of
the effective surface energies at the time instant of initial
loading and the time of consideration;
~
E t is an operator that
10
0
10
1
10
2
10
3
10
4
10
5
0
0.5
1
1.5
2
2.5
3
t, t
0

(
t
,

t
0
)
2.567 (inner layer)
1.821 (outer layer)
2.205 (inner layer)
1.564 (outer layer)
1.665 (inner layer)
1.181 (outer layer)
1.297 (inner layer)
0.920 (outer layer)
Fig. 6 Creep coefcient of the layers (h
0
91 and 168 mm, RH95%) for the non-linear analysis.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 485
converts the instantaneous plane-stress state into the state
developed near a crack-tip under linear creep action;
m t; t
0
11:2 is the factor dening the increase in strength
at time t from the action of the preceding stress [56]; and
st=st
0
is the stress ratio manifesting the time variation of
the effective load near a crack-tip [56].
The rst term in function (68) evaluates the solidication
of the material due to the hydration of cement paste, the
second term accounts for the damage to the material due to
the growing cracks, and the last term denes the relaxation
or increase in stress near the crack-tip.
According to [56], an inversion of t; t
0
yields the ratio of
the materials long-term strength, f
c
t; t
0
, to its strength at
the loading time, f
c
t
0
; thus

u
t; t
0
t; t
0

1

f
c
t; t
0

f
c
t
0

: 69
We adopt this approach to predict the long-term strength
of the composite. Arutyunyans [9] theorem states that the
presence of linear creep does not change the instantaneous
equilibrium equations of the plane-stress state of the classi-
cal theory of elasticity when external loading is applied.
In this case, the creep strain only inuences the total strain.
This theorem allows us to account for the linear creep strain,

cr
i
t; t
0
, in eq. (68) by using the operator 1=
~
E t , which is
expressed from the total strain,
i
t
c
t , dened in eq. (4)
10
0
10
1
10
2
10
3
10
4
10
5
0
5
10
15
20
25
t, t
0

i
(
t
,

t
0
)
9.982 (outer layer)
10.434 (outer layer)
5.432 (inner layer)
4.980 (inner layer)
11.515 (inner layer)
10.219 (inner layer)
19.076 (outer layer)
20.372 (outer layer)
13.804 (inner layer)
12.005 (inner layer)
21.787 (outer layer)
23.586 (outer layer)
14.778 (inner layer)
12.748 (inner layer)
22.637 (outer layer)
24.667 (outer layer)
10
0
10
1
10
2
10
3
10
4
10
5
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
x 10
3
t, t
0

c
(
t
)
0.0011814
0.0012564
0.0017687
0.0013689
0.0016788
0.0012252
0.0015060 0.0015132
Fig. 7 Development of stress (a) and strain (b) in a three-layered composite structure for non-linear creep (dashed lines) and
linear creep (solid lines) behavior.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 486
for the i-th layer stress at t as follows:
1=
~
E
i
t

c
t
s
i
t

1
E
i
t

s
;i
t; t
0

s
i
t
C
n
i
t; t
0
: 70
Numerically, this relation may be expressed using eq. (23).
To include the effect of stress relaxation (or its increase) near
the crack-tip of each layer, the ratio of s t =s t
0
should be
computed from the non-linear solution of (47) because it simply
evaluates the effective load developed at time t (see [56]).
Following [55], [56], we can benecially introduce a
damage index for the i-th layer as follows:
M
i
t; t
0

i
t
0

i
t; t
0


i
t
0

u;i
t; t
0

71
A value of M
i
t
n
i
; t
0
_ _
1 means that the initial stress-
strength level,
i
t
0
s
i
t
0
=f
c;i
t
0
, causes the delayed failure
of the i-th layer at time t
n
i
, where
u;i
t; t
0
is the long-term
stress-strength ratio. The stress dened at the failure time t
n
i
is the long-term strength of the layer f
c
t; t
0
s
i
t
n
i
_ _
.
For M
i
t; t
0
o1, the applied stress-strength ratio
i
t
0
is too
low and cannot cause layer failure at any t.
The theoretical prediction of the long-term strength
according to eq. (68) and failure time t
n
for the i-th layer
depends on the aging and creep properties, the abilities of
cement hydration to heal the crack-tip and the ability to
redistribute the stress near the crack-tip. In particular, the
stress relaxation produces growth of
i
t; t
0
and, in turn,
growth in the long-term strength. Meanwhile, the stress
increase near the crack-tip results in the long-term strength
reduction.
Other more complicated techniques based on viscous-
elastic damage and rheological modeling may be found in
the literature for plain concrete. For example, a nonlinear
creep damage approach was implemented by Mazzotti and
Savoia [57] for plain concrete under uniaxial compression.
In their study, the creep strain was modeled using a modied
version of the solidication theory with a damage index
based on the positive strains. Recently, viscous-elastic
damage modeling has also been implemented by Verstrynge
et al. [19] for masonry structure analysis. A multiscale
modeling of early age concrete behavior was reported in [22].
In Fig. 8, a theoretical long-term strength prediction is
illustrated. The graphs shown in Fig. 8(a, b) are attributed to
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0

i
(
t
)
/
f
c
i
(
t
0
)

i
(
t
)
/
f
c
i
(
t
0
)
0.700 (outer layer)
1.387 (inner layer)
1.011 (outer layer)
0.688 (outer layer)
0.711 inner layer)
0.872 (inner layer)
0.634 (outer layer)
0.718 (inner layer)
0.852 (inner layer)
0.685 (outer layer)
0.614 (outer layer)
0.725 (inner layer)
0.878 (inner layer)
0.633 (outer layer)
Outer layers failure
0
0.5
1
1.5
2
2.5
0.466 (outer layer)
0.369 (inner layer)
1.138 (outer layer)
0.432 (outer layer)
0.734 (outer layer)
0.535 (inner layer)
1.208 (inner layer)
2.400 (inner layer)
0.707 (outer layer)
0.600 (inner layer)
1.089 (inner layer)
0.418 (outer layer)
0.719 (outer layer)
0.636 (inner layer)
1.076 (inner layer)
0.411 (outer layer)
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
10
0
10
1
10
2
10
3
10
4
10
5
t, t
0
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
M
i
(
t
,
t
0
)
M
i
(
t
,
t
0
)
0.496 (inner layer)
0.692 (outer layer)
1.085 (outer layer)
0.816 (inner layer)
1.115 (outer layer)
0.843 (inner layer)
1.077 (outer layer)
0.827 (inner layer)
Outer layers failure
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.154 (inner layer)
0.409 (outer layer)
0.443 (inner layer)
0.588 (outer layer)
0.551 (inner layer)
0.592 (outer layer)
0.591 (inner layer)
0.572 (outer layer)
Fig. 8 Function
i
t; t
0
, specifying the state of material under sustained loading and the damage index M
i
t; t
0
for each layer:
a-b) composite composed of middle/low/middle strength concrete; c-d) composite composed of high/low/high strength concrete.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 487
the structure composed of layers that possess middle/low/
middle strengths: f
cm;1
33 MPa, f
cm;2
16 MPa and
f
cm;3
33 MPa. The plots in Fig. 8(c, d) represent the long-
term strength prediction of the high/low/high layer strengths.
All material properties were given in the previous analysis.
The middle/low/middle strength composite structure
(Fig. 8a, b) was loaded using an initial stress-strength level at

2
t
0
0:94 and
1;3
t
0
0:58. This level was held constant for
all ages of t
0
considered. As shown in (Fig. 8a), the early age
structure (loaded at t
0
2 days) has an extensive ability for
stress redistribution over the layers from the development of
creep strain, aging and cement hydration. In particular, the
inner (low strength) layers stress-strength ratio,
2
t
0
0:94,
immediately falls to 0.8 after load application, and it nally
reduces to 0.688. In the interim, the outer layers stress-
strength levels increase from the initial value of
1;3
t
0
0:58
to
1;3
t 0:7.
The limit value for the long-term stress-strength ratio (dotted
bold lines in Fig. 8) should be used as a minimum for
u
t; t
0
(see,
[56]). Hence, the comparison of
u;1;3
t; t
0
0:940:7 and

u;2
t; t
0
140:688 (for t30000 and t
0
2 days in Fig. 8a) indi-
cates that long-term failure is not reached in all layers. The
destruction index, M
1;2;3
t; t
0
o1, more visually shows this con-
clusion for t
0
2 days in Fig. 8b. In additional analyses, when the
load was increased to almost instantaneous failure, i.e.,

2
t
0
0:98 and
1;3
t
0
0:62, the long-term failure of the outer
layer was not obtained in any layer because of the aforemen-
tioned stress redistribution.
The destruction factor, M
1;3
t; t
0
1 (horizontal line indi-
cating the failure), shown in Fig. 8(b) occurs for the outer
layers loaded at ages t
0
28, 180 and 740 days. The delayed
long-term failure of these layers occurs after a certain period
under sustained loading (intersection point of horizontal line
with function M
i
t; t
0
in Fig. 8b). Theoretically, the failure of
the outer layers occurs when the crack length increases to its
critical length, while the inner layer withstands the sustained
compression because its stress relieves throughout. The
practical conclusion from Fig. 8(b) is that the initial loading
level applied for all ages of t
0
28, 180 and 740 days is too high
and should be diminished to less than its ultimate values,

i
t
0

i
t
n

u;i
t; t
0
, to avoid long-term failure.
In the analysis of the high/low/high strength composite
(Fig. 8c,d), the initial stress-strength levels were introduced
near the instantaneous failure,
2
t
0
0:98, for the inner layer
and
1;3
t
0
0:33 for the outer layers. The results shown in
Fig. 8(c, d) illustrate that no long-term failure occurs (i.e.,
M
1;2;3
t; t
0
o1) at any t for any layer; consequently, no long
term failure occurs for the entire composite either.
Based on these results, after the nal period of the sus-
tained loading, when the delayed failure was avoided, the
applied load could be increased to cause the failure of the
composite structure. This is important when decisions regard-
ing the structure strengthening should be given after the
period of exploitation. The plots in Fig. 8(b, d) show that the
long-term strength of the composite can be greater than its
instantaneous value when M
1;2;3
t; t
0
o1. As shown, the initial
instantaneous load value after the nal period of action may
be increased by at least 1020% to reach M
i
t; t
0
1. This effect
can be explained in the following way. The low-strength layer
has a large creep strain, while the high-strength material is
characterized by a low creep strain (Fig. 5a). Thus, because of
the layers strain compatibility, the stress in the low-strength
layer allows the additional loading to be borne by the compo-
site after the period of sustained loading. This feature will be
conrmed below based on the experimental investigations
found in the literature. Additionally, a low creeping of the
high- strength concrete layers also governs the increased
presence of crack propagation in comparison with the low-
strength layer (cf., Fig. 8(a) and (c), using
u;1;3
t; t
0
).
6.4. The experimental data
To verify the theoretical prediction of the long-term strength
or delayed failure time of a multi-layered composite element,
experimental investigations are required. Unfortunately,
experimental tests involving long-term strength tests of
layered composites under sustained compressive loading
are not currently available in the literature. However, an
indirect qualitative verication can be discussed, relying on
some published experimental data.
Prokopovich and Zedgenidze [36] have summarized the
available experimental data on reinforced concrete beams
subjected to sustained bending and concluded that the
beams having a high reinforcement ratio (more than 3%)
and when subjected to a high sustained load (0.9 of the
instantaneous failure load), the beams did not collapse after
periods of 648, 1048 and 390 days. These beams were able
to withstand a portion (1.121.42% of the instantaneous
failure load) of the additional loading to reach the failure.
The authors related the increase in the load-carrying capacity
of the RC beams to the increased abilities of concrete to
relieve stress under high levels of sustained loading.
In the short-term experimental and numerical investiga-
tions of two-layered composite beams involving high-
performance concrete for the top layer and normal strength
reinforced concrete for the bottom layer, apko et al. [24] also
concluded that because of a redistribution of stress between
the layers, the composite beams withstood a higher load then
those made of the normal strength concrete without layers.
Experimental evidence demonstrating that plain high-
strength concrete is more prone to crack propagation and
exhibits increased long-term strength in comparison with
low-strength concrete was also previously reported by Iravani
and MacGregor [58] and Smadi et al. [59]. The latter authors
determined that
u
t; t
0
0:65 and
u
t; t
0
0:8 for concrete
with strengths of f
c
20 MPa and f
c
60 MPa, respectively. The
empirical relations found in the literature cannot capture the
effect of concrete strength on its long-term degradation.
These experimental results are an indirect qualitative
verication of the results presented in Section 9. In this case,
a relevant test series should be conducted in the future to
obtain a direct verication of the theoretical prediction of the
long-term strength, strain and the delayed failure time of
multi-layered composite elements.
7. Conclusions
Both linear and nonlinear creep models for predicting the time-
dependent behavior of a concrete composite were proposed.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 488
The presented one- and two-dimensional models use the
analytical and iterative analyses of the Volterra integral equa-
tion, implying the validity of the principle of superposition.
The analytical approach is based on the age-adjusted
effective modulus method. The model was validated through
a direct calculation of the time-dependent stresses and
strains for a three-layered composite wall structure, compar-
ing the theoretical values with those determined numerically.
In particular, the analytical approach yields identical values
for the time-dependent stress strain state parameters. It has
recently been found that negative values for the aging
coefcient can be obtained when the stress redistribution
between the layers is governed by a combination of consider-
ably different creep strains and aging of the layers for the
early age multi-layered composite.
Under the plane strain state, the two-dimensional creep
analysis of the multi-layered composite results in the same
vertical stress-time history as that in a one-dimensional case
if the Poisson ratios of the layers are equal. This result holds
even though an average value of the vertical stress to satisfy
the Volterra integral term is dependent on the Poisson ratio
of the layers. In particular, the evolution of vertical stress
with time is dependent only on the vertical strain and
compatibility conditions in a direction parallel to the
lamination.
The fracture mechanics approach has been adopted to
study the gradual degradation of the materials and the long-
term strength of the multi-layered composite under sus-
tained compression. Particularly, it was dened that the
long-term failure mainly starts from the layers having less
deformability because the more deformable layers can relieve
the initial stresses. In particular, the stress redistribution near
the crack-tips during the nal period of high sustained
loading may result in the need to apply additional compres-
sive stress to reach failure of the composite. Therefore, the
long-term strength of the composite may exceed its instan-
taneous strength for some cases, such as the early age
composite or the composite made of layers possessing con-
siderably different creep and aging properties. This feature is
important in the design of layered wall structures to ensure
the proper selection of materials for the layers.
The proposed analysis and methods provide a basis for
further investigations on the time-dependent behavior of a
concrete multi-layered compressive composite.
r e f e r e n c e s
[1] W.A. Woishnis, D.C. Wright, Select plastics to avoid product
failure, Advanced Materials and Processes (1994) 3940.
[2] H-G. Kwak, J-K. Kim, Nonlinear behavior of slender RC
columns (1), Numerical formulation, Construction and
Building Materials 20 (2006) 527537.
[3] I. Cypinas, Non-linear Deformation and Stability of
Reinforced Concrete Column under the Long-time Load,
Journal of Civil Engineering and Management (Statyba) 5 (3)
(1999) 176182.
[4] R. Baleviius, V. Simbirkin, Non-linear physical relations for
reinforced concrete elements under long-term loading,
International Journal of Applied Mechanics and Engineering
10 (1) (2005) 520.
[5] M. Mariukaitis, R. Baleviius, Stress, strain and creep
analysis of layered composite structures under sustained
loading, Engineering Structures and Technologies 1 (3) (2009)
123134 in Lithuanian.
[6] U. Pendharkar, S. Chaudharyb, A.K. Nagpal, Prediction of
moments in composite frames considering cracking and
time effects using neural network models, Structural
Engineering and Mechanics 39 (2) (2011) 2041.
[7] L. Jukneviius, G. Mariukaitis, J. Valivonis, Inuence of
technological factors on the state of stress and strain in
three-layer reinforced concrete structures, Journal of Civil
Engineering and Menagement 12 (3) (2006) 195204.
[8] L.S. Castillo, A. Aguado de Cea, Bi-layer diaphragm walls:
evolution of concrete-to-concrete bond strength at early
ages, Construction and Building Materials 31 (2012) 2937.
[9] N. Kh. Arutyunyan, Some Problems in the Theory of Creep,
Pergamon Press, Oxford, 1966.
[10] R. Baleviius, An average stress strain approach to creep
analysis of RC uncracked elements, Mechanics of Time-
dependent Materials 14 (2010) 6989.
[11] P. Majda, J. Skrodzewicz, A modied creep model of epoxy
adhesive at ambient temperature, International Journal of
Adhesion and Adhesives 29 (4) (2009) 396404.
[12] S.V. Aleksandrovskyi, Analysis of plain and reinforced
concrete structures for temperature and moisture effects
with account of creep, Stroiizdat, Moscow, 1973 In Russian.
[13] V.V. Vasiliev, E.V. Morozov, Mechanics and analysis of
composite materials, Elsevier, Amsterdam, 2001.
[14] R.M. Jones, Mechanics of composite materials, Taylor &
Francis, Philadelphia, 1999.
[15] O. Faber O, Plastic yield, shrinkage and other problems of
concrete and their effects on design, Minutes of Proc. of the
Inst. of Civil Engineers 225 (1) (1927) 2773.
[16] Z.P. Baant, Prediction of concrete creep effects using age-
adjusted effective modulus method, ACI Journal 69 (1972)
212217.
[17] Z.P. Baant, L.J. Najjar, Comparison of approximate linear
methods for concrete creep, Journal of the Structural
Division 99(ST9) (1973) 18511874.
[18] G.A. Chami, M. Theriault, K.W. Neale, Creep behaviour of
CFRP-strengthened reinforced concrete beams, Construction
and Building Materials 23 (2009) 16401652.
[19] E. Verstrynge, L. Schueremansa, D. Van Gemerta, M.A.N
Hendriks, Modelling and analysis of time-dependent
behaviour of historical masonry under high stress levels,
Engineering Structures 33 (2011) 210217.
[20] Z.P. Baant, S.S. Kim, Nonlinear creep of concrete
adaptation and ow, Journal of structural division 105(EM3)
(1979) 429446.
[21] F. Benboudjema, F. Meftah, J-M. Torrenti, G. Heining, A.
Sellier, A basic creep model for concrete subjected to
multiaxial loads, in: 4th International Conference on
Fracture Mechanics of Concrete and Concrete Structures,
2001.
[22] Ch. Pichler, R.A. Lackner, Multiscale Micromechanics Model
for Early-Age Basic Creep of Cement-Based Materials,
International Journal of Computers and Concrete 4 (2008)
295328.
[23] M. Yamada, H. Matsuura, H. Kumai, Hybrid reinforced
concrete beams by high-strength and normal-strength
concrete, in: Proceedings of the 5th Symposium on High
Performance and High Strength Concrete, 1999.
[24] A. apko, B. Sadowska-Buraczewska, A. Tomaszewicz,
Experimental and numerical analysis of exural composite
beams with partial use of HSC-HPC, Journal of Civil
Engineering and Management 11 (2) (2005) 115120.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 489
[25] B. Sadowska-Buraczewska, New generation concrete as a
strengthening layer in beam bending elements, Civil and
Environmental Engineering 2 (2011) 389392.
[26] Ch. Pichler, R. Lackner, H.A. Mang, Multiscale Model for
Autogenous Shrinkage of Early-Age Cement-Based Materials,
Engineering Fracture Mechanics 74 (2007) 34581-2 74 (2007)
3458.
[27] E. Julio, F. Branco, V.D. Silva, J. Lourenco, Inuence of added
concrete compressive strength on adhesion to an existing
concrete substrate, Build Environ 41 (2006) 19341939.
[28] W. Lorenc, E. Kubica, M. Kozuch, Testing procedures in
evaluation of resistance of innovative shear connection with
composite dowels, Archives of Civil and Mechanical
Engineering 10 (3) (2010) 5163.
[29] G. Mariukaitis, J. Valivonis, An analysis of the joint
operation of a CFRP concrete in exural elements, Mechanics
of Composite Materials 43 (2) (2007) 467478.
[30] D. Breton, J. Ballivy, J. Grandet, Contribution to the formation
mechanism of the transition zone between rock-cement
pastes, Cement and Concrete Research 23 (1993) 335346.
[31] G.A. Rac, B.K.B. Prasad, Inuence of the roughness of
aggregate surface on the interface bond strength, Cement
and Concrete Research 32 (2) (2002) 253257.
[32] D.P. Bentz, E.J. Garboczi, Simulation studies of effects of
mineral admixtures on the cement paste aggregate
interfacial zone, ACI Material Journal 88 (5) (1991) 518529.
[33] G. Mariukaitis, V. Popovas, Estimation of multi-layered wall
slenderness depending on behaviour of exible ties, Journal of
Civil Engineering and Management (Statyba) 5 (1) (1999) 310.
[34] D. Pawlus, Critical loads calculations of annular three-
layered plates with soft elastic or viscoelastic core, Archives
of Civil and Mechanical Engineering 11 (4) (2011) 9931009.
[35] D. Zabulionis, R. Baleviius, E. Dulinskas, Model for Time-
Dependent Stress and Strain Analysis of Layered Composite
Element, Solid State Phenomena 113 (2006) 571576.
[36] I.E. Prokopovich, V.A. Zedgenidze, Applied Theory of Creep,
Stroyizdat, Moscow, 1980 (In Russian).
[37] V.M. Bondarenko, S.V. Bondarenko, Engineering Methods of
Non-linear Theory of Reinforced Concrete, Stroyizdat,
Moscow, 1982 In Russian.
[38] Specications for evaluation of the concrete creep and shrinkage
in the analysis of concrete and reinforced concrete structures.
Concrete and Reinforced Concrete Ressearch and Technological
Institute, NIIZHB, Gosstroy, Moscow, 1988 (In Russian).
[39] R. Baleviius, D. Dulinskas, On the prediction of non-linear
creep strains, Journal of Civil Engineering and Management
16 (3) (2010) 382386.
[40] H. Trost, Auswirkungen des Superpositionsprinzips auf
Kriech- und Relaxationsprobleme bei Beton und Spannbeton,
Beton- und Stahlbetonbau 62(10): 230-238 62 (11) (1967)
261269.
[41] M . Jirasek, Z.P. Baant, Inelastic analysis of structures, Wiley,
Chichester, 2002.
[42] G. Mariukaitis, E. Dulinskas, The stress and strain state of
prestressed reinforced concrete members during steam
curing, Mokslas, Vilnius, 1974 In Russian.
[43] G. Mariukaitis, E. Dulinskas, The stress and strain state of
prestressed reinforced concrete members during steam
curing, Book review, ACI Journal 2 (1977) 220222.
[44] A. Kawano, R.F. Warner, Model formulation for numerical
creep calculations for concrete, Journal of Structural
Engineering 122 (3) (1996) 284290.
[45] S.V. Aleksandrovskyi, Creep and shrinkage of plain concrete
and reinforced concrete structures, Stroiizdat, Moscow, 1976
In Russian.
[46] S.A. Slobodianiuk, E.A. Yatsenko, Interaction of pre-stressed
reinforcement with concrete and creep analysis of one-
layered cellular concrete wall panels, Porogi,
Dnepropetrovsk, 2007 (in Russian).
[47] S.A. Slobodianiuk, A method of initial parameters with an
explicit time factor, in: VI Polish-Ukrainian (Ed.), Seminar
Theoretical Foundation of Civil Engineering, , 1998, pp.
547550.
[48] D. Zabulionis, A. Gailius, Numerical modelling of creep
functions of laminated composites, Mechanika 65 (3) (2007)
511.
[49] EN 1992-1-1:2004: Part1-1. European Committee for
Standardization. Eurocode 2. Design of concrete structures.
General rules and rules for buildings, CEN, Brussels; 2004.
[50] A. Ghali, R. Favre, M. Elbadry, Concrete structures: stresses
and deformations, Spon Press, London, 2002.
[51] M. Elices, J. Planas, Fracture Mechanics Parameters of
Concrete, Advanced Cement Based Materials 4 (1996)
116127.
[52] Z.P. Baant, Concrete fracture models: testing and practice,
Engineering Fracture mechanics 69 (2002) 165205.
[53] M. Kamiski, Computational mechanics of composite
materials, Springer-Verlag, London, 2003.
[54] . Figiel, M. Kamiski, B. Lauke, Analysis of a compression
shear fracture test for curved interfaces in layered
composites, Engineering Fracture Mechanics 71 (2004)
967980.
[55] F.H. Wittmann, Y.B. Zaitsev, Behaviour of hardened cement
paste and concrete under sustained load, in: Proceedings of
Symposium on Mechanical Behaviour of Materials, 4, 1971; p.
84-95.
[56] Y.B. Zaitsev, Simulation of strains and strength of concrete
by fracture mechanics, Stroyizdat, Moscow, 1982 In Russian.
[57] C. Mazzotti, M. Savoia, Nonlinear Creep Damage Model for
Concrete under Uniaxial Compression, Journal of
Engineering Mechanics 129 (9) (2003) 10651075.
[58] S. Iravani, J.G. MacGregor, Sustained load strength and short
term strain behaviour of highstrength concrete, ACI
Material Journal 95 (5) (1998) 636647.
[59] M.M. Smadi, F.O. Slate, H.N. Nilson, High-, medium- and low-
strength concretes subjected to sustained overloads
strains, strengths, and failure mechanisms, ACI Journal 84 (3)
(1985) 657664.
a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 3 ( 2 0 1 3 ) 4 7 2 4 9 0 490

Das könnte Ihnen auch gefallen