Вы находитесь на странице: 1из 9

Materials and Structures/Mat6riaux et Constructions,Vol.

30, April 1997, pp 139-147

Confined concrete model under cyclic load


j. Enrique Martlnez-Rueda 1, 2 and A.S. Elnashai 2
(1) On leavefrom Facultadde In qenien'a, UniversidadAutanoma del Estado de M&ico, Toluca, M{xico.
(2) Department of Civil Engineering, Imperial College, London, UK

ABSTRACT RI~SUMI~
A uniaxial cyclic model for confined and unconfined Un rnodFle cyclique uniaxial pour le b6ton confin6 et non
concrete is developed. Starting from an existing model, confin~ est d6velopp6. Sur la base d' un module existant, les r~gles
the cyclic rules to define inelastic strain and degradation cycliques qui d~finissent la d~formation inSlastique et la d~grada-
of strength and stiffness are first reviewed and then mod- tion de la rigidit8 et de la r&istance sont analys&s et ensuite
ified to produce a new concrete model with enhanced rnod~'~es afin de produire un nouveau module plus performant.
performance. Comparisons at the stress-strain level Des comparaisons entre le nouveau module et des modules exis-
between the new and existing concrete models reveal the rants au niveau de la contrainte-d~formation montrent l'ad~qua-
adequacy of the new model in terms of numerical stabil- tion du nouveau module en ce qui concerne la stabilitd nurn&ique
~," , I . . . . I I ,
ity and increasing degradation of strength and stiffness et l evolutlon de la degradatlon de la rt2dtte et de la reststance
under increasing cyclic strain. The proposed model is sous d~formation cyclique. Le module propos~ est validd par com-
validated by comparing analytical predictions with paraison des pr&ision analytiques avec des r&ultats expMmen-
experimental results of reinforced concrete members taux sur des ~l&nents de b&on arm~ soumis a des charges
under cyclic and dynamic loading. Good agreement is cycliques et dynamiques. Une bonne concordance est observ&
observed between analysis and experiments, confirming entre l'analyse et les expMences, confirmant ainsi la capacit~ du
the ability of the model to predict the cyclic and module h prMire le cornporternent cyclique et dynamique des ~l~-
dynamic behaviour of reinforced concrete members with ments en b&on arm~ soumis a l'action combin& des efforts
mixed axial-flexural response characteristics. axiaux et deflexion.

1. INTRODUCTION ling of critical (plastic hinge) zones up to very large


deformations is essential. This imposes extremely taxing
The assessment of the behaviour of engineering requirements on analytical investigation tools, both on
structures under complex loading requires the utilisation the local and the global structural levels. In reinforced
of both experimental and analytical investigation tools. concrete analysis, the subject of this paper, the funda-
Whilst final confirmation of new concepts and mental requirements for use of an analysis model in
approaches to design require experimental verification, earthquake applications are as follows:
the role of analysis is central to the process of reaching - Provide accurate estimates of stiffness and strength
the point of confirmation. This is mainly due to the under cyclic loading at aW strain level. This is necessary
impracticality of investigating structural performance in for accurate evaluation of lateral response and damage
a parametric fashion using labour- and capital-intensive accumulation modelling.
testing procedures, especially under dynamic loading. - Exhibit stability and accuracy at very high strain levels.
Therefore, development of advanced analytical model- This is required for the calculation of the collapse load of
ling techniques and highly controlled testing methods the overall structure, which may correspond to one or
should go in tandem. Whereas analytical procedures for more zones reaching very high levels of deformation.
the study of structures and components in the vicinity of - Account accurately for the effect of confinement. This
their yield limit state have reached a satisfactory level of is required for the distinction between detailing of criti-
development, accurate analysis up to structural collapse cal and non-critical zones.
is still an active research area, particularly in earthquake The above characteristics have been the subject of
engineering, as discussed further herein. extensive research in the past two decades or more, as
To enable the evaluation of ductility demand, in reflected by the number of research publications on con-
order to provide the necessary supply, accurate model- stitutive relationships for concrete [1]. However, models

0025-5432/97 9 PdLEM 139


Materials and Structures/Mat&iaux et Constructions, Vol. 30, April 1997

applicable to variable amplitude cyclic loading, hence 3.1 Monotonic or envelope curve
dynamic loading, are rather scarce. In this paper, an
existing model [2] for confined concrete under cyclic For monotonic loading, the compressive concrete
loading is reviewed and modifications are introduced in stress fc is given by the Popovics curve [5] expressed as
order to develop a new concrete model with enhanced
L- fc'xr (1)
performance.
r-l+x"
where the parameters of equation (1) are defined by
equations (2) through (7) as
2. CONCRETE BEHAVIOUR UNDER
t _
CYCLIC LOADING f cc kf'co (2)
x : ec/8cc (3)
Uniaxial tests carried out on concrete short columns
have allowed the observation of strength and stiffness
degradation of concrete due to cyclic loading.
Experimental results [3, 4] have shown that the envelope
of the stress-strain relationships of concrete under cyclic r- E~ (5)
load is identical with the stress-strain curve obtained under E~ -E~o~
constantly increasing strain. It was also observed that the
degradation of strength is characterised by a locus of com- Er = 5000, f/~-o (6)
mon points defined by the intersection between loading
and reloading branches. Fig. 1 shows a typical stress-strain E~oc----
f~'~ (7)
curve for concrete under cyclic loading. gcc
where k is the confinement factor, f'co is the compres-
sive strength of unconfined concrete, f'cc is the compres-
fc'
sive strength (peak stress) of confined concrete, 8~ is the
longitudinal concrete strain, Eco is the strain at uncon-
fined stress f~o, e~ is the strain at maximum concrete
stress f'~c, E~ is the initial modulus of elasticity of con-
crete and Esec is the secant modulus of elasticity of con-
crete at peak stress.

3.2 Inelastic strain

Irrecoverable or inelastic9 strain 8_P...1 corresponds , to


zero stress on the compressive unloading or reloading
branches and is defined as
Fig. 1 - Typical stress-strain curve for concrete under cyclic
loading. (8un+ 8a)fu. (8)
gpl = gun fun + E c e a
From the analytical point of view, a particular con- where
crete model for cyclic loading is characterised by the way
the degradation of strength and stiffness are modelled.
g~ = a ~ (9)
This is achieved by adopting specific criteria in the for-
mulation of the monotonic or envelope curve, the
unloading and reloading branches and the rules to define
a=Maxl [gcc +gun
.09Gn }
inelastic strain and strength degradation. 8un is the reversal envelope strain (also referred to as the
maximum concrete strain ever experienced ), fu, is the
reversal envelope stress and G is the focal strain defined
3. THE MODEL OF MANDER, PRIESTLEY in Fig. 3.
AND PARK
3.3 Compressiveunloadingand reloading
Mander et al. [2] have proposed a unified stress-strain
model for confined concrete members subjected to uni-
branches
axial cyclic loading. This model follows the approach of
Fig.1 shows the unloading and reloading branches.
the model of Karsan and Jirsa [4], but in a simplified
The unloading branches are defined by
way, with a different envelope curve and with some
refinements to account for confinement effects. fc = fu~ funXr (11)
r-l+x r

140
Martinez-Rueda,Elnashai

where where
gc -- gun x = gc - ~re (22)
x = - - (12)
gpl -- gun
fun -- fnew
Eu 2 +/ g
g r e = g u n " [ - f cr ( @ (23)
r = Eu_Es~ c (13) /. fco)

Er -Ere
gsec _ fun (14)
gun -- gpl A= _4[(fnew _ fre) _ E r (gun _ ere)] (24)
E u = bcE c (15)
Ere is the tangent modulus at the returning point
b = fu~ > 1 (16) (~re, fre) in the monotonic stress-strain curve.
f&

c= ~ / ~ <1 (17)
3.4 Tensile unloading branches
gun
The deterioration rule for tensile strength due to pre-
where fun is the reversal (unloading) compressive con-
vious compressive strain histories assumes that upon
crete stress and E u is the tangent modulus at the begin-
unloading from a compressive branch, the tensile con-
ning of the unloading branch. Coefficients a, b and c in
crete stress becomes:
equations (10), (16) and (17) respectively were evaluated
by trial and error to produce the "best fit" of equation
(11) to experimental curves. ,t :,(1 /25/
The reloading branches are defined by a linear rela-
!
tionship followed by a parabolic transition. The linear where f t is the tensile concrete strength.
relationship is given by If gpl > gcc then ft = 0 and the stress-strain relation
becomes:
fc = fro + E,-(ec - ero) (18)
where ft = Et(ec - gpl) (26)
where
E r _ fro - fnew (19)
gro -- gun E, =--ft (27)
gt
fnew = 0.92 fun + 0.08 fro (20)
(r fro) are the co-ordinates of the point of reloading ft
t

G = E-7 (28)
located either in an unloading branch or in a cracked
state, as s h o w n in Fig.2. T h e degraded stress f, ew When the tensile strength is9 exceeded, i' e' e C > (et - ep1),
accounts for cyclic degradation and was obtained using cracks open and zero tensile strength is assumed for subse-
the same experimental results used to calibrate the quent loadings.
unloading branches.

3.5 Review of the performance of the model


of Mander et al.
A realistic concrete model must show increasing
degradation of strength and stiffness due to cyclic effects.
This degradation depends primarily on the definition of

/tII II// k <


.'ll "xo f_)
the inelastic strain and the rule for cyclic degradation of
strength. A detailed review of the d e g r a d a t i o n of
strength and stiffness of the model of Mander et al. [2]
V/7 reveals the following weak points:
1. The rule to degrade strength considers a uniform degree
V %1 s
of strength decay regardless of the level of strain. At full
Fig. 2 - Stress-strain curves for unloading and reloading branches reversal, i.e. when ~o = 0, equation (20) predicts a strength
in the model ofMander et al. degradation ratio fuJ~ew equal to 0.92. However, experi-
mental evidence [4] indicates that both strength and stiff-
The parabolic transition between the point of degraded ness degradation are a function of the accumulated damage
stress (gun, fnow)and the returning point (ere, f~e)is given as: and hence the strength degradation ratio is not constant
but depends on the magnitude of the reversal strain.
fc = fre + treX + Ax2 (21) 2. The rule to define inelastic strain is inadequate, ben-

141
Materials and Structures/Mat6riaux et Constructions,Vol.30, April1997

erating increasing stiffness for large strains. This is dis-


cussed further in section 5.
3. The returning strain 8~e defined by equation (23) in fc ( ~ ' ~ ~
some cases is smaller than a returning strain defined by
the intersection of the extension of the initial linear
reloading branch and the envelope. This results in a local
and unrealistic gain of stiffness due to cyclic effects.
Additionally, the parabolic transition defined by equa-
tion (21) in some cases overpasses the envelope.
4. The initial stiff ess of the unloading branches predicted
by equation (15) results in very high values of the tangent
modulus. Experimental studies [3, 4, 6] show" that equa-
1
tion (15) overestimates the initial unloading stiffness.

'e ff)
4. IMPLEMENTATION OF A NEW
Fig. 3 - Stress-straincurves for unloading and reloading branches
CONCRETE MODEL in the proposed new model.
The lack of numerical stability of the model of
Mander et al. [2], in particular with respect to stiffness, For the high strain range the rule to define .8_ I o1 was
may lead to convergence problems under large displace- adapted from the model of Yankelevsky and R.elnhardt
ments when implemented into a non-linear program [7], as shown in Fig. 3. In the new proposed rule, the
following a fiber element approach. focal point (gf, re) is defined based on the upper limit of
To overcome the weak points of the model of the intermediate strain range. This allows a continuous
Mander et al. [2], a new concrete model which predicts transition between the intermediate and high strain
continuing cyclic degradation of strength and stiffness ranges. Accordingly, the co-ordinates of the focal point
was developed. The new model follows the simplified are given as
approach of the model ofMander et al. [2], but the rules
for cyclic degradation of strength, inelastic strain and the fcl'gplcr
shape of the unloading branches are different. [e,.I= Ec(ec,.%,cr)_ fc, (32)

Iffl = Be I~cl (33)


4.1 Inelastic strain
8~r - 2.58~ (34)
Three different rules are proposed to define inelastic where Eplcr is the inelastic strain corresponding to the
strain; they reflect the level of damage achieved and upper limit of the intermediate strain range Ecr.
depend on the maximum strain ever experienced accord- Until detailed experimental studies on cyclic behav-
ing to equations (29) through (31), which correspond to iour of concrete for different strain ranges become avail-
low, intermediate and high strain range, respectively. able, it is proposed that equations (29) through (31) pro-
vide a reasonable estimate Of~pl consistent with the level
gpl DR gun -- fu~
E--
7 , 0 -< Gun <- e35 (29) of damage experienced by the concrete structure.

Gun -F ga
gpl =eu. fun +E~g, , g35-<e~,,-<2.58c~ (30) 4.2 Unloading branches
The adopted unloading curves shown in Fig. 3 are
fc,.o-I<
8Pl- fc,-+fu, , 2.58c~_<8u~ (31)
second degree parabolas joining the strain reversal point
(gun, fun) with the current full reversal point (~pl, 0) and
The inelastic strain at low strain range is obtained on with zero slope at (8pl, 0), namely
the assumption of elastoplastic behaviour. This is numeri-
~c -- gpl
cally convement since when the maximum concrete strain fc = fu, (35)
is smaller than 835 (a strain correspondent to a stress of
0.35 f'c in the ascending branch) the cyclic behaviour of The strain reversal point (~un, t~m)may belong either
concrete is essentially elastoplastic, eliminating the need for to a reloading branch or to the envelope. If strain reversal
an elaborate model within this strain region. occurs from a reloading branch, then the current level of
In the intermediate strain range, 8_F1 is defined using inelastic strain 8171is still used. The inelastic strain 8pl }s
the formula of the model of Mander et al. [2], which updated every time the maximum strain ever experi-
agrees well with experimental results obtained by Karsan enced is updated. Therefore, this updating procedure
and Jirsa [4]. The upper limit of this range was chosen as may take place only either in the envelope or in the sec-
the highest strain achieved in this experimental work. ond reloading branch described below.

142
Martinez-Rueda, Elnashai

4.3 Loadingand reloading branches in terms ofa normalised strain and may be used to repre-
sent both confined and unconfined concrete. Therefore,
Fig. 3 shows the loading and reloading branches used since the mechanisms of damage are represented in the
in the proposed model. For strains smaller than the max- envelope, comparisons for unconfined concrete can be
imum strain ever experienced eun, a straight line is fitted directly extended to confined concrete and allow the use
between the reloading point (ero, fro) and the degrading of the extensive experimental results reflected in the
strength point (gun, fnew)' The degraded stress fnew is model of Karsan and Jirsa [4].
defined following an approach similar to the model of
Karsan and Jirsa [4]. It is additionally assumed that fnew
lies in the curve of common points and therefore the 5.1 Reloadingstiffness
intersection between unloading and reloading branches
occurs at a strain equal to the maximum strain ever Fig. 4 shows a comparison of stiffness degradation
experienced. The curve of common points is then predictions for the case of unconfined concrete. The
defined by a modified version of the Popovics equation stiffness considered for comparison is the reloading
[5], namely secant stiffness Esrc between the full reversal point (epl , 0)
and the common point (enew,fnew),namely
fnew -- fcc2 X r (36)
r-l+x r g src = . fnew (44)
Ene w -- I~pl
where
fcc2 = 0.9 f'cc (37)
1.25
x= s~ (38)
cc2 1.00 Karsan and lirsa
Mander
ecc 2 = 0.9gcc (39) E ~ 0.75 Proposed Model
For strains larger than eun, a straight line is fitted
between the degrading strength point (%,, fnew) and a Ec 0.50
returning point (ere, fre)" To compensate for the lack of a
smooth transition between the reloading branches and 0.25
i
the envelope the returning strain ere is set to the average
0.00
value between eun and the returning strain e're obtained f
using the empirical equations ofKarsan andJirsa [4]. -0.25 I I I m I' , I , I
0 2 4 6 8 10
~;~t + 8.. (40)
s 2 E~

Fig. 4 - Comparison o f secant reloading stiffness at common


where e're : Sreu,~ (41) point for different concrete models.
Sr = 0. 00273 + 1.2651S e (42)
Se = u_~
S (43)
IF-cc
It is evident from Fig. 4 that in the low strain range the
three models differ from each other, the model of Karsan
in which Sr is the returning strain ratio and So is the and Jirsa [4] being the most conservative. However, while
unloading strain ratio. the model ofKarsan andJirsa [4] and the model of Mander
et al. [2] tend to overestimate and underestimate Esrc,
respectively, the proposed model predicts an expected null
5. PERFORMANCE OF THE PROPOSED degradation of stiffness at zero strain.
AND EXISTING MODELS In the intermediate strain region there is no signifi-
cant difference among the three models. For the high
In this section, the models ofMander et al. [2], strain region, both the model of Karsan and Jirsa [4] and
Karsan and Jirsa 141 and the proposed model are com- the model of Mander et al. [21 fail to predict increasing
pared to assess the adequacy of the models in predicting stiffness degradation under increasing cyclic strain. In
cyclic stiffness and strength degradation. Due to the dif- contrast, the proposed new model predicts increasing
ferent monotonic curve used in the model of Karsan and stiffness degradation under increasing cyclic strain for
Jirsa [4], a direct comparison of this model is not possi- any level of strain. The model of Mander et al. [2] not
ble. However, the general trend exhibited by this model only stops degrading stiffi~ess but slightly increases stiff-
provides guidance about the expected cyclic behaviour ness very early in the high strain range at about a cyclic
in the low and intermediate strain ranges. strain equal to 38cc The model of Karsan and Jirsa [4]
The model of Mander et al. [2] and the proposed continues to predict degradation of stiffness for strains
model use the Popovics curve [5] to represent the enve- well in excess of the lower limit of the high strain range.
lope. As shown by equation (1), this curve is expressed However, the inherent rule to degrade stiffness degener-

143
Materials and Structures/Mat6riaux et Constructions, Vol. 30, April 1997

ates at about 5~cc and predicts an asymptotic increase of


stiffness. This may be attributed to the fact that this 1.0
__ Monotonic
model was not calibrated for the large strain region. ." ". -- Mander
The findings discussed in the previous paragraph ' " - "- 1
0.8
pinpoint the importance of using different rules to
define the inelastic strain according to the level of dam-
age associated with the maximum strain ever experi- fc___0"6
enced. It is also clear that the model of Karsan and Jirsa
fo~ 0 . 4
[4], in its present form, cannot be extended to model
confined concrete for strains in excess of 5acc.
0.2

5.2 Unloading stiffness 0.0 I J I , I I I I I

0.0 0.5 l.O 1.5 2.0 2.5 3.0


Fig. 5 shows a comparison of stiffness degradation pre- Ec
dicted by different concrete models for the case of uncon-
fined concrete. The stiffness considered for comparison is
Fig. 6 - Comparison between the monotonic curve and the
the unloading secant stiffness Esu between the unloading envelope of degraded stress at full reversal predicted by the
point (~un,fun) and the full reversal point (r 0), namely model ofMander et al. and the proposed new model.

Esu - fu. (45)


I~un -- Epl damage regardless of the maximum strain ever experi-
enced. Experimental evidence [4] reveals that this
strength degradation rule appears to be conservative in
1.25
the low strain range and unconservative in the interme-
Karsan and Jirsa diate and high strain ranges. The inadequacy of this
1.00
Mander strength degradation rule is in part responsible for the
0.75 Proposed Model degeneration of the stiffness degradation rule of the
E= model for reloading.
"~'c 0.50 As expected, the proposed model approximates the
trend observed in experimental results reported by
0.25 Karsan and Jirsa [4], and hence a more realistic cyclic
strength degradation is observed.
0.00
e
I i I, i I
-0.25
0 2 4 6 8 lO 6. VALIDATION OF THE PROPOSED
Ec
MODEL
Eoe

Fig. 5 - Comparison of unloading secant stiffness predicted by The proposed model has been implemented in the
different concrete models. finite element analysis program ADAPTIC [8, 9]. The
program is capable of predicting the large displacement
behaviour of 2D and 3D structures under static and
dynamic loads, taking into account geometric nonlinear-
It is evident from Fig. 5 that the observed perfor- ities and the effect of material inelasticity. Perfect bond
mance of the models is similar to the case of reloading. between concrete and steel is assumed in the evaluation
However for the case of unloading, the model of Mander of the transverse section response.
et al. [2] coincides with the proposed model in the low To validate the proposed concrete model, the experi-
and intermediate strain regions and tends to predict null mental response of reinforced concrete members is com-
degradation of stiffness at zero strain. This was expected pared with analytical results obtained with ADAPTIC.
since for unloading the unrealistic rule to degrade The specimens used for comparison are consistent with
strength in the model ofMander et al. [2] does not affect the behaviour assumed in the finite element formula-
Esu which depends mainly on the rule for inelastic strain. tion. Hence test results of specimens with response con-
trolled mainly by a combination of flexure and axial load
are used as a benchmark.
5.3 Strength degradation
Fig. 6 compares the degraded stress due to full strain 6.1 Validation under cyclic loading
reversal predicted by the model of Mander et al. [2] and
the proposed model. As predicted by equation (20) the Tests conducted on four nearly full-size reinforced
model ofMander et al. [2] considers a uniform degree of concrete columns with various quantities of transverse

144
Martinez-Rueda, Elnashai

In general, the analysis overestimates the strength of


P = 1815 KN the column at low displacements and the difference in
r ~ HI2 L L -'//Sh response decreases for larger displacements. At maxi-
Reaction mum displacement the analysis underestimates the
~h~ c . F . E mesh
strength of the specimen by less than 5%.
I 1A The analytical model is unable to predict the pinch-
B Lateral load
ing observed in the experimental results. Some other
d
s ~ "q~"4~ H ~ hffi550mm" I minor effects (in this case) not considered in the model,
including distribution and layout of inclined cracking,
~=~5~,, I~ ,,,'"~ bond slip and shear distortion at the plastic hinge region,
A A I~ 12.24 " E
| may explain some of the differences between analysis
Reaction I le and experiment. Nevertheless, within reasonable limits,
it has been proposed [12] that variations in the shape of
~P =AmLle,d hysteresis loops of RC frame structures do not have a
and dimensions b. Materials and transverse section major influence on the inelastic dynamic response under
severe earthquake excitation.
Fig. 7 - Test specimen elevation, transverse section, material
properties, method of loading and F.E. model.
6.2 Validation under dynamic loading
reinforcement have been described by Park et al. [10]. Dynamic tests on a medium-scale reinforced con-
From this experimental work the test specimen referred crete column have been described by Gauvin et al. [13]
to as "unit 1" was selected as a benchmark to compare and have been used in the past to validate a P,,C element
analytical predictions. Fig. 7 shows the test specimen for nonlinear analysis [14]. The characteristics of the
elevation, transverse sections, material properties and specimen of this experimental programme are compati-
method of loading. ble with the assumed element behaviour and hence these
The F.E. model used to represent the test specimen test results are a good benchmark to compare the perfor-
(Fig.7.c) is an assembly of four cubic inelastic elements. mance of the proposed concrete model under random
Convergence studies showed there was no need to refine strain history induced by seismic loading.
the mesh further. The transverse section was divided Fig. 9 shows the test specimen elevation, transverse
into 75 fibers and the confinement factors for the core section, material properties and method of loading. The
were estimated as 1.4 and 1.2 for the hoop spacings of 80 specimen was tested under simulated earthquake loading
and 135 mln respectively. The longitudinal reinforce- imposed by a shaking table. Eight tests using the same
ment was modelled using a multisurface steel model record scaled to different PGA of increasing magnitude
[11]. The parameters for the steel model were estimated were conducted. Fig. 10 shows the acceleration record
by scaling the multisurface parameters for typical high used, normalised with respect to PGA. Table 1 shows a
strength steel bars with fy = 400 MPa [11]. summary of experimental results in terms of top peak
Fig. 8 shows a comparison between experimental and displacement.
analytical results. The same trend of both strength and The F.E. model used to represent the specimen is an
stiffness degradation and energy absorption is observed, assembly of four cubic inelastic elements [Fig. 9c].
and hence good representation of parameters of dynamic
response is provided by the analytical model.

c. F.E. mesh

-
-1~00 '
980
' '
-20
" '

Horizontal displacement at benen stub [mm]

Fig. 8 - Comparison o f experimental results and analytical


. .
,

a. loading and dimensions


'"l l-
9

9
12~8
fc =42 MPa
fy=445 MPa
9

9
h

b. Materials and trar~erse section


Fig. 9 - Test specimen elevation, transverse section, material
predictions for a column under cyclic loading. properties and method of loading for dynamic tests.

145
Materials and Structures/Mat~riaux et Constructions,Vol,3 0 , April1997

Table1 - Summaryofshakingtabletest s
Test
resultsreportedbyGauvinetal.[13]
PGA Maximum Maximum Damping
~IiIIIi6
Displacement Drift ~ ~
No [g] [mm] [%] [%]
1 0.265 04.46 0.31 3,1
2 0.418 13.25 0.92 3,7 ~0
3 0.530 21.46 1.49 4.0
4 01724 28122 1196 4.2 ~;2l"
5
6
0"836
1.000
38102
46.80
2'3.25
64 4"6
4.6
{ -4 ~ IIIIVll
IIIIII

7
8 1.3 5 65.95 4.58 5.4 i-6I 11' ;"

4 6 8 10
1.0
Tune

Fig. 11 - Comparison between analysis and experiment for


O.5 shaking table test No 3.

9o AL * A[ l[ ] J~ 1~. J Fig.// shows a comparison between analytical and


experimental results for the case of the third test, in
which PGA was equal to 0.53g. The comparison is pre-
sented in terms of the lateral displacement at column
mid-height. The same trend of lateral response is
Z -i.0 ! observed in the analytical and experimental results and
0.0 O,5 I.O 1,3 ZO 2.3 3.0 3.3
good agreement is obtained with respect to the number
Tkne [se~l of displacement reversals, frequency of vibration and
Fig. 10 - Acceleration record used in the shaking table tests. location of peaks. Due to the limitations in the model-
ling of viscous damping, a higher difference between
analytical and experimental results occurs after the 3.5
Convergence studies showed there was no need to refine seconds of ground motion. From a practical point of
the mesh further. The transverse section was divided view, the modelling of the transient response is more
into 75 fibers and a uniform confinement factor equal to i m p o r t a n t than the free vibration response.
1.15 was estimated for the confined core. As in the case Nevertheless, the analytical predictions of the free vibra-
of the validation under static loading, a multisurface steel tion response show similar frequency and the same ten-
model with estimated parameters was used for the longi- dency with respect to residual displacement.
tudinal reinforcement. Viscous damping proportional to Similar comparisons were established for the first six
stiffness was included in the model. As shown in tests, also showing good agreement between analytical
Table 1, experimental results indicate that viscous damp- and experimental results. Fig.12 shows a global compari-
ing increases as the specimen is subjected to higher levels son between experimental and analytical results in terms
of damage, i.e. higher peak displacement. From these of peak top displacement. In general, it is observed that
results, viscous damping equal to 4% was considered in
the modelling of the first three tests and 5% when mod-
elling the eight tests together. 100 I -

To account for the effect of accumulated damage due 90t-

to previous excitations, the test specimen was analysed


under earthquake loading using a cumulative accelero- 0
701-

gram created by assembling the individual test accelero- E 401-

grams one after another. Between each of the scaled e~


301-

accelerograms a lapse of 6.5 seconds of zero acceleration 401-

was included to allow the free vibration response to 301"

decay to a low displacement level close to the expected 201-


e~t
residual displacement. g io I-
--0- Analysk
A preliminary analysis showed that the yield displace- 01-
, I , I , I i I
ment of the specimen was about 14 mm. Therefore, the 0 2 4 6 8

third test, in which the maximum top displacement was Test number
21.6 mm, is the first test in which both concrete and Fig. 12 - Global comparison of results between the analytical
steel exhibit significant non-linear behaviour. model and the shaking table tests.

146
Martinez-Rueda, Elnashai

the analytical model slightly overestimates the top dis- The first author gratefully acknowledges the support
placement for the first six tests, which cover a wide provided by CONACYT-Mdxico and the Facultad de
range of lateral drifts as high as 3.25%. For tests seven Ingenierfa - Universidad A u t 6 n o m a del Estado de
and eight, higher differences between analysis and exper- M&ico.
iment are observed. These may be attributed to cumula-
tive errors in the estimation of residual displacements
between consecutive tests.
Nevertheless, within a reasonable range of displace-
ments, the analytical predictions are quite accurate, con-
firming the ability of the analytical model to predict the
dynamic non-linear response of R C elements with REFERENCES
behaviour dominated by flexure and axial force.
[1] Comit~ Euro-International du B&on, 'Response of R.C. critical
regions under large amplitude reversed actions', Bulletin
D'Information ]61 (CEB, Laussane,1983).
7. CONCLUSIONS [2] Mander, J.B., Priestley, J.N., and Park, R.,'Theoretical stress-
strain model for confined concrete', ASCEI. Struct. Div. 114
Starting from an existing concrete model, a new (12) (1989) 1804-1826.
cyclic stress-strain model for both confined and uncon- [3] Sinha, B.P., Gerstle, K. H. and Tulin, L.G. 'Stress-strain relation
for concrete under cyclic loading',/. Amer. Concr. Inst. 61 (2)
fined concrete has been introduced. The model has been (1964) 195-211.
implemented into a sophisticated non-linear analysis [4] Karsan, I.D., and Jirsa, J.O. 'Behaviour of concrete under com-
program adopting a fiber element approach. Based on pressive loadings',-/. Struct. Div. 95 (12) (1969) 2543-2563.
the comparisons performed at stress-strain level and at [5] Popovics, S. 'A numerical approach to the complete stress-strain
member level under static and dynamic conditions, the curves for concrete', Cem. and Concr. Res. 3 (5) (1973) 583-500.
[6] Desayi, P., Sundara Raja Iyengar, K.T. and Sanjeeva Red@, T.
following conclusions were drawn: 'Stress-strain characteristics of concrete confined in steel spirals
1. The proposed concrete model exhibits unconditional under repeated loading', Mater. Strua. 12 (71) (1979) 375-383.
numerical stability and predicts increasing strength and [7] Yankelcvsky, D.Z. and Reinhardt, H.W. 'Model for cyclic com-
stiffness degradation under cyclic loading for any level of pressive behaviour of concrete', ASCE-/. Struct. Eng. 113 (2)
(1987) 228-239.
strain. [8] Izzudin, B.A. 'Nonlinear dynamic analysis of framed structures',
2. The proposed concrete model, in conjunction with PhD. Thesis, Imperial College, London (1991)
advanced non-linear analysis techniques, is able to pro- [9] Izzudin, B.A., Karayannis, C.G. and Elnashai, A.S. 'Advanced
vide a good estimation of the cyclic response of R C non-linear fornmlation for reinforced concrete beam-colmnns',
members with behaviour dominated by flexure and axial ASCE.]. Struct. En,q~. 120 (10) (1994) 2913-2934.
force under static and dynamic conditions. [10] Park, R., Priestley, J.N. and Gill, W.D. 'Ductility of square-
confined concrete columns', ASCEJ. Struct. Div. 108 (4) (1982)
3. The study conducted on concrete models at stress- 929-950.
strain level lead to the development of a new model with [11] Petersson, H. and Popov, E.P., 'Constitutive relations for gener-
continuous degradation of both strength and stiffiless. alized loadings', ASCEJ. En~. Mech. Div. 103 (4) (1977) 611-
This methodology may be adopted as a standard proce- 627.
[12] Park, R. 'Evaluation of ductility of structures and structural
dure to verify new concrete models before their imple-
assemblages from laboratory testing', Bull. N. Z. National Soc..for
mentation into a non-linear program. Earth. Engf. 22 (3) (1989) 155-165.
[13] Gauvin, J., Jeandidier, C., Gaubert, J.C., Quevat, J.c. and
Vaghi, H. 'Essais Sismiques de Poteaux en B&on Armd', CEA
ACKNOWLEDGEMENTS Note Technique D.M.T./78/177/(1978)
[14] Fleury, F., lie, N., Merabet, O. and Reynouard, J.M. 'A R/C
element for the nonlinear structural seismic analysis', in
The authors wish to acknowledge the assistance of 'Proceedings of the Second European Conference on Structural
Dr. B. A. Izzudin in the implementation of the proposed Dyamics-EURODYN'93', Trondheim, June, 1993 (A.A.
model into the non-linear program ADAPTIC. Balkema, 1993) 161-167.

147

Вам также может понравиться