Construction

and Building

Construction and Building Materials 20 (2006) 527–537

MATERIALS
[Link]/locate/conbuildmat

Nonlinear behavior of slender RC columns

(1). Numerical formulation
Hyo-Gyoung Kwak *, Jin-Kook Kim
Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology, 373-1 Guseong-dong, Yuseong-gu,
Daejeon 305-701, Korea

Received 28 November 2003; received in revised form 3 November 2004; accepted 31 January 2005
Available online 21 March 2005

Abstract

This paper introduces an analytical model which can simulate the nonlinear behavior of slender reinforced concrete (RC) col-
umns. The layer approach is adopted to determine the equilibrium conditions in a section and to consider the different material
properties across the sectional depth effectively. The material nonlinearity including the cracking of concrete is taken into consid-
eration, and geometric nonlinearity due to the P–D effect is taken into account by using the initial stress matrix. In advance, the
creep deformation of concrete is described in accordance with a first-order algorithm based on the expansion of a degenerate kernel
of the compliance function. To verify the analytical results, correlation studies with previous numerical results and experimental
data are conducted, and numerous parameter studies are followed to discuss the structural responses of slender RC columns accord-
ing to the changes in design variables. Finally, the necessity for a rigorous nonlinear analysis is emphasized for more accurate pre-
diction of the ultimate resisting capacity of slender RC columns.
 2005 Elsevier Ltd. All rights reserved.

Keywords: RC column; Ultimate resisting capacity; Slenderness ratio; P–D effect; Creep

1. Introduction building structures, either throughout an entire building

or in some parts of a structure, e.g., the exterior of build-
A reinforced concrete (RC) column, which is a pri- ings and the interior of lobbies. Moreover, the use of
mary structural member, is subjected to the axial force high strength steel and concrete has led to an increased
and bending moment which may be due to end restraint use of slender members. However, as slender RC col-
arising from the monolithic placement of floor beams umns may fail due to not only material failure in a sec-
and columns or due to eccentricity from imperfect align- tion but also instability of a structure, they require more
ment. Due to the combination of axial force and bend- rigorous numerical analyses which consider secondary
ing moment, the column section must be designed to effects such as the P–D effect and creep deformation of
ensure that the acting forces in a member exist inside concrete in order to reserve their strength and service-
the P–M interaction diagram representing the resisting ability.
capacity of the column. Recently, because of architec- There has been a lot of research on the behavior and
tural aesthetics and efficiency in use of space, relatively design of slender RC columns. On the basis of the force
slender columns have frequently been used in many equilibrium equation and the strain compatibility condi-
tion at a section, Bazant et al. [3,5] analytically calculated
*
Corresponding author. Tel.: +82 42 869 3621; fax: +82 42 869 3610.
the resisting capacity of slender RC columns by assum-
E-mail addresses: khg@[Link] (H.-G. Kwak), glory95@ ing a deflection curve with a sinusoidal function. Mate-
[Link] (J.-K. Kim). rial nonlinearity of steel and concrete was taken into

0950-0618/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2005.01.036
528 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537

account and an excellent discussion of slenderness effects

on interaction diagrams was provided. Kim et al. [10,11]
introduced a numerical method considering material and
geometric nonlinearities by using the layer model and
carried out an experimental study to verify the exactness
of the algorithm they developed. Drysdale and Huggins
[8] conducted experimental and analytical studies for
both short-term and long-term behaviors of RC columns
with relatively high slenderness ratio and discussed de-
crease of the ultimate resisting capacity due to the P–D
effect and creep deformation of concrete. Recently, Yal-
cin and Saatcioglu [21] developed an analytical model
that considers the influence of anchorage slip and plastic
Fig. 1. Behavior of slender RC column.
hinge length on the nonlinear behavior of RC columns.
In this paper, an analytical model to predict the
resisting capacity of slender RC columns is proposed. In the case of an RC column with a relatively large
The layer approach is adopted to simulate the different slenderness ratio, instability failure, which means a fail-
material properties across the sectional depth. Material ure before reaching the P–M interaction diagram of an
nonlinearity due to the cracking of concrete and yield- RC cross-section (the solid line envelope in Fig. 1),
ing of reinforcing bars, and geometric nonlinearity due may occur. A typical description for the long-term fail-
to the P–D effect are taken into account. Concrete ure is illustrated in Fig. 1. When an axial load PD with
creep is evaluated by the first-order algorithm based initial eccentricity e acts on a slender RC column at
on the expansion of the compliance function [9], and t = t0, the instantaneous lateral deflection will be e1
the aging effect of concrete properties is included in due to second-order effects (P–D effect). Moreover, the
the evaluation. The validity of the numerical model creep deformation during t0  t1 will increase the lateral
proposed in this paper is established by comparing deflection in spite of the absence of additional axial
the analytical predictions with results from previous load. If an additional live load is applied at t = t1, its
analytical studies [5,10], and numerical analyses for increment terminates at point C, located outside the
slender RC columns are conducted. On the basis of cross-section interaction diagram, because the strength
the numerical results obtained, the necessity of rigorous for a live load applied after a period of creep under con-
nonlinear analysis is emphasized for more accurate pre- stant load PD is usually higher than the short-term
diction of the ultimate resisting capacity of slender RC strength. The larger the column slenderness ratio, the
columns. greater is the reduction in the axial force resistance.
For not too slender columns, the failure occurs at
points rather close to the material strength. For very
2. Structural behavior of slender RC columns slender columns, on the other hand, the failure occurs
well within the cross-section interaction diagram be-
Generally, the ultimate compressive force P0 and the cause of a pronounced second-order effect. Referring
ultimate bending moment M0 for an RC column section to the structural behavior of RC columns, material non-
are related to each other by means of an interaction dia- linearities of steel and concrete and time-dependent
gram (P–M interaction diagram). In the absence of sec- deformation of concrete are taken into account for more
ond-order effects (P–D effect), as in very short columns, exact estimation of the ultimate resisting capacity of RC
the cross-section would undergo proportional loading columns. Moreover, it needs to be assumed that the fail-
until reaching the material strength at point A of the ure of RC columns occurs due to material failure only
cross-section interaction diagram (see Fig. 1). Slender because the slenderness ratio in most RC columns de-
columns, however, will follow the loading path up to signed in practice is smaller than the critical ratio that
point B where the material strength is reached. Point causes instability failure, even if the slenderness ratio
B is on the cross-section interaction diagram but is at ranges to a large value.
a smaller axial load, Pso, than it would be if L/r were
actually zero. Unlike a steel column, a concrete column
accompanies relatively large time-dependent deforma- 3. Modeling of material properties
tion, such as creep. This time-dependent deformation
gradually increases the lateral deflection caused by the 3.1. Concrete
initial eccentricity e and the P–D effect, and finally de-
creases the ultimate resisting capacity and serviceability Based on the principle of superposition, total uniaxial
of slender RC columns. concrete strain ec(t) at any time t is assumed to be
 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537 529

composed of the mechanical strain em c ðtÞ caused by short- ing to the mechanical strain can be calculated by using
term service loads, and the nonmechanical strain enm c ðtÞ the stress–strain relation of concrete.
consists of creep strain ecr sh
c ðtÞ and shrinkage strain ec ðtÞ. X
m
 
C ðt; sÞ ¼ ai ðsÞ  1  eki ðtsÞ : ð3Þ
ec ðt Þ ¼ em nm m cr sh
c ðt Þ þ ec ðt Þ ¼ ec ðt Þ þ ec ðt Þ þ ec ðt Þ: ð1Þ i¼1

Shrinkage strain can be evaluated directly by utilizing The response of RC columns under loads depends to a
the shrinkage models proposed in the design codes [1,7] large extent on the stress–strain relation of the constitu-
since it is defined as the volume change that occurs inde- ent materials and the magnitude of stress. Since concrete
pendently of imposed stresses. On the other hand, creep is used mostly in compression, the stress–strain relation
is defined as an increase in strain under sustained stress. in compression is of primary interest. Of many mathe-
Introduction of an analytical model to calculate the matical models currently used in the analysis of RC
creep strain is inquired, and many studies have been per- structures, the monotonic envelope curve introduced
formed [4,9]. In this paper, the first-order algorithm by Kent and Park and later extended by Scott et al.
based on expansion of creep compliance, proposed by [18] is adopted in this paper because of its simplicity
Kabir and Scordelis [9], has been adopted because this and computational efficiency. In this model, as shown
model can simulate the stress history effectively in spite in Fig. 2(a), the monotonic concrete stress–strain rela-
of its simplicity in application. tion in compression is described by three regions, where
The increment of creep strain from time tn  1 to time eco is the concrete strain at maximum stress, K is a factor
tn for uniaxial stress state can be expressed as follows which accounts for the strength increase due to confine-
[2,4]: ment, and Z is the strain softening slope.
On the other hand, it is assumed that concrete is line-
X
m
  arly elastic in the tension region. Beyond the tensile
Decn ¼ Ain1 1  eki Dtn ; Ain ¼ Ain1 eki Dtn þ ai ðsÞDrn ;
i¼1
strength, the tensile stress decreases linearly with increas-
ing principal tensile strain (see Fig. 2(b)). Ultimate failure
ð2Þ
is assumed to take place due to cracking, when the prin-
where Decnis the increment of creep strain, ki are inverse cipal tensile strain exceeds the value e0 = 2 Æ Gf/ft Æ ln(3/b)/
retardation times, ai(s) are constants depending on the (3  b) in Fig. 2(b), where b is the element length. Gf is the
age at loading s, m is a number of time steps, and Ai with fracture energy that is dissipated in the formation of a
initial values Ai1 ¼ ai ðt1 Þ  Dr1 at n = 1 represent hidden crack of unit length per unit thickness and is considered
state variables by which the effects of past time steps to be a material property. The value of e0 is derived from
are considered. the fracture mechanics concept of equating the crack
Before starting the calculation of creep strain by Eq. energy release with the fracture toughness of concrete
(2), parameters such as m, ai(s) and ki must be deter- Gf [12]. The experimental study by Welch and Haismen
mined. Since the use of the compliance function [20] indicates that for normal strength concrete, the value
(J(t, s) = 1/E(t) + C(t, s)) in the form of a Dirichlet series of Gf/ft is in the range of 0.005–0.01 mm. If Gf and ft are
induces some numerical difficulties caused by not con- known from measurements, e0 can be directly determined
sidering a separate term to represent the instantaneous from the equation in Fig. 2. More details for the concrete
deformation, the creep compliance (C(t, s)) is used di- model can be found elsewhere [13].
rectly in this study, as shown in Eq. (3). Hence, m = 4
is taken, and the assumed corresponding retardation 3.2. Steel
times are 8.0, 80.0, 800.0, and 8000.0, respectively. Be-
sides, the values of ai(s) are determined by the method Reinforcing steel is modeled as a linear elastic, linear
of least squares using Kabirs Dirichlet series creep com- strain hardening material with yield stress fy. The rea-
pliance [9]. After determination of nonmechanical strain sons for this approximation are: (1) the computational
increments, the concrete stress at each layer correspond- convenience of the model; and (2) the behavior of RC

(a) (b)

Fig. 2. Stress–strain relation of concrete.

530 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537

members is greatly affected by the yielding of reinforcing equilibrium condition of normal force only. To determine
steel when the structure is subjected to a monotonic the neutral axis while considering bending effects, the
bending moment [17,19]. Thermal strain is the only non- mechanical strains of concrete ðem m t
c Þ and steel ðes ¼ es Þ
m t
mechanical strain expected for steel. However, it is not need to be partitioned into an axial strain ðeca ; esa Þ and a
considered in this study, and thus the mechanical strain bending strain ðem t
cb ; esb Þ, as represented in Fig. 3.
can be directly calculated from the total strain of steel Since the axial strain is constant across the section
ðes ðtÞ ¼ em
s ðtÞÞ. and the bending strain is zero at the neutral axis, the
bending strains of concrete and steel at any layer can
be calculated by em m m t nm m t
cb ¼ ec  eca ¼ ec  ec  eca and esb ¼
4. Determination of neutral axis t t
es  esa , respectively. Based on the assumed neutral axis,
the stress corresponding to the mechanical bending
In order to formulate the constitutive relationships in
strain can be calculated from the stress–strain curves
the section of an RC column, the following simplified
of the constitutive materials, and iterations using the
assumptions have been made: (1) The section of an ele-
bisection method are repeated until errors for the axial
ment is divided into imaginary layers to describe the dif-
force and bending moment calculated by the axial strain
ferent material properties; (2) plane sections remain plane
and bending strain are within the given tolerances [14].
to represent the linearity in the strain distribution on any
In addition, time rate in a stress increment has been
section at any loading history; (3) a perfect bond between
considered. If a stress increment Drc(t0) is introduced
the concrete matrix and reinforcing bars is assumed; and
at time t0 and sustained without a change in magnitude,
(4) the constitutive materials are assumed to carry uniax-
the time variation of strain in concrete follows the
ial stress only. In addition, shear deformation is not taken
continuous line ABC in Fig. 4, and the total strain at
into account in the formulation because the shear effect is
time t, instantaneous plus creep, can be represented by
expected to be very small in slender RC columns.
Unlike a beam element subjected to a bending moment Drc ðt0 Þ
only, a column element is subjected to both axial force Decr
c ðtÞ ¼ ð1 þ /Þ; ð4Þ
Ec ðt0 Þ
and bending moments so that the neutral axis of a
column section cannot be calculated directly by the where / = /(t, to) is the creep coefficient.

(a) (b)

Fig. 3. Strain components at a section.

Fig. 4. Time variation of strain caused by a stress increment Drc.

 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537 531
 
On the other hand, when a stress increment Drc(t,t0) v
yw;x . Hence, x displacement U(x,y) and y displace-
is gradually introduced between t0 and t, the strain var- h
iation with time can be represented by the dashed line in ment V(x,y) may be expressed in terms of the displace-
Fig. 4. The total strain produced during the period t0  t ment column vector r
can be obtained by U ðx; yÞ ¼ ½/; yw;x   r; V ðxÞ ¼ ½0; w  r; ð7Þ
Drc ðt0 ; t0 Þ where w,x means the first order derivative of w with re-
Decr
c ðt; t 0 Þ ¼ ð1 þ v/Þ; ð5Þ
Ec ðt0 Þ spect to x.
where v is the concrete aging coefficient which accounts In addition, the axial strain e(x, y) can be defined by
for the effect of aging on the ultimate value of creep for  2
dU ðx; yÞ 1 dV ðxÞ
stress increments or decrements occurring gradually eðx; yÞ ¼ þ ; ð8Þ
dx 2 dx
after the application of the original load. It is found that
an average value of v = 0.82 can be used for most prac- where the second term represents the nonlinear displace-
tical problems where the creep coefficient lies between ment effect.
1.5 and 3.0 and t0 is greater than 5 days. An approxi- When a finite change in the joint displacement Dr oc-
mate value of 0.8 has frequently been used for v, and curs, corresponding changes in the strain De can be ex-
the same value of v = 0.82 is adopted in this study [16]. pressed by
dDU
De ¼
5. Construction of equilibrium equation dx
  1  T  
¼ /;x ; yw;xx  Dr þ DrT  0; w;x  0; w;x  Dr
Based on the assumed displacement field formula- 2
tion, all the constitutive equations including the element 1 T T
¼ B  Dr þ Dr  c  c  Dr; ð9Þ
stiffness matrix are derived. As shown in Fig. 5, the no- 2
dal displacement vectors of a two-dimensional beam ele- where B = [/,x, yw,xx] = [1/L, 1/L, y(1  2p)(6/L2),
ment in its local coordinate system can be expressed by y(1 + 2p)(6/L2), y(2  3p)(2/L), y(1  3p)(2/L)]. More-
u = {u1, u2}T, v = {v1, v2}T, and h = {h1, h2}T and the no- over, the incremental strain–displacement relationship
dal displacements of an element may be expressed as the of de = B Æ dr + drT Æ cT Æ c Æ Dr = drT Æ (B + cT Æ c Æ Dr)
column vector r = {u, v, h}T. can be developed by taking the differential from Eq. (9).
Assuming that the independent axial displacement Applying the virtual work principle of drT Æ
U0(x) varies linearly with x, and that the small rotation (R + DRj) = V de Æ (r + Dr) dV to a finite element on
j

hi at each node can be calculated by derivation of the the basis of the virtual displacement dr and neglecting
vertical displacement vi with respect to x, the displace- a higher order incremental term, the incremental of no-
ments, U0(x) and V(x) at any point within the element, dal force vector DRj applied at node j can be written
can be represented by from Eq. (9) as
  Z Z
v j T
U 0 ðxÞ ¼ /  u; V ðxÞ ¼ w  ; ð6Þ DR ¼ B Dr dV þ cT rc dV  Dr
h V V
Z Z
where / = [(1  p), p] and w = [(1  3p2 + 2p3), ¼ T t nm
B Eðde  de Þ dV þ cT rc dV  Dr
(3p  2p ), L(p  2p + p ), L(p + p3)] represent the
2 3 2 3 2
V V
Z Z
displacement shape functions, and the nondimensional T T
parameter p denotes x/L, that is, the position along ¼ B EB dV þ c rc dV  Dr
V V
the axis of the beam element. Z
T nm
Then, by adopting the plane section hypothesis, the x  B E de dV ; ð10Þ
displacement U(x) at any point can be written by the V

relation of U ðx; yÞ ¼ U 0 ðxÞ  y  dV ðxÞ=dx ¼ /  u where E is the tangent modulus of constitutive material.

Fig. 5. Displacement components in a beam element.

532 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537

Finally, the equilibrium equations can be rewritten in

differential form as
dR ¼ dRj þ dRnm ¼ ½K   dr; ð11Þ
where
Z Z
½K ¼ ½K e  þ ½K g  ¼ BT EB dV þ cT rc dV ;
V V
Z Z
T T
½K e  ¼ B EB dV ; ½K g  ¼ c rc dV ;
ZV V

dRnm ¼ BT E denm dr;

V

2 3
EA=L EA=L 0 0 0 0
6 EA=L 0 0 0 0 7
6 7
6 7
6 12EI=L3 12EI=L3 6EI=L2 6EI=L2 7
½K e  ¼ 6
6
7;
6 12EI=L3 6EI=L2 6EI=L2 7
7
6 7
4 sym: 4EI=L 2EI=L 5
4EI=L
ð12Þ
2 3
0 0 0 0 0 0
6 0 0 0 7 0 0
6 7
6 7
6 6P =5L 6P =5L P =10 P =10 7
½K g  ¼ 6
6
7;
6 6P =5L P =10 P =10 7
7
6 7
4 sym: 2PL=15 PL=30 5
2PL=15
ð13Þ
while calculating the elastic stiffness [Ke] and the geomet-
ric stiffness [Kg], the value of E at each layer is assumed
to be held constant along the element length, and thus
the volume integration in Eq. (10) can be represented
by the inner product of the line integration along the ele-
Fig. 6. Solution procedure.
ment length and the area integration across the sectional
depth. Moreover, since the layer approach is employed,
wherein a typical section is divided into imaginary lay-
ers, the sectional stiffness terms of EA and EI in Eq. ture depends on the displacement increments, the solu-
(12) can beR evaluated tion of the equilibrium equations is typically
Pnby c
summation Pns over all layers, accompanied by an iterative method through the con-
i.e.,
R EA ¼ P
A
E dA ¼ i¼1 EPns
c i A c i þ i¼1 E si Asi and EI ¼
Ey 2
dA ¼
nc
E c y 2
A c þ E s y 2
A , where nc and vergence check. The nonlinear solution scheme selected
A i¼1 i c i i i¼1 i s i si
ns denote the number of concrete and steel layers respec- in this paper uses the tangent stiffness matrix at the
tively, Ai and Ei are the sectional area and elastic mod- beginning of each load step and each time step in
ulus of ith layer, yi is the distance from the centroid, and combination with a constant stiffness matrix during
P refers to the applied force. the subsequent correction phase, that is, an incremen-
tal-iterative method.
The criteria for measuring the convergence of the
6. Solution algorithm iterative solution are generally based on the accuracy of
satisfying the global equilibrium equations or on the
Every nonlinear analysis algorithm consists of four accuracy of determining the total displacements. The
basic steps: the formation of the current stiffness matrix, accuracy of satisfying the global equilibrium is controlled
the solution of the equilibrium equations for the dis- by the magnitude of the unbalanced nodal forces. Hence
placement increments, the state determination of all ele- the convergence criteria for the unbalanced nodal forces
ments in the model, and a convergence check. These are used in this paper, and these can be expressed as
steps are presented in some detail in the flow diagram
of Fig. 6. Since the global stiffness matrix of the struc- F max
unbal: 6 Tol:F ; M max
unbal: 6 Tol:M ; ð14Þ
 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537 533

where jF max max

unbal: j and jM unbal: j are the absolute values of the
maximum unbalanced axial force and bending moment,
respectively, and Tol.F and Tol.M are the specified toler-
ances corresponding to the axial force and bending mo-
ment. Tol.F = Tol.M = 0.01 are taken in this study, and
more details for the solution procedures can be found
elsewhere [14].

7. Analytical and experimental verification

7.1. Short-term loading

The experimental results from several hinged RC col-

umns tested by Kim and Yang [11] are used to investi-
gate the validity of the analytical model proposed in
this paper. More details from the material properties
Fig. 7. Comparison of analytical model with test results.
used in the experimental procedure for each specimen
can be found elsewhere [11].
The second group of specimens used to validate the
The ultimate loads of columns measured experimen-
proposed analytical model is a series of columns with
tally are compared with those obtained by the analytical
a width · depth of 20 cm · 30 cm. These columns were
model in Table 1 in which the third column shows two
tested by Chuang and Kong [6]. More details of the
test results measured from two specimens with the same
material properties as well as the configuration of the
specification for each slenderness ratio. The good agree-
specimens can be found elsewhere [6]. From Table 2,
ments for the individual columns, regardless of the com-
which shows the experimental results and the analytical
pressive strength of concrete and the slenderness ratio,
predictions, it can be seen that the proposed numerical
lead to the conclusion that the ultimate loads of hinged
model accurately predicts the ultimate load regardless
RC columns can be accurately predicted by the pro-
of the eccentricity ratio.
posed numerical model. In addition, Fig. 7 representing
An additional comparison with numerical calcula-
a relation between the axial force and lateral deflection
tions introduced by Bazant et al. [3,5] is conducted to
at the mid-span, shows that the proposed numerical
verify the accuracy of the proposed numerical model.
model not only gives accurate predictions for the ulti-
The steel ratio and cross-section dimensions of the se-
mate load but also effectively simulates the nonlinear
lected columns with slenderness ratios of L/r = 10, 70
behavior of simply supported slender RC columns as
and 100 are represented in Fig. 8. The same material
the axial force increases from zero to its ultimate value.
properties of concrete and steel as those in the previous
analytical study are used and have the following values:
Table 1
fc0 ¼ 5000 psi (352 kgf/cm2), Es = 29 · 106 psi (2.04 ·
Comparison of computed ultimate strength of RC columns with test
results from Kim and Yang [11] 106 kgf/cm2) and fy = 60 000 psi (4220 kgf/cm2).
For the design of slender RC columns, ACI318 [1]
fc0 ðkgf=cm2 Þ L/r Test results Pu,t Analysis results Pu,a Pu,a/Pu,t
(·103 kgf) (·103 kgf) recommends a simple approximate formula based on
(Kim et al.) (This study) the moment magnification factor. When a column is
259.9 60 6493 6600 1.02 subjected to ultimate loads of Pu and Mu, the load
6697 0.99 and moment used in the design of the column section
are assumed to be Pu and d Æ Mu, where d is the moment
100 3894 3780 0.97 magnification factor, and is calculated from d = Cm/
3568 1.06
(1  Pu//k Æ Pcr) P 1; Pcr is the elastic buckling load;
647.3 60 10 479 10 400 0.99 and Cm is an equivalent uniform moment diagram
11 570 0.90 factor. Note that Cm = 1 for the columns used in this
example.
100 4608 4840 1.05 Moreover, as mentioned in a previous study [15], the
4852 1.00
stiffness reduction factor /k designed to consider the
878.7 60 12 446 11 840 0.95 inevitable random variability of the materials has not
12 610 0.94 been incorporated for the purpose of comparison with
the numerical results (i.e. /k = 1.0 is assumed).
100 5535 5520 1.00 As shown in Fig. 8, the results from the proposed
5596 0.99
numerical model are in good agreement with those
534 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537

Table 2
Comparison of computed ultimate strengths of RC columns with test results from Chuang and Kong [6]
Specimen L/r e/h q (%) Test results Pu,t (·103 kgf) Analysis results Pu,a (·103 kgf) Pu,a/Pu,t
A-17-0.25 58.9 0.25 3.27 120.3 123.5 1.03
A-18-0.25 62.4 110.1 113.8 1.03
A-19-0.25 65.8 121.8 107.3 0.88
A-17-0.50 58.9 0.50 91.9 92.0 1.00
A-18-0.50 62.4 86.7 84.0 0.97
A-19-0.50 65.8 83.1 82.0 0.99

B-17-0.25 58.9 0.25 1.34 110.8 106.0 0.96

B-18-0.25 62.4 100.7 100.0 0.99
B-19-0.25 65.8 106.6 92.0 0.86
B-17-0.50 58.9 0.50 48.8 50.0 1.02
B-18-0.50 62.4 48.8 50.0 1.02
B-19-0.50 65.8 47.1 49.0 1.04

cross-section dimensions of the tested columns are rep-

resented in Fig. 9, and those columns have a slenderness
ratio of L/r = 107. Moreover, the following material
properties, which have the same values with those used
in the experiments, are used: fc0 ¼ 4000 psi (282 kgf/
cm2), Es = 29 · 106 psi (2.04·106 kgf/cm2), and fy =
56 000 psi (3940 kgf/cm2). To trace the time-dependent
behavior of RC columns, creep and shrinkage of con-
crete are considered with the aging effect of concrete,
and an ultimate creep coefficient of cu = 3.0 and an ulti-
6
mate shrinkage strain of e1 sh ¼ 600  10 are used on
the basis of the ACI model.
Fig. 9 shows a plot of Pou/Put versus bd where
bd = PD/(PD + PL) = ratio of dead load to total load,
Pou = short-term failure load, and Put = long-term fail-
ure load. When an RC column is subjected to a small
Fig. 8. Failure envelopes for pin-ended RC columns. sustained load, which means that bd has a small value,
the points of failure caused by the subsequent sudden
obtained by Bazant and Xiang [5], and leading to the overload generally indicate a substantial increase in
conclusion that the ultimate loads of slender RC col- strength. This phenomenon, however, is almost nonexis-
umns can be accurately predicted by the proposed tent in the tests of Drysdale et al. because drying
numerical model. Fig. 8 also leads to the following con-
clusions: (1) As the slenderness ratio increases, the differ-
ence between the ACI strength interaction curve and the
proposed model gradually increases; (2) the ACI method
may underestimate the resisting capacity of slender RC
columns; and (3) the ACI method does not achieve a
uniform safety margin, defined in this study as the uni-
form difference between the results predicted by the
ACI method and the results calculated by a rigorous
analysis over the entire interaction diagram for columns
with L/r = 70 and 100.

7.2. Long-term loading

To verify the accuracy of the proposed numerical

model for long-term behavior, correlation studies be-
tween analytical and experimental results by Drysdale
and Huggins [8] are conducted. The geometry and Fig. 9. Long-term resistance of RC columns.
 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537 535

prevents the increase of concrete strength due to hydra- ties are assumed: fc0 ¼ 360 kgf=cm2 , Es = 2.1 · 106 kgf/
tion. This means that the analytical results may repre- cm2 and fy = 4350 kgf/cm2. The resulting strength inter-
sent lower values of Pou/Put than those from the action curves in terms of primary bending moments are
experimental data along the whole range of bd because given in Fig. 10.
the increase in concrete strength can accurately be con- As shown in Fig. 10, which represents the relation be-
sidered in the numerical analyses by implementing the tween the axial force and the primary bending moment,
age effect. In this aspect, the two models proposed by the P–D effect appears more significant as the steel ratio
the ACI and Bazant et al. give slightly conservative re- decreases. This seems to arise from the fact that columns
sults and underestimate the resisting capacity of slender with relatively small steel ratios have smaller bending
RC columns as bd increases. In advance, since the dead stiffnesses, EI, at the post-cracking state, and this leads
load takes possession of 50% of the total design load in to an increase in the lateral deflection and accompanying
most RC columns (bd P 0.5), direct application of the the P–D effect. On the other hand, the P–D effect gradu-
ACI model or the Bazant model may result in a conser- ally disappears as the slenderness ratio converges to zero,
vative design. so that a minor influence of the steel ratio on the P–D ef-
On the other hand, the numerical results from the fect appears in columns with low slenderness ratios (i.e.,
proposed analytical model agree well with the experi- L/r 6 30). This effect also disappears when the applied
mental results. Especially at bd = 1, the numerical result axial force P becomes zero. More details related to the
shows 30% reduction in the ultimate resisting capacity P–D effect can be found in the companion paper.
for the axial load. This coincides well with the experi- Fig. 11 shows the structural response of slender RC
mental results, which represent reductions ranging from columns according to changes in the compressive
20% to 40%. strength of concrete. A more significant decrease of
the ultimate resisting capacity with an increase of the
slenderness ratio appears in the RC columns using con-
8. Numerical analyses crete with relatively high compressive strength. Since the
stress–strain relation of concrete represents more brittle
8.1. Short-term loading behavior as the compressive strength increases, a slender
high strength concrete column has a relatively small en-
The ultimate resisting capacity of slender RC col- ergy absorption capacity when it is subjected to a large
umns is affected by many variables in addition to the lateral deflection accompanied by the P–D effect. This
compressive strength of concrete, such as the slenderness is why a significant decrease of the ultimate resisting
ratio, steel ratio, eccentricity, etc. In order to isolate the capacity occurs in the high strength concrete column.
effects of these variables, a parametric study is con- The obtained results also show that the use of high
ducted. The same cross-section dimensions as those used strength concrete in slender RC columns is not as effec-
in the analytical verification under short-term loading tive as in short columns.
are used (see Fig. 10). Two values of steel ratio
(qs = 0.03 and 0.08) are investigated and the slenderness 8.2. Long-term loading
ratio is limited to a maximum L/r = 70, because the slen-
derness ratio of RC columns generally used in design In order to study the long-term behavior of slender
practice is less than 70. The following material proper- RC columns, time-dependent analyses using the

(a) (b)

Fig. 10. P–M interaction diagrams of RC columns in accordance with the steel ratio.
536 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537

(a) (b)

Fig. 11. P–M interaction diagrams of RC columns in accordance with the compressive strength of concrete.

proposed model are conducted in the case of bd = 1 be- are the same as those used in the short-term loading
cause this case represents the most severe decrease in the case. Figs. 12 and 13 show the P–M interaction dia-
ultimate resisting capacity for the axial load (see Fig. 9). grams, which represent the relation between the axial
The cross-section dimensions and material properties force and the primary bending moment.

(a) (b)

Fig. 12. P–M interaction diagrams of RC columns in accordance with the steel ratio.

(a) (b)

Fig. 13. P–M interaction diagrams of RC columns in accordance with the compressive strength of concrete.
 H.-G. Kwak, J.-K. Kim / Construction and Building Materials 20 (2006) 527–537 537

As shown in Fig. 12, greater strength reduction due Acknowledgements

to the creep deformation of concrete appears in RC col-
umns with relatively small reinforcement because the The research reported in this paper was made possi-
creep deformation in an RC section increases as the steel ble by the financial support from the Smart Infrastruc-
ratio decreases and larger creep deformation accompa- ture Technology Center funded by the Korea Science
nies greater increase of the lateral deformation. In ad- and Engineering Foundation. The authors would like
vance, this figure shows that more significant strength to express their gratitude to this organization for the
reduction due to creep deformation occurs in RC col- financial support.
umns with larger slenderness ratios, and this tendency
seems to be maintained always, regardless of changes
in the variables. Fig. 13 also shows a decrease of the ulti- References
mate resisting capacity of RC columns according to the
creep of concrete. It can be seen that the strength reduc- [1] ACI Committee 318 Building Code Requirements for Reinforced
tion due to the creep effect is increased proportionally to Concrete (ACI 318–99). Detroit: American Concrete Institute;
the compressive strength of concrete. Moreover, from a 1999.
[2] ASCE. State-of-art report on finite element analysis of reinforced
comparison of Figs. 12 and 13 with Figs. 10 and 11 for
concrete. New York: American Society of Civil Engineers; 1982.
the short-term loading case, it can be found that all the [3] Bazant ZP, Cedolin L, Tabbara MR. New method of analysis for
characteristics for the resisting capacity of RC columns slender columns. ACI Struct J 1991;88(4):391–401.
represented in short-term loading are maintained in the [4] Bazant ZP, Wu ST. Dirichlet series creep function for aging
long-term behavior. concrete. J Eng Mech ASCE 1973;99(EM2):367–87.
[5] Bazant ZP, Xiang Y. Inelastic buckling of concrete column in
braced frame. J Struct Eng 1997;123(5):634–42.
[6] Chuang EH, Kong FK. Large-scale tests on slender, reinforced
concrete columns. Struct Eng 1997;75(4/5):410–6.
9. Conclusions [7] Comite Euro-International du Beton CEB-FIP Model Code.
London: Thomas Telford Service; 1993.
A numerical model to simulate the nonlinear behav- [8] Drysdale RG, Huggins MW. Sustained biaxial load on slender
concrete columns. J Struct Div, ASCE 1971;97(No. ST5).
ior of slender RC columns considering the long-term
[9] Kabir AF, Scordelis AC. Nonlinear analysis of reinforced
deformations of concrete is presented in this paper, concrete panels, slabs and shells for time dependent effects.
and the proposed model is verified by comparison with Report No. UC–SEEM 76–6. Berkeley: University of California;
results from previous analytical and experimental stud- 1976.
ies. Moreover, on the basis of the numerical results in [10] Kim JK, Lee SS. The behavior of reinforced concrete columns
subjected to axial force and biaxial bending. Eng Struct
this limited investigation, the following conclusions are
2000;22(11):1518–28.
obtained: (1) the use of high-strength concrete in slender [11] Kim JK, Yang JK. Buckling behaviour of slender high-strength
RC columns is not as effective as in short columns since concrete columns. Eng Struct 1995;17(1):39–51.
the resisting capacity is significantly decreased in high- [12] Kwak HG, Filippou FC. Finite element analysis of reinforced
strength concrete columns; and (2) an increase of the concrete structures under monotonic loads. Report No. UCB/
SEMM-90/14. Berkeley: University of California; 1990.
steel ratio leads to a relatively small decrease in the ulti-
[13] Kwak HG, Kim SP. Bond-slip behavior under monotonic
mate resisting capacity of slender RC columns and im- uniaxial loads. Eng Struct 2001;23(3):298–309.
proves the structural behavior under short-term and [14] Kwak HG, Seo YJ. Long-term behavior of composite girder
long-term loading. bridges. Comput Struct 2000;74(5):583–99.
Since the nonlinear behavior of slender RC columns [15] Mirza SA, Lee PM, Morgan DL. ACI stability resistance
factor for RC columns. J Struct Eng, ASCE 1987;113(9):
is dominantly affected by many design variables such
1963–1976.
as the slenderness ratio, steel ratio, compressive strength [16] Neville AM, Dilger WH, Brooks JJ. Creep of plain and
of concrete, eccentricity and the magnitude of ultimate structural concrete. London and New York: Construction
creep coefficient, etc., sophisticated numerical methods Press; 1982.
considering material and geometric nonlinearities will [17] Ngo D, Scordelis AC. Finite element analysis of reinforced
concrete beams. ACI 1967;64(3):152–63.
play an increasingly important role and will become
[18] Scott BD, Park R, Priestley MJN. Stress–strain behavior of
the standards for final design checks. Nevertheless, the concrete confined by overlapping hoops at low and high strain
introduction of a simple design formula that can be rates. ACI 1982;79(1):13–27.
effectively used in practice to determine an initial section [19] Vedo A, Ghali A. Moment–curvature relation of reinforced
of slender RC columns may be required because the de- concrete slabs. J Struct Div, ASCE 1977;103(3):515–31.
[20] Welch GB, Haisman B. Fracture toughness measurements of
sign formulas noted in the current design codes still have
concrete. Report No. R42. Sydney: University of New South
some limitations in considering numerous design vari- Wales; 1969.
ables and they give slightly conservative results for slen- [21] Yalcin C, Saatcioglu M. Inelastic analysis of reinforced concrete
der RC columns. columns. Comput Struct 2000;77(5):539–55.