Documente Academic
Documente Profesional
Documente Cultură
www.elsevier.com/locate/engstruct
Received 14 June 2004; received in revised form 3 February 2005; accepted 3 February 2005
Available online 8 March 2005
Abstract
Two similar continuum plasticity material models are used to examine the influence of material modelling on the seismic response
of reinforced concrete frame structures. In the first model reinforced concrete is modelled as a homogenised material using an isotropic
Drucker–Prager yield criterion. In the second model, also based on the Drucker–Prager criterion, concrete and reinforcement are included
separately. While the latter considers strain softening in tension the former does not. The seismic input is provided using the Eurocode 8
elastic spectrum and five compatible acceleration histories. The results show that the design response from response history analyses (RHAs)
is significantly different for the two models. The influence of compression hardening and strength enhancement with strain rate is also
examined for the two models. It is found that the effect of these parameters is relatively small. In recent years there has been considerable
research in nonlinear static analysis (NSA) or pushover procedures for seismic design. The NSA response is frequently compared with that
obtained using RHA, which also uses the same material models, to verify the accuracy of the static procedure. A number of features exhibited
by reinforced concrete during dynamic or cyclic loading cannot be easily included in a static procedure. The design NSA and RHA responses
for the two material models are compared. The NSA procedures considered are the Displacement Coefficient Method and the Capacity
Spectrum Method. A comparison of RHA and NSA procedures shows that there can be a significant difference in local design response even
though the target deformation values at the control node are close. Moreover, the difference between the mean peak RHA response and the
pushover response is not independent of the material model.
© 2005 Elsevier Ltd. All rights reserved.
Keywords: Seismic design; Continuum plasticity; Response history analysis; Pushover methods
0141-0296/$ - see front matter © 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2005.02.003
P. Pankaj, E. Lin / Engineering Structures 27 (2005) 1014–1023 1015
Fig. 1. The four-storey frame used: (a) dimension; (b) beam cross-section; (c) column cross-section.
essentially based on the Drucker–Prager yield criterion [10], Theoretically, for a general nonlinear multiple degrees
the latter is capable of incorporating complex features such of freedom system, the peak seismic response (required for
as strain softening in tension, hardening in compression and design) can only be approximated by a static procedure.
stiffness degradation. The influence of material modelling There has been considerable research directed towards
on seismic response was considered earlier briefly by the improving pushover procedures so they can reflect various
authors [11] and in this paper this influence is examined in aspects of a nonlinear dynamic analysis. For example,
detail for a simple reinforced concrete plane frame. Chopra and Goel [16,17] proposed a modal pushover
Nonlinear response history analysis for several possible procedure to include contribution of higher modes. Chopra
ground motions, as prescribed by a number of codes, and Goel [18] provided a method to determine a capacity-
makes seismic design of structures very complicated. As demand diagram, in which the displacement demand was
a result, there has been considerable research to develop determined by analysing inelastic systems in place of
displacement based nonlinear static analysis (NSA) or equivalent linear systems. The suggested method used the
pushover procedures that can provide seismic design values. constant-ductility design spectra and was shown to be an
NSA response is frequently compared with that obtained improvement over the ATC-40 [15] procedures. Farfaj and
using response history analysis (RHA), which also uses the co-workers [19,20] extended the CSM procedure to include
same material models, to verify the accuracy of the static cumulative damage and called the method N2. The method
procedure. A number of features exhibited by reinforced has been shown to be a significant improvement over CSM
concrete during dynamic or cyclic loading (e.g. progressive and in many studies N2 is referred as a method distinct
degradation with each cycle of loading, influence of strain from CSM. This paper examines this difference between the
rate) cannot be easily included in a static procedure. design RHA and NSA response for both DCM and CSM
Therefore it is important to examine whether the difference procedures for a simple frame.
between the design RHA and NSA response is influenced by
the choice of material models. In other words, the hypothesis 2. The test structure and material modelling
that the comparison between a given NSA and RHA
procedure will show similar trends for different material The test structure used to evaluate the influence of
models needs to be tested. material modelling was a single-bay, four-storey frame.
Displacement-based NSA procedures exist in several The reinforced concrete members were modelled using
codes and guidelines in one form or the other [12–15]. Drucker–Prager plasticity and concrete damaged plasticity.
The existing nonlinear static techniques can be broadly In each case a number of variations were considered.
divided into two categories: Displacement Coefficient
Method (DCM) [13,14,16,17] and Capacity Spectrum 2.1. The test structure
Method (CSM) [15,18–20]. The common feature of
these techniques is that appropriately distributed lateral The test structure is shown in Fig. 1. The total mass
forces are applied along the height of the building, including live load for the frame is 97 000 kg. The columns
and then monotonically increased with a displacement were assumed fixed at the base. A damping ratio of 5%
control until a certain deformation is reached. The was assumed. The finite element model used two-node cubic
key difference between the CSM and DCM procedures beam elements. The finite element mesh comprised of four
is that the former usually requires formulation in an elements (for two columns) in each storey and four elements
acceleration–displacement format. representing beams at each floor level.
1016 P. Pankaj, E. Lin / Engineering Structures 27 (2005) 1014–1023
Table 2
Material properties used with the CDP model
3. Earthquake loading
considered. However, in this study the loading is applied indicates that the response induced by the peculiar nature
according to the first mode pattern only. FEMA 273 does of a time history can sometimes cause a strain rate that is
not provide a clear methodology for the determination of sufficiently significant to affect peak response. The influence
yield displacement and strength from the pushover curve. of the hardening parameter also varies significantly from
The bilinear curve determined from the pushover curve is one excitation to another. The maximum variation due to
often sensitive to the target displacement. This has been the hardening parameter is for excitation V. It is interesting
recognised in FEMA 274 [14]. In this study an iterative to note that in the dynamic environment hardening can
process was used to evaluate the yield values. Since the cause either an increase or decrease in the peak deformation
process is load controlled, it is often necessary to use the response. Comparing the peak responses from different
Riks procedure [8] to avoid problems with convergence. excitation histories with the mean values shows the largest
The CSM procedure adopted is numerical (rather than difference for the case DP-PP with strain rate effect included
graphical) based on the studies of Fajfar [20], Chopra and for the excitation history X. In general, the peak values vary
Goel [18] and Vidic et al. [30]. far more significantly when different spectrum compatible
time histories are used than due to inclusion of hardening or
strain rate.
5. Influence of material modelling on dynamic response
Table 4
The peak base shear (kN) in the DP structure
Table 5
5.2. CDP material model
The peak top deformation (m) in the structure modelled using CDP
Some typical responses of the structure subjected to Model Strain rate Earthquake history Mean
included
excitation V and modelled using CDP are shown in Figs. 6 V W X Y Z
and 7. The nomenclature used in these figures is similar
to that used earlier, i.e. HP and PP stand for hardening No 0.18 0.17 0.27 0.18 0.25 0.21
CDP-HP
Yes 0.18 0.17 0.26 0.17 0.24 0.21
and perfect plasticity respectively; ‘0’ and ‘001’ indicate
exclusion and inclusion of strain rate effects respectively. CDP-PP
No 0.18 0.17 0.27 0.17 0.25 0.21
For this model the response histories show that there is Yes 0.18 0.17 0.26 0.16 0.25 0.21
negligible influence of hardening parameter or strain rate on
the design parameters.
Once again the peak values of various response quantities
were examined. For example Table 5 lists the peak top Table 6
deformations. From Table 5 it can be seen that there is The peak base shear (kN) in the structure modelled using CDP
little influence of strain rate for any of the five earthquakes. Model Strain rate Earthquake history Mean
Comparing the response between the hardening and perfect included
plasticity, it can be seen that the differences are again V W X Y Z
small with maximum for earthquake Y (∼5%). The major No 122 128 124 119 136 126
CDP-HP
difference in the peak response is again due to different Yes 125 129 125 123 138 128
excitation histories. For example the top deformation of No 121 125 122 118 134 124
earthquake history X is around 28% higher than the mean CDP-PP
Yes 125 127 125 122 137 127
peak value. The analysis showed that the peak strain rate
during seismic excitation was around 0.004 per second.
However, this did not appear to influence the peak response The response of a local parameter, namely the peak mo-
significantly. Similarly it can be seen that the influence of ment at a base node (not shown), indicated a slightly higher
strain rate on base shear (Table 6) is small for different variation due to the strain rate effect (maximum ∼9%),
1020 P. Pankaj, E. Lin / Engineering Structures 27 (2005) 1014–1023
Table 7
Peak responses from RHA, DCM and CSM for CDP and DP structures
CDP-HP DP-HP
Response quantity RHA DCM CSM RHA DCM CSM
pattern that assumes that the response is controlled by the A comparison of RHA response with that obtained using
fundamental mode even in the post-elastic regime. This is DCM and CSM procedures shows that there can be a
consistent with previous findings [16]. significant difference in the internal force response between
The second and perhaps a more interesting feature dynamic and static procedures even though the target
demonstrated by the results is that the difference between deformation values at the control node match. Moreover, the
pushover and RHA response is not independent of the difference between the mean peak RHA response and the
material model. In other words this means that even if pushover response is not independent of the material model,
pushover and RHA responses closely match for a particular i.e. the static and dynamic procedures can yield similar
material model they may be different for another. The cause values for one material model and fairly dissimilar values
of these relative differences can be understood by examining for another.
the two material models used in this study. The DP model
essentially behaves like a bilinear force–deformation model
8. Notation
of the kind used in previous pushover studies [16,17]. During
monotonically increasing lateral loading of a pushover
analysis both branches of hysteretic force–deformation The following abbreviations and symbols have been used
relationship are utilized in a manner not too different from in this paper:
a cyclic loading situation. Thus a simple model of this CDP Concrete damaged plasticity (model)
kind is more likely to provide a better match between CSM Capacity spectrum method
the pushover and RHA response as the key attributes of
DCM Displacement coefficient method
the model are captured by the pushover procedure. Indeed
DP Drucker–Prager (model)
examining Fig. 12 it can be seen that the drift trend for RHA
and pushover procedures are similar along the height. In HP Hardening plasticity
fact the difference is largely due to the target deformation PP Perfect plasticity
that is underestimated by the pushover procedures (Table 7). E Young’s modulus of reinforced concrete (for
On the other hand the CDP model presents attributes that homogenised DP model)
cannot be captured by the pushover procedures used. In this Ec Young’s modulus of concrete (for CDP model)
model, while the reinforcement behaves in a bilinear manner Es Young’s modulus of reinforcement (for CDP
concrete does not. During a loading cycle elements undergo model)
compression hardening on one face and tensile strength fc Compressive yield strength of concrete
degradation on the other, followed by tensile degradation fy Yield stress for reinforcement (for CDP model)
and hardening on respective faces. These complex attributes HC Hardening modulus
of the model are only available in a cyclic loading regime
HT Softening modulus for concrete in tension (for CDP
and not in a monotonically increasing lateral load procedure.
model)
As a result the trend for storey drift for RHA and pushover
procedures can be seen to be different along the height in Kc Ratio of the second stress invariant on the tensile
Fig. 11 even though the target displacements are close. meridian to that on the compressive meridian for
concrete (for CDP model)
NSA Nonlinear static analysis
7. Conclusions RHA Response history analysis
β Friction angle
This simple study shows that the influence of strain rate
on the seismic analysis of reinforced concrete structures is ν Poisson’s ratio of concrete
small. The inclusion of a small value of hardening parameter ψ Dilation angle
has negligible influence on the RHA response for the CDP σb0 /σc0 Ratio of initial equibiaxial compressive yield stress
model and a small influence for the DP model. For a to initial uniaxial compressive yield stress (for CDP
given material model the peak RHA response from different model)
excitation histories causes significantly larger variation than σt 1 Initial tensile crack stress (for CDP model)
does inclusion or exclusion of compression hardening and
strain rate parameters. However, when the RHA response
of the two material models is compared a significant References
difference is observed. In the CDP model reinforcement is
included separately and it also includes strain softening in [1] CEB. Behaviour and analysis of reinforced concrete structures under
tension, while the DP model treats reinforced concrete as a alternate actions inducing inelastic response – vol. 1, General models.
Bull d’ Inf CEB, 210, Lausanne, 1991.
homogenized continuum. It is found that although the peak
[2] Penelis GG, Kappos AJ. Earthquake-resistant concrete structures.
deformation response (represented by the mean peak RHA E&FN Spon; 1997.
values) is fairly close, the internal force peak response from [3] Takeda T, Sozen MA, Nielsen NN. Reinforced concrete response to
CDP is significantly lower than that obtained from DP. simulated earthquakes. J Struct Eng Div, ASCE 1970;96(12):2557–73.
P. Pankaj, E. Lin / Engineering Structures 27 (2005) 1014–1023 1023
[4] Saiidi M. Hysteresis models for reinforced concrete. J Struct Eng Div, estimating seismic demands for buildings. Struct Dyn Earthq Eng
ASCE 1982;108(5):1077–87. 2002;31(3):561–82.
[5] Lee J, Fenves GL. Plastic-damage model for cyclic loading of concrete [18] Chopra AK, Goel RK. Capacity-demand-diagram methods for
structures. J Eng Mech 1998;124(8):892–900. estimating seismic deformation of inelastic structures: SDF systems.
[6] Lubliner J, Oliver J, Oller S, Oñate E. A plastic-damage model for Pacific Earthquake Engineering Research Center, University of
concrete. Int J Solids Struct 1989;25(3):229–326. California, Berkeley, Report No. PEER-1999/02; 1999.
[7] Lowes LN. Finite element modeling of reinforced concrete beam- [19] Fajfar P, Gaspersic P. The N2 method for the seismic damage analysis
column bridge connections. Ph.D. thesis. Berkeley: University of of RC buildings. Earthq Eng Struct Dyn 1996;25:31–46.
California; 1999. [20] Fajfar P. Capacity spectrum method based on inelastic demand
[8] ABAQUS V6. 3. ABAQUS/Standard user’s manual. Pawtucket (RI): spectra. Earthq Eng Struct Dyn 1999;28:979–93.
Hibbitt, Karlsson & Sorensen Inc; 2002. [21] Pankaj, Moin K. Exact prescribed displacement field solutions in
[9] Pivonka P, Lackner R, Mang HA. Shapes of loading surfaces of Mohr Coulomb elastoplasticity. Eng Comput 1996;13:4–14.
concrete models and their influence on the peak load and failure mode [22] Pankaj, Bicanic N. Detection of multiple active yield conditions for
in structural analyses. Int J Eng Sci 2003;41(13–4):1649–65. Mohr Coulomb elasto-plasticity. Comput Struct 1997;62:51–61.
[10] Drucker DC, Prager W. Soil mechanics and plastic analysis or limit [23] Correnza JC, Hutchinson GL, Chandler AM. A review of reference
design. Q Appl Math 1952;10(2):157–65. models for assessing inelastic seismic torsional effects in buildings.
[11] Lin E, Pankaj P. Nonlinear static and dynamic analysis – the influence Soil Dyn Earthq Eng 1992;11:465–84.
of material modeling in reinforced concrete frame structures. In: [24] Bischoff PH, Perry SH. Compressive behaviour of concrete at high
Thirteenth world conference on earthquake engineering. 2004, Paper strain rates. Mater Struct 1991;24:435–50.
no 430. [25] Bicanic N, Pankaj. Some computational aspects of tensile strain
[12] Eurocode 8, CEN. Design provisions for earthquake resistance localisation modelling in concrete. Eng Fracture Mech 1990;35(4–5):
of structures, Part 1.1, General rules – seismic actions and 697–707.
general requirements for structures, Draft for development. European [26] Basu S et al. Recommendations for design acceleration response
Committee for Standardization; 1998. spectra and time history of ground motion for Kakrapar site.
[13] FEMA-273. NEHRP Guidelines for the seismic rehabilitation Earthquake Engineering Studies, EQ∼83-5(revised), Department of
of buildings. Washington (DC): Federal Emergency Management Earthquake Engineering, University of Roorkee; 1985.
Agency; 1997. [27] Hibbitt HD, Karlsson BI. Analysis of Pipe Whip, EPRI, Report NP-
[14] FEMA 274. NEHRP commentary on the guidelines for the seismic 1208; 1979.
rehabilitation of building seismic safety council. Washington, DC; [28] Hilber HM, Hughes TJR, Taylor RL. Collocation, dissipation and
1997. ‘overshoot’ for time integration schemes in structural dynamics.
[15] Applied Technology Council. Seismic evaluation and retrofit of Earthq Eng Struct Dyn 1978;6:99–117.
concrete buildings. Report ATC 40, CA, USA; 1996. [29] Pankaj, Kumar A, Basu S. Interpolation of design accelerogram
[16] Chopra AK, Goel RK. A modal pushover analysis procedure to esti- for direct integration analysis of concrete structures. In: Bicanic N,
mate seismic demands for buildings: theory and preliminary evalua- Mang H, de Borst R, editors. Computational modelling of concrete
tion. Pacific Earthquake Engineering Research Center, University of structures, EURO-C, vol. 2. Pineridge Press; 1994. p. 1091–101.
California, Berkeley, Report No. PEER- 2001/03; 2001. [30] Vidic T, Fajfar P, Fischinger M. Consistent inelastic design spectra:
[17] Chopra AK, Goel RK. A modal pushover analysis procedure for strength and displacement. Earthq Eng Struct Dyn 1994;23:507–21.