Practical implementation of generalized pushover analysis for multimodal pushover analysis *Fırat Soner Alıcı1), Kaan Kaatsız2) and Haluk Sucuoğlu3) 1), 2), 3) Department of Civil Engineering, METU, Ankara, Turkey 1) [email protected] ABSTRACT A generalized pushover analysis procedure was previously developed for estimating the inelastic seismic response of structures under earthquake ground excitations. The procedure consists of applying different generalized force vectors separately for each story to the structure in an incremental form with increasing amplitude, until an estimated target interstory drift demand is obtained. A generalized force vector is expressed as a combination of modal forces, and simulates the instantaneous force distribution acting on the system when the interstory drift at a given story reaches its maximum value during dynamic response to a seismic excitation. Each nonlinear static analysis under a generalized force vector activates all modes of a multi degree of freedom system, simultaneously. Correspondingly, inelastic actions develop in members with the contribution of all instantaneous modes in the nonlinear response range. A practical implementation of the proposed generalized pushover analysis is presented in this study. The method is applied to a 12 story frame and a 20 story frame-wall system, and the results are obtained under six spectrum-compatible ground motions. First, it has been demonstrated that generalized pushover analysis is successful in estimating maximum member deformations and exact in determining member forces under a ground excitation with reference to nonlinear response history analysis. Then, it is shown that the results obtained from generalized pushover analysis by using the mean spectrum of six ground motions are almost identical to the mean of the results obtained from separate generalized pushover analyses under six ground motions. Moreover, these mean results are very close to the mean of the results obtained from six nonlinear response history analyses. 1. INTRODUCTION The ease of application and conceptual simplicity of single mode conventional pushover analysis enables to develop multimodal pushover analysis procedures which supersede nonlinear response history analysis in practical application (Sasaki et al. 1998, Chopra and Goel 2002, Gupta and Kunnath 2000, Aydınoğlu 2003, Antoniou and Pinho 2004). Generally, these methods, except one, are adaptive. In adaptive methods, 1), 2) Research Assistant, Ph.D. Candidate 3) Professor eigenvalue analysis is conducted at each step of load increment according to the formation of nonlinear deformations in the system. This approach removes the simplicity of pushover analysis procedure, and requires special programming or modification of the current analysis programs. On the other hand, in multimodal pushover analysis, results are obtained for each mode independently, and then the modal results are combined by statistical rules (SRSS or CQC) for combining elastic modal responses. There are several shortcomings in the modal combination of inelastic modal responses. They are approximate, and the internal forces obtained by statistical combinations exceed capacities, hence they require correction at each load increment. In this study, a recently developed multimodal pushover analysis (Sucuoğlu and Günay, 2011) which accounts for the contribution of all significant modes to inelastic seismic response is described, and its practical implementation is developed. In this procedure, a set of pushover analyses are conducted by employing different generalized force vectors, derived for each story. Therefore, for an N story building, N number of pushover analyses is required. Each generalized force vector is derived as a different combination of modal lateral forces in order to simulate the effective lateral force distribution when the interstory drift at a selected story reaches its maximum value during seismic response. Pushover analysis of a story proceeds in a step-by-step manner until the target interstory drift value of that story is achieved. Target interstory drift in any story can be estimated from linear elastic response spectrum analysis by considering the equal displacement rule. The contribution of the first mode response can also be obtained from the solution of inelastic SDOF system response. This suggested procedure can be implemented by using any structural analysis software (SAP, OpenSees, Drain2D, etc.) which facilitate displacement controlled pushover analysis. This study focuses on the practical implementation of generalized pushover analysis (GPA) which requires less computational effort. Performance of GPA and its practical implementation (RGPA) is compared with the benchmark nonlinear response history analysis, based on the results obtained from the analyses of a twelve story RC frame. 2. GENERALIZED PUSHOVER ANALYSIS The generalized pushover procedure is based on an effective force vector acting on the system when the interstory drift at the j‟th story reaches its maximum value during dynamic response. This is a generalized force vector since it includes contributions from all modal forces at the time of maximum response of the interstory drift at the j‟th story. If this force vector is defined, it can then be applied to the system as an equivalent static force in order to produce the maximum value of the j‟th interstory drift. In GPA, generalized force vectors are derived from the dynamic response of linear elastic MDOF systems under earthquake ground excitations by using the response spectrum analysis (RSA) procedure. ∑( ) (1 Here, ; is the n‟th mode shape; is the mass matrix, is the influence vector, and is the spectral acceleration at the n‟th mode. In Eq. (1), is the n‟th mode contribution to the maximum interstory drift of the j‟th story determined from RSA, and is the quadratic combination of the spectral modal drift terms according to Eq. (2). ∑[ ] (2 The target interstory drift demand at the j‟th story can be obtained consistently with the generalized force vector , ∑ (3 by invoking the modal scaling rule of GPA: (4 is the modal displacement amplitude at t which satisfies the equation of max motion of the SDOF system representing the n‟th mode under a ground excitation ̈ , and is the spectral displacement at the n‟th mode. satisfies the equation of motion, ̈ ̇ ̈ (5 where is the time when reaches its maximum value under ̈ In order to improve the prediction of Eq. (3), the first mode linear elastic spectral displacement demand can be replaced with the first mode inelastic spectral displacement demand . This operation requires conducting an „a priori‟ first mode pushover analysis. Then, can be estimated from the nonlinear response history analysis of the equivalent SDOF system which is representing the first mode behavior. GPA uses the higher-order interstory drift parameter as target demand rather than the lowest order story (roof) displacement parameter. Accordingly, when the associated generalized force vector pushes the system to the target drift , the system adopts itself in the inelastic deformation range, and the first and second order deformation parameters (rotations and curvatures) and the second and third order force parameters (moments and shears) take their inelastic values with more effective contributions from the higher modes. On the other hand, if story (roof) displacement is used as target demand, the contribution of higher modes becomes less significant. If different local response parameters are primary considerations during inelastic dynamic response, interstory drift values are more effective representatives of local maximum response parameters as target deformation because they are well synchronized with the local response parameters. N number of pushovers is conducted in GPA for each story sequentially. In the j‟th story GPA, the structural system is pushed incrementally with the force vector Displacement controlled pushover analysis is conducted until reaches . After completing all GPA from 1 to N, member deformations and member internal forces are determined by taking envelopes of the related GPA results, and these envelope values are registered as the maximum seismic response values. 3. REDUCTION IN THE NUMBER OF GENERALIZED PUSHOVERS: INSTANTS OF MAXIMUM RESPONSE GPA is based on the general assumption that the interstory drift ratios occur independently at each story (j=1-N) at different instants t , j=1-N. However if there j,max are n modes contributing significantly to the total dynamic response (n<N), then there are only 2(n-1) possible combinations of the n modes leading to the maximum values of interstory drifts at specific stories. Hence there are 2(n-1) independent instants t for max calculating and in the N DOF system. The reduction concept in generalized pushover analysis is introduced herein with implementation to a 12-story reinforced concrete building having four identical frames and a symmetrical floor plan, shown in Fig. 1. The first three modal periods are 2.39, 0.82, and 0.48 seconds, respectively. Fig. 1 a) Elevation, b) floor plan of the 12 story building Let‟s consider the linear elastic spectral drift profiles of the first three modes for the 12-story RC frame under the CHY006-E strong motion component of the 1999 Chi-Chi earthquake shown below in Fig. 2. 12 11 10 9 8 # 7 y 6 or St5 4 3 2 1 0 0 0.005 0.01 0.015 0.02 -0.01 -0.005 0 0.005 0.01 -0.01 -0.005 0 0.005 0.01 Interstory Drift Ratio Interstory Drift Ratio Interstory Drift Ratio Δ1 Δ2 Δ3 Fig. 2 Drift profiles for the first three modes under CHY006-E (drift ratios) Here, [ ( )] (6 is the n‟th mode spectral drift ratio at the j‟th story, where is obtained from the linear elastic response spectrum of CHY006-E. There are 2(n-1) possible combinations for n number of significant modes contributing to interstory drifts. For n=2 and n=3, possible combinations are shown in Fig. 3 and Fig. 4, respectively. It should be noted that these are the absolute maximum combinations of modal drifts. During actual dynamic response, maximum modal drift distributions do not occur synchronously. However this situation does not affect the approach presented. According to Fig. 3, the combinations Δ +Δ and Δ -Δ control the system in two 1 2 1 2 ranges along its height, the lower 1st-5th stories and the upper 6th-12th stories, respectively. On the other hand, according to Fig. 4, different combinations of the first three modes control the system in four different ranges. Fig. 4 reveals that Δ -Δ +Δ 1 2 3 combination controls the upper (9th-12th) stories whereas Δ +Δ +Δ combination 1 2 3 controls the lower (1st-3rd) stories. For the middle stories, Δ +Δ -Δ and Δ -Δ -Δ 1 2 3 1 2 3 combinations control the interstory drift maxima of the 4th-5th, and 6th-8th stories, respectively. This condition is further established with the elastic time history analysis (ETHA) results. In Fig. 5, drift profiles are plotted when the interstory drift values in each range attain their maximum values during linear elastic dynamic response analysis under CHY006-E for the related four combinations of the first three modes. The control stories are selected as the 2nd, 5th, 7th and 11th from the first, second, third and fourth story ranges, respectively. The envelope of maximum interstory drift values obtained from ETHA by considering all 12 modes are also plotted on the same figure. It is clear from Fig. 5 that the maximum interstory drifts occur at four different instants with the drift profiles approximately shown in Fig. 4. The intention for such comparison is to show that the reduced number of pushovers (from 12 to 4) with the related force vectors would be sufficient to capture the maximum dynamic response of the system under a ground motion. When the second story interstory drift reaches its maximum, the first and the third stories also reach their maxima synchronously in the first range as shown in Fig. 5. This is almost the case for the other ranges as well. The error at the ninth story is somewhat more significant, due to falling between two ranges. 12 12 11 11 Δ1+Δ2 10 10 9 Δ1 9 Range 2 8 8 # yro67 # yro67 tS5 tS5 4 4 Δ1-Δ2 3 Range 1 3 2 2 1 1 Δ1 0 0 -0.02 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 Interstory Drift (m) Interstory Drift (m) Fig. 3 Combinations of the first two modes contributing to interstory drift Accordingly, one story level from each story range can be selected, and the associated four (2(n-1), n=3) force vectors can be employed in GPA instead of the N numbers of force vectors, i.e., N number of generalized pushovers. The ranges are directly identified from the intercepts of the mode shapes as shown in Fig. 3 and Fig. 4. For the 12-story frame, considering the combinations of the first three modes, 2nd, 5th, 7th and 11th stories are selected from each story range (middle stories of each range), and only four generalized pushover analyses are conducted by applying f , f , f and f 2 5 7 11 in accordance with Eq. (1). Finally, the envelopes of these four generalized pushover analyses are employed for calculating the maximum response parameters. Thus, the computation effort in GPA is considerably reduced from 12 to 4 pushovers for this case. 12 12 11 Δ1+Δ2+Δ3 11 Δ1+Δ2-Δ3 10 Δ1+Δ2 10 Δ1+Δ2 9 9 8 8 # yro 67 # yro 67 tS 5 tS 5 4 4 Range 2 3 3 2 Range 1 2 1 1 0 0 -0.02-1E-160.02 0.04 0.06 0.08 0.1 -0.02 0 0.02 0.04 0.06 0.08 0.1 Interstory Drift (m) Interstory Drift (m) 12 12 11 11 Range 4 10 10 9 9 8 8 # yro 67 Range 3 # yro 67 tS 5 tS 5 34 Δ1-Δ2-Δ3 34 Δ1-Δ2+Δ3 2 2 1 Δ1-Δ2 1 Δ1-Δ2 0 0 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1 Interstory Drift (m) Interstory Drift (m) Fig. 4 Combinations of the first three modes contributing to interstory drift Fig. 5 Comparison of ETHA maximum drift envelope with the maximum drift profiles in four ranges 4. GROUND MOTIONS EMPLOYED IN CASE STUDIES The ground motion set employed in the case studies contains six different spectrum matched ground motions. Acceleration records of these six ground motions were generated from the selected reference data set of ground motions which have similar properties (Hancock and Bommer, 2006). The spectrum of each ground motion is adjusted with the Turkish Earthquake Code (TEC 2007) design spectrum. These reference ground motions were selected to be capable of generating higher mode effects on the structural systems, and downloaded from the PEER strong motion database (PEER, 2010). Important features of both the reference and the spectrum matched ground motions are presented in Table 1. The spectrum matched synthetic ground motions were derived by using the RSPMatch2005 (Hancock et al. 2006) software program. The derivation process is based on changing the frequency range of the reference ground motion in order to match the original acceleration spectrum with the given target acceleration spectrum. Fig. 6 shows the acceleration spectra of synthetic ground motions, their mean spectrum and TEC (2007) design spectrum. Acceleration spectrum of the reference GM4, scaled by 1.5 is also added to Fig. 6. This ground motion is used to represent maximum seismic intensity in the following analysis. Table 1 Properties of reference and spectrum matched ground motions PGA (g) PGV (cm/s) PGD (cm) Station - CD Site # GM Code Earthquake (Mw) Component (km) Geol. Spect. Spect. Spect. Ref. matched Ref. matched Ref. matched 1 CLS090 Loma Prieta, 1989 (7) Corralitos-090 3,9 A 0,479 0,449 45,2 53,2 11,3 20,3 2 LEX000 Loma Prieta, 1989 (7) Lex. Dam-000 5,0 A 0,420 0,496 73,5 66,2 20,0 24,9 3 PCD254 San Fer., 1971 (6.6) Pac. Dam-254 2,8 B 1,160 0,567 54,1 48,1 11,8 17,0 4 CHY006-E Chi-Chi, 1999 (7.6) CHY006-E 9,8 B 0,364 0,513 55,4 59,9 25,6 20,0 5 Bolu000 Duzce, 1999 (7.1) Bolu-000 12,0 D 0,728 0,541 56,4 55,9 23,1 18,7 6 ERZ-EW Erzincan,19 92 (6.9) Erzincan-EW 4,4 D 0,496 0,518 64,3 64,3 21,9 23,4 Fig. 6 Acceleration response spectrum matched ground motions, original GM4x1.5, mean acceleration spectrum, and TEC2007 design spectrum 5. COMPARATIVE RESULTS FROM CASE STUDIES The reduced GPA procedure is employed in two case studies, a 12 story frame and a 20 story frame-wall system, both designed according to the capacity design principles for the design spectrum given in Fig. 6 and detailed according to the enhanced ductility requirements of the Turkish Earthquake Code (2007). Solutions are obtained under six ground motion components, and the results are evaluated in comparison with the results from nonlinear response history analysis. Maximum interstory drifts, maximum average plastic rotations of beam ends in a story, maximum average chord rotations of column ends in a story, beam end moments and shear forces, column (shear wall) end moments and shear forces are obtained under the six ground motion components. Maximum average plastic or chord rotations at each story level are calculated by first averaging the member end rotations at the corresponding story at each load (or time) step, and then using the maximum of these values as the story maximum. The results of nonlinear response history analysis (NRHA), GPA and the reduced algorithm of generalized pushover analysis (RGPA) introduced herein are compared with each other for each ground motion in the set. The target drift in GPA is determined by employing the inelastic maximum displacement amplitude for the first mode in Eq. (3). 5.1 Case Study I: Twelve Story RC Frame The elevation and plan view of the frame were shown in Fig. 1. The member dimensions for beams are 300x550 mm for the first four stories, 300x500 mm for the second four stories, and 300x450 mm for the last four stories. The columns dimensions are 500x500 mm, 450x450 mm, and 400x400 mm in the first four, the second four, and the last four stories, respectively. The slab thickness is 140 mm, and live load is 3.5 kN/m2. Concrete and steel characteristic strengths are 25 MPa and 420 MPa, respectively. Analytical model of the twelve story frame is generated by using the OpenSees software (2005 . Structural frame members are modeled with “Beam with Hinges” element. For beam sections, bi-linear moment curvature relationships are defined along the hinge lengths by considering the designed section properties of each beam type. On the other hand, column sections are defined by utilizing fiber sections along the hinge lengths. The remaining part in between the plastic hinges of a member is defined with elastic section properties. In order to describe cracked section stiffness for each member, the gross moment of inertia values are multiplied by 0.4 and 0.6 for beams and columns respectively. Rigid diaphragms are assigned to each story level, and P-∆ effects are taken into account in the model. In time history analyses, Rayleigh damping is computed by considering 5% damping in the 1st and 3rd modes. Maximum interstory drift ratios and maximum average beam-end plastic rotations are compared for the results of full GPA, RGPA and NRHA in Fig. 7 and Fig. 8, respectively. It can be observed from these figures that the results of RGPA are very close to those of the full GPA. However RGPA results deviate from the NRHA results by as much as 30% at the stories above the mid height where NRHA calculates higher plastic deformations due to the higher mode effects. The basic reason is the difference between the plastic hinge formation sequences of the two procedures. Nonlinear static procedures cannot perfectly simulate the sequence which occurs during actual dynamic response. Fig. 7 Comparison of maximum interstory drift ratios obtained under six ground motions Fig. 7 Continued Fig. 8 Maximum values of average beam end plastic rotations under six ground motions