Finite Elements in Analysis and Design 43 (2007) 879 887 www.elsevier.

com/locate/nel

3-D pushover analysis of a collapsed reinforced concrete chimney

Wei Huang a, , Phillip L. Gould b
a KPFF Consulting Engineers, Los Angeles, CA 90045, USA b Department of Civil Engineering, Washington University, Saint Louis, MO 63130, USA

Received 12 May 2006; received in revised form 6 April 2007; accepted 30 May 2007 Available online 17 July 2007

Abstract During the Izmit (Kocaeli) Earthquake of August 17, 1999, a 115 m high reinforced concrete chimney or heater stack, located at the Tpras Renery, collapsed. This stack was designed and constructed according to international standards and is representative of similar structures at reneries throughout the world, including those in earthquake-prone regions. This structure is of particular interest because several similar chimneys at the site survived the shock with only moderate damage. The particular distinction of this chimney appears to be an unusually larger rectangular opening, located about 1 of the height above the base, which appeared to be the region of collapse initiation. 3 The main focus of the research is the dynamic response of the stack due to an earthquake motion recorded at a nearby site. A new 3-D pushover analysis procedure is proposed in this paper and the results will be compared with those of a nonlinear dynamic analysis. Results are presented that show the importance of the 3-D interaction effects in the dynamic response of the stack. The results also conrm that the stack could readily fail under the considered earthquake and are consistent with the debris pattern. 2007 Elsevier B.V. All rights reserved.
Keywords: Finite element; Chimney; Nonlinear 3-D pushover analysis; Failure

1. Introduction and objectives This study is focused on a 115 m high reinforced concrete chimney that collapsed during the 1999 Kocaeli earthquake. This earthquake caused great damage to inhabited structures and the regional transportation system that has been well documented. The coincident damage to industrial facilities did not produce a high death toll, but the economic repercussions were enormous. Furthermore, many of these facilities were designed and constructed to international standards and provide information that is readily transferable to other developed countries. The reinforced concrete chimney shown near the center of Fig. 1 collapsed during the earthquake. The debris cut many lines, which fueled res that shut down the renery for months. This structure is of particular interest because several similar chimneys at the site survived the shock with only
Corresponding author. Tel./fax: +1 310 745 1587.

moderate damage. As shown in Fig. 1, the collapsed heater stack is shown next to a similar structure that survived. The particular distinction of this chimney appears to be an unusually larger rectangular opening, located about 1 of the height above 3 the base, which appeared to be the region of collapse initiation. The remnants of the stack are shown in Fig. 2. The overall objectives of the study are four-fold: (1) To evaluate the original design of the collapsed chimney, known as the Tpras stack, using current analysis techniques. (2) To evaluate the design of a similar size chimney representative of US practice. (3) To explain why the single stack in question did indeed collapse while several similar structures in the same vicinity survived with minimal damage through the use of advanced seismic evaluation tools. (4) To extend the pushover analysis procedure for chimney structures by taking into account the higher modes and the 3-D interaction effects.

E-mail address: w_huang@hotmail.com (W. Huang). 0168-874X/$ - see front matter doi:10.1016/j.nel.2007.05.005 2007 Elsevier B.V. All rights reserved.

880

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887

Fig. 1. Heater stacks before earthquake.

Fig. 2. Heater stacks after earthquake.

The input for the study is a single strong motion record recorded at a site near the failed stack, named the YPT (A PetroChemical Plant in Krfez, Turkey) record. No other nearby record is available, so this record is adopted as the input motion for the analysis of the Tpras stack. For the YPT longitudinal spectrum (YPTx) and transverse spectrum (YPTy), along with a modern design code spectrum, UBC 97 (1997 Uniform Building Code) spectrum, several demand curves are plotted in the spectral acceleration vs. spectral displacement domain (ADRS) as shown in Figs. 3 and 4. The ADRS spectrum is converted from the ordinary response spectrum using the relationship between acceleration, displacement and period from simple harmonic motion in order to permit comparisons with the capacity curve. Note that any radial line from the origin represents a constant period on all of the intersected demand curves. Also included on the gures are the rst mode pushover capacity curves for different loading conditions, among which a prole oriented 90 to the opening is the critical pushover case. The construction of these curves is explained in later paragraphs. The damping ratios for intact reinforced concrete chimney structures are fairly low, typically in the range of 25%. Following the usual practice for reinforced concrete, 5% damping was used. The rst two objectives were addressed earlier [13] by a response spectrum analysis based on the unsmoothed YPT record as well as the UBC 97 design spectrum. The comparative capacities of the Tpras stack and a comparably sized US stack were based on their respective structural designs, which are somewhat different in approach due to changes in practice over the years. Elevations of the stacks are shown in Fig. 5. For the third objective, a contemporary nonlinear technique was applied. This method leads to a comparison of demand and capacity as well. In this case, the demand was based on the YPT record and also a smoothed spectrum derived from

UBC-97 YPTx Mean+1 STDV Mode1 Withhole 90 Deg.

YPTx Mode1 Nohole

YPTx Mean Mode1 Withhole 0 Deg.

1.60 1.40
T = 1.0 s

Spectral Acceleration (g)

1.20 1.00 0.80 0.60 0.40 0.20 0.00 0 20 40 Spectral Displacement (cm) 60 80
T = 2.0 s

Fig. 3. Mode 1 capacity curves vs. YPTx demand curves.

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887
UBC-97 YPTy Mean+1 STDV Mode1 Withhole 90 Deg. YPTy Mode1 Nohole YPTy Mean Mode1 Withhole 0 Deg.

881

1.60 1.40
T = 1.0 s

Spectral Acceleration (g)

1.20 1.00 0.80 0.60 0.40 0.20 0.00 0 20 40 Spectral Displacement (cm) 60 80
T = 2.0 s

Fig. 4. Mode 1 capacity curves vs. YPTy demand curves.

Fig. 5. Chimney elevations.

a statistical earthquake simulation, so as not to overemphasize the local peaks and valleys [4]. The capacities were provided by the pushover curves shown in Figs. 3 and 4. This approach attempted to distinguish the differences in the deformation capacity of a chimney with a large opening hole, as opposed to a chimney without such an opening. Also, the direction of the push, either perpendicular to the plane of the opening or 90 to that direction, was studied.

Pushover analysis is a nonlinear static procedure (NSP) in which a lateral load pattern is applied to the structure and then incrementally increased until the structure reaches the target displacement or collapses. Due to its conceptual simplicity and computational attractiveness as compared to nonlinear dynamic analysis, pushover analysis has been gaining popularity as a tool for seismic design and performance evaluation of structures. However, it has been shown by many researchers [57]

882

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887

that despite its efciency and applicability, it also exhibits signicant limitations when higher modes and 3-D interaction effects play important roles in the dynamic response of the structure. For a chimney structure like the Tpras stack, this may be the case. How to take those effects into account and develop a simple but improved pushover procedure for seismic design and evaluation of a chimney structure is the fourth objective of the study and the focus of this paper. Following a summary of the 2-D pushover analysis study considering the higher mode effects, a new 3-D pushover analysis procedure is introduced in this paper and some comparisons with nonlinear dynamic analysis results are presented. The stacks were analyzed using the nite element (FE) analysis program ABAQUS. Shell elements having nonlinear concrete and reinforcement properties have been used for both 2D and 3-D pushover analyses. A sensitivity study was carried out in order to obtain the appropriate mesh size for the models and a more rened mesh was used around the opening area. The same FE mesh was retained for the nonlinear dynamic analysis (NDP). In general, ABAQUS provides advanced nonlinear concrete and reinforcement modeling features compared with some other commercial FE programs. A more detailed description of the FE modeling of the stacks can be found elsewhere [9].

120

100

80 Height (m)

60
First Mode Uniform Triangular ELF SRSS(YPTx) SRSS(YPTy)

40

20

0 0.0E+00

5.0E+04 1.0E+05 Lateral Force (N)

Fig. 6. Tpras stack pushover load patterns.

2. 2-D pushover analysis Nonlinear dynamic analysis has traditionally been regarded as the most precise method for estimating the response of a structure to a particular seismic event. Since the nonlinear properties of the elements of the structure are considered when modeling the structure, it should represent the behavior of the structure at each time during and after the onset of the ground motion. Therefore this method has been accepted as the benchmark for seismic structural analysis. However, because of the following reasons, it is generally used only for the evaluation of special, complicated, and important structures: (1) material modeling can be complex. For example, the hysteretic behavior modeling of the concrete components under earthquake loading is complicated and involves stiffness degradation, strength deterioration, etc.; (2) selection of an appropriate ground motion or suite of ground motions is vital to the analysis. The nonlinear response of the structure is extremely sensitive to the frequency contents and phasing properties of the earthquake motions; (3) the analysis is time consuming and costly since it must be performed for small intervals of time to accurately follow the nonlinear response of the structure. Additionally, a number of runs may be required to get decisive results for a suite of earthquake inputs. The NSP in ATC-40 [8] is based on the capacity spectrum method, which has become a standard technique for the evaluation of existing structures due to its utility in predicting inelastic dynamic performance. This method requires that both the capacity of the structure, derived from a pushover analysis, and the demand, given by a selected ground motion spectrum, be compared in the spectral acceleration vs. spectral displace-

ment (ADRS) domain. The NSP results for the Tpras stack were directed toward a plausible quantitative explanation of the unique failure of the stack. In the 2-D pushover analysis studies [3,9], the rst step was to evaluate the collapsed Tpras stack and identify the possible failure mode during the earthquake. The second step was to apply the different lateral load patterns commonly considered, including rst mode distribution, multimode combination (SRSS) distribution, uniform distribution, triangular distribution, and equivalent lateral force (ELF) distribution as well as the recently developed modal pushover procedure (MPA) [10]. In the MPA procedure, the seismic response of the structure due to each mode is determined by pushing the structure to its modal target displacement using an invariant modal lateral force distribution. Then, using an appropriate modal combination rule, e.g. the SRSS rule, the overall peak responses of the structure is estimated by combining the peak response for each mode. All of these pushover load patterns are shown in Fig. 6. The analysis then compares the corresponding seismic demand to the exact nonlinear dynamic response history analysis (NL RHA) results in order to evaluate the capabilities of the various patterns to represent higher mode effects and possible redistribution of inertial forces in a structure due to nonlinearity under earthquake load. Since the stack is asymmetric due to the large opening and since the real seismic loading condition for the structure is a 3-D earthquake input, the 3-D interaction effects may not be negligible. Therefore, the third step is to extend the 2-D pushover analysis to a 3-D procedure to account for the 3-D interaction effects. This step is addressed in the next sections. The results for the rst two steps of the pushover analysis stated above can be found elsewhere [3,9]. Some of the conclusions are summarized as follows: 1. A 2-D pushover analysis produced a variety of demand capacity comparisons, which revealed that the strength of the stack, including the effect of the opening, at best barely

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887

883

met the demand and in some cases fell short. Phase-byphase cracking patterns for several models under different loading directions showed clearly that the corresponding development of cracking patterns changed the capacities signicantly. The results provided a plausible explanation for the failure in the direction 90 from the opening, which is also consistent with the observed debris eld. 2. For the 2-D pushover analysis of the model without the opening, the uniform distribution underestimated the target displacement by up to around 40%. The rst mode, ELF, and triangle distributions gave fair predictions, with errors around 1020%. Taking into account the higher mode effects, the SRSS distribution showed the ability to predict the target displacement within about 10%, and the MPA procedure provided the best estimation of the target displacements, for most cases within 10%. As for the peak deections, once again, the MPA procedure and SRSS distribution provided better estimations than the other distributions. 3. For the 2-D pushover analysis of the model with the opening pushed 90 to the opening direction shown in Fig. 5, the SRSS distribution gave better predictions for failure displacements than other distributions, with around 10% error.

Fig. 7. 3-D pushover load pattern.

3. 3-D pushover analysis 3. 3.1. A new 3-D pushover analysis procedure In traditional pushover analysis, only the distribution of forces equivalent to those produced by earthquake action in one direction is applied to the structure to represent the inertia forces experienced during the earthquake. This procedure has provided insightful results for symmetric structures. But for asymmetric structures, pushover analysis considering two directional earthquake inputs may be more appropriate, since the structure has different dynamic properties in each direction. For the Tpras stack, with the large opening at the 30 m level, the stack would have undergone different lateral motions simultaneously and the 3-D interaction effects may not be negligible. There is very little research focusing on improving the pushover analysis by considering 3-D interaction effects [11,12], so the need for developing improved pushover analysis procedures considering 3-D interaction effects for asymmetric structures is evident. In this study, a new 3-D pushover analysis method is proposed to extend the traditional 2-D pushover procedure for the analysis of the asymmetric Tpras stack. The validity of the proposed method will be assessed by comparing the results with those from an exact 3-D step-by-step nonlinear dynamic analysis. The basic procedure is as follows: 1. Carry out a 3-D modal analysis using a FE model with the initial geometry and material properties. Obtain the natural frequencies and fundamental modes for each direction. 2. Now, two types of lateral load patterns may be selected based on the patterns shown in Fig. 6. One type is a fundamental

4.

5.

6.

7.

8.

mode, usually Mode 1, and the other type may be one of the patterns in Fig. 6 other than Mode 1. For a lateral load pattern other than the fundamental mode patterns, apply the lateral forces to the structure, and perform the pushover analysis for each direction. Plot the pushover curves in the spectral displacement vs. spectral acceleration domain (ADRS). The equivalent SDF (single degree of freedom) period for the lateral load pattern in each direction is then taken as the initial secant for the pushover curve before yielding. For each direction, given the fundamental frequencies for the fundamental modes and equivalent SDF system frequencies for the other load patterns, locate the corresponding spectral acceleration values from the response spectrum in each direction. (In this case, the longitudinal and transverse directions of the YPT spectrum.) Apply two directional lateral forces for each load pattern to the structure as illustrated in Fig. 7, proportional to the spectral acceleration values obtained from Step 4. For each load pattern, perform the 3-D pushover analysis using the lateral load forces described in Step 5, and plot the capacity curve for each direction. Compare the capacity curves with the smoothed mean demand curves of the spectra for each direction to obtain the target displacement of the structure for the various load patterns. Determine the response over the height of the structure using the 3-D pushover analysis results for the selected patterns at the respective target displacements.

The validity of the proposed method will be assessed by comparing the results with a 3-D nonlinear dynamic analysis of the stack.

884

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887

Table 1 Target displacements for the model without an opening

Method 3-D pushover Disp. (m) Mode 1 Uniform ELF Triangle SRSS MPA NL RHA 0.522 0.416 0.680 0.701 0.662 0.565 0.601 Error (%) 13.1 30.8 13.1 16.6 10.1 6.0 0 2-D pushover Disp. (m) 0.498 0.316 0.492 0.476 0.629 0.528 Error (%) 17.1 47.4 18.1 20.8 4.7 12.2

120 100 80 Height (m) 60 40 20 0 0.00

NL RHA Mode 1 Uniform ELF Triangle SRSS MPA

3.2. 3-D pushover analysis results First, the 3-D pushover procedure is applied to the model without an opening and compared to 3-D nonlinear dynamic analysis with two directional inputs. Then the procedure is extended to predict the failure of the model with the opening. The failure analysis for the model with the opening is carried out by 3-D nonlinear dynamic analysis as well. Once again, the records used in the nonlinear dynamic analysis are from a suite of simulated earthquakes based on the YPT record [4]. 3.2.1. Model without an opening 3.2.1.1. Target displacement Similar to the traditional 2-D pushover analysis, target displacements are calculated by the proposed 3-D pushover analysis based on different lateral load patterns as well as the MPA procedure. These target displacements using the 3-D and 2-D pushover analysis as given in Table 1 are the magnitudes of the displacement of the stack at the top. As shown in the table, 3-D pushover analysis results are based on the YPT earthquake input in both directions while 2-D results are based on YPTy earthquake input in one direction. The 3-D results are calculated by combining the target displacements for each direction. The errors are obtained by comparison to the NL RHA results. As seen in Table 1, for the 3-D pushover analysis results the error from the uniform distribution is the largest, while the Mode 1 distribution, ELF distribution, and triangle distribution errors are less. Taking into account the higher mode effects, the SRSS distribution gives a good prediction for the target displacement and the error from the MPA procedure is less than 10%. In general, the new 3-D pushover analysis provides better estimations for the target displacement when compared to 2-D pushover analysis since the pushover load patterns are simulating the earthquake inputs in both directions. 3.2.1.2. Peak deections The peak deections over the height of the stack for each direction are obtained from the structural deection at the target displacement in that direction, as determined by the capacity-demand comparison. The mean values of the peak response from the nonlinear dynamic analysis results based on a suite of simulated records are compared with the pushover results.

0.10

0.20

0.30 0.40 0.50 Displacement (m)

0.60

0.70

0.80

Fig. 8. 3-D peak deections for the model without an opening.

120 100 80 Height (m) 60 40 20 0 -10.0 0.0 Error (%)

NL RHA Mode 1 Uniform ELF Triangle SRSS MPA

-40.0

-30.0

-20.0

10.0

20.0

30.0

Fig. 9. 3-D peak deection errors for the model without an opening.

The differences between the estimated values using 3-D pushover analysis and exact values from 3-D nonlinear dynamic analysis are shown in Figs. 8 and 9. As seen in Fig. 8, the MPA procedure and SRSS distribution give better estimates of the peak deections during the earthquake than the others, while the uniform distribution underestimates the total response up to 30%. In general, the 3-D pushover analyses predict the target displacements and the peak deections of the structure fairly well, as compared to 3-D nonlinear dynamic analysis. The MPA procedure and SRSS distribution using 3-D pushover analysis provide good estimates, especially for the dynamic response of the structure under two directional earthquake inputs that are not easily predicted using traditional 2-D pushover analysis. The ability of the procedure to predict the failure of the stack will be discussed in the following section.

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887
Table 2 3-D failure displacements for the model with the opening
Pattern YPT mean Failure Disp. (m) Mode 1 Uniform ELF Triangle SRSS NL RHA 0.667 0.745 0.646 0.645 0.597 0.525 Error (%) 27.0 41.8 23.1 22.8 13.8 0

885

3.2.2. Model with the opening Since the stack failed in an earthquake having different lateral loading components acting simultaneously, it is appropriate to analyze the structure in multiple directions. The failure displacement and the cracking pattern recorded at the failure point from 3-D nonlinear dynamic analysis will be used to validate the 3-D pushover procedure. The 3-D pushover procedure was applied to the model with the opening, where the lateral load patterns in the two directions are proportional to the response spectrum values based on the equivalent SDF system. Also, a 3-D nonlinear dynamic analysis was carried out using two directional inputs, a suite of YPT longitudinal records in the direction 0 to the opening, and a suite of YPT transverse records in the direction 90 to the opening, based on the orientation of the opening from the site reference. The failure displacement at the top and the cracking patterns are compared between the 3-D pushover analysis prediction and the 3-D nonlinear dynamic analysis results. Because the pushover analysis results for the three fundamental modes cannot be combined for failure analysis since the failure displacement is determined for each mode by pushing the structure to its capacity in that mode, the MPA procedure is not considered here. 3.2.2.1. Failure displacement Incremental lateral loads in two directions for different loading patterns were applied on the structure until failure. The magnitudes of the top displacement at the point of failure predicted by the different pushover patterns are shown in Table 2. The errors relative to the 3-D NL RHA are listed as well. As shown in Table 2, where results taking into account higher mode effects in both directions are summarized, the SRSS distribution provides the best prediction, with less than 14% error. 3.2.2.2. Cracking pattern In Figs. 10 and 11, the cracking patterns for 3-D pushover analysis using different load patterns are plotted at the failure. On the coarse plots of Figs. 10 and 11, vertical and diagonal cracks superimposed on the horizontal exural cracks cause the dense crack regions. The failure cracking pattern for the nonlinear dynamic result is shown as NL RHA in Fig. 11. In the failure cracking pattern from nonlinear dynamic analysis, there are more long critical shear cracks around the opening area than there are exural cracks along the height. This nd-

Fig. 10. 3D cracking patterns for failure of the model with the opening.

ing conrms the initial prediction by 2-D pushover analysis; the critical shear cracks developed at the opening area caused the stack to fail during that earthquake. The cracking patterns from 3-D pushover analysis show the existence of the critical shear cracks around the top left and bottom right corner of the opening. Considering the limitation of monotonic loading, we would expect a symmetric cracking pattern for the other direction, so the overall cracking patterns around the opening under cyclic loading match well with the nonlinear dynamic analysis results. Even though all lateral patterns give good estimations at the opening level, the SRSS distribution, by taking into account the higher mode contribution, better predicts the cracks developed from the opening level to about the 65 m level. 4. Conclusions Using a demandcapacity comparison, a nonlinear static pushover analysis was used to investigate the collapsed Tpras stack. The demand was represented by an acceleration displacement response spectrum based on the YPT record motion as well as some smoothed adaptations typical of design spectra. The capacities were calculated from pushover curves using a nonlinear reinforced concrete FE analysis. A new 3-D pushover analysis procedure was proposed and the results

886

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887

eral patterns were in an acceptable range. The SRSS distribution resulted in the lowest error, between 10% and 20%. All of the lateral load patterns successfully captured the shear cracks developed around the opening, along with exural cracks. 3. From the failure cracking pattern for the 3-D nonlinear dynamic analysis, there were more long critical shear cracks around the opening area than there were the exural cracks along the height. This conrmed the initial prediction by 2D pushover analysis that the critical shear cracking around the opening area, along with the concentrated exural cracking, was prominent in the failure. The 3-D nonlinear dynamic analysis results conrmed that the Tpras stack could not survive the YPT earthquake inputs under both directions. Acknowledgments The authors wish to thank the United States National Science Foundation for the support of this study under Grant CMS-0084737. Additionally the generous support of our ofcial Turkish collaborator, Prof. Dr. Semih S. Tezcan and his colleague Professor Sami A. Kilic is appreciated. The cooperation of the designer of the Tpras stack, Mr. T. Tunca was indispensable to our effort and the consent of General Manager H. Danis and other ofcials of the renery to make the photos, calculations and drawings available to us were generous. References
Fig. 11. 3D cracking patterns for failure of the model with the opening (continued). [1] P.L. Gould, W. Huang, R. Martinez, G.S. Johnson, Investigation of the collapse of a heater stack during the Izmit (Kocaeli) Turkey earthquake of August 17, 1999, in: Proceedings of the 7th US National Conference on Earthquake Engineering, Boston, MA, July 2002. Also presented in: Proceedings of the 12th European Conference on Earthquake Engineering, London, UK, September 2002. [2] P.L. Gould, W. Huang, G.S. Johnson, Nonlinear analysis of a collapsed stack, in: Proceedings of the 13th World Conference on Earthquake Engineering, Vancouver, British Columbia, Canada, August 2004. [3] W. Huang, P.L. Gould, R. Martinez, G.S. Johnson, Nonlinear analysis of a collapsed reinforced concrete chimney, Earthquake Eng. Struct. Dyn. 33 (2004) 485498. [4] G.I. Schuller, H.J. Pradlwarter, Estimation of the evolutionary process of strong nonstationary earthquake records, in: Proceedings of the 9th World Conference on Earthquake Engineering, Tokyo-Kyoto, Japan, vol. 2, 1988, pp. 777782. [5] S. Kim, E. D Amore, Push-over analysis procedure in earthquake engineering, Earthquake Spectra 15 (3) (1999) 427434. [6] H. Krawinkler, G.D.P.K. Seneviratna, Pros and cons of a pushover analysis of seismic performance evaluation, J. Eng. Struct. 20 (46) (1998) 452464. [7] R.S. Lawson, V. Vance, H. Krawinkler, Nonlinear static pushover analysisWhy, when and how?, in: Proceedings of the 5th US Conference on Earthquake Engineering, Chicago, IL, vol. 1, 1994, pp. 283292. [8] ATC-40, Seismic Analysis and Retrot of Concrete Buildings, vol. 1, Applied Technology Council, Redwood City, CA, November 1996. [9] W. Huang, Nonlinear analysis of a collapsed reinforced concrete chimney, Doctoral Thesis, Washington University, Saint Louis, MO, May 2005. [10] A.K. Chopra, R.K. Goel, A modal pushover analysis procedure for estimating seismic demands for buildings, Earthquake Eng. Struct. Dyn. 31 (2002) 561582.

were compared with those from a nonlinear dynamic analysis. Based on these pushover analyses, some of the conclusions are summarized as follows: 1. A new 3-D pushover analysis procedure was proposed and applied to models of chimneys with and without an opening. Various lateral load patterns were considered. For the target displacement of the model without the opening, the error from the uniform distribution was the largest, while the Mode 1 distribution, ELF distribution, and triangle distribution provided somewhat better estimates. The SRSS distribution gave a good prediction, with an error around 10% and the error from the MPA procedure was even less than 10%. As to the peak deections, the MPA procedure and SRSS distribution provided the best estimates, while the uniform distribution underestimated the total response by up to 30%. The Mode 1 distribution, ELF distribution, and triangle distribution gave similar estimates. Compared to a 2-D pushover analysis, the new 3-D pushover analysis procedure provides a better estimation for target displacements. 2. For the 3-D pushover analysis on the model with the opening, the failure displacements predicted using different lat-

W. Huang, P.L. Gould / Finite Elements in Analysis and Design 43 (2007) 879 887 [11] A.G. Ayala, E.A. Travera, A new approach for the evaluation of the seismic performance of asymmetric buildings, in: Proceedings of the 7th US National Conference on Earthquake Engineering, Boston, MA, July 2002 (CD-ROM).

887

[12] A.S. Moghadam, W.K. Tso, Pushover analysis for asymmetrical multistory buildings, in: Proceedings of the 6th US National Conference on Earthquake Engineering, Seattle, EERI, Oakland, 1998.