You are on page 1of 13

THE STRUCTURAL DESIGN OF TALL AND SPECIAL BUILDINGS Struct. Design Tall Spec. Build.

20, 164176 (2011) Published online 12 May 2009 in Wiley Online Library (wileyonlinelibrary.com/journal/tal). DOI: 10.1002/tal.520

Evaluation of peak structural responses based on consistent elastic and inelastic design spectra
Scott Swensen1 and Kevin Wong2,*,
Department of Civil & Environmental Engineering, University of Utah, Salt Lake City, Utah, USA

SUMMARY Many structures built in seismically active regions are expected to behave nonlinearly during an earthquake event. Because of the complex and cumbersome nature of nonlinear time history structural analysis, more recent research has been focused on a way of producing inelastic response spectra that will accurately estimate the peak structural responses. In this paper, peak acceleration and displacement values for three different two-dimensional steel frames were determined using linear and nonlinear time history analysis, and the results were then compared with peak responses found using elastic and inelastic response spectrum analysis. In order to compare the accuracy of these methods, 12 ground motions were adjusted using spectral matching to produce 12 different ground motion histories with nearly identical 5% damped elastic design spectra for the University of Utah location in Salt Lake City, as specied by the United States Geological Survey. It was determined that linear response spectrum analysis proves to be a fairly accurate estimate of time history peak responses. While inelastic response spectrum analysis was found to provide a somewhat less accurate estimation of time history peak responses, the method still shows promise as a simpler and less computationally intensive alternative to time history analysis. Copyright 2009 John Wiley & Sons, Ltd.

NOMENCLATURE a(t) = reference time history fj(t) = set of adjustment functions Fa = short-period site coefcient Fv = long-period site coefcient Fmax,e = elastic strength of SDOF system Fy = yield strength of SDOF system R = strength-reduction factor Ri = absolute value of the peak response S1 = mapped maximum considered earthquake, 5% damped, spectral response acceleration parameter at a period of 1 s SS = mapped maximum considered earthquake, 5% damped, spectral response acceleration parameter at short periods t = time T = natural period of the system um = maximum displacement uy = yield displacement Vy = yield strength of MDOF system a = strain hardening ratio da(t) = adjustment time history

* Correspondence to: Kevin Wong, Department of Civil & Environmental Engineering, University of Utah, 122 S. Central Campus Dr., Salt Lake City, UT 84112 E-mail: kfwong@civil.utah.edu Copyright 2009 John Wiley & Sons, Ltd.

EVALUATION OF PEAK STRUCTURAL RESPONSES

165

1. INTRODUCTION Since the early 1930s, response spectrum analysis has been widely used to characterize ground motions and predict their inuence on structures. The simplicity of response spectrum analysis made it the primary means of performing linear seismic analysis before the widespread use of computers to carry out intensive time history analysis. Time history analysis is particularly burdensome for inelastic systems, in that rigorous computation is needed and many plausible ground motions should be tested in order to accurately gauge strength demand. The straightforward method of response spectrum analysis, however, has also become the standard means adopted by building code agencies to characterize acceptable design behaviour and predict ground motions in different areas of the world. The response spectrum analysis method does, however, have shortcomings. Because most structures cannot be modelled as single-degree-of-freedom (SDOF) systems, linear response spectrum analysis relies on approximations used when combining modal combinations to nd the total peak responses (Chopra, 2007). For structures responding in an inelastic manner, the challenges only increase. While the response spectra used to determine inelastic structural behaviour for SDOF systems can be easily estimated by reducing the elastic spectrum by a reduction factor, multi-degree-of-freedom (MDOF) systems require a much more intensive analysis to accurately estimate these peak responses (Nasser and Krawinkler, 1991). If response spectrum analysis results are proven to closely estimate the peak responses found using time history analysis, then computational time history analyses would likely be unnecessary for design purposes. In this paper, the accepted means of determining peak acceleration and displacement responses by means of elastic and inelastic response spectrum methods are compared to peak responses calculated through computational time history analysis. 2. ORIGINAL GROUND MOTION RECORDS Twelve ground motion records were selected from the Pacic Earthquake Engineering Research Strong Motion Database to be used in the analysis. These 12 records were chosen because they have similar peak accelerations and durations. These records were also deemed favourable because they have response spectra that are similar to a target spectrum to be matched. All of the records also have the same time step (0.005 s). The selected original time histories and their properties are shown in Table 1. 3. GROUND MOTION RECORD ADJUSTMENT TO TARGET RESPONSE SPECTRUM The selected ground motion histories in Table 1 were adjusted to create 12 different records with similar 5% damped elastic response spectra. This was done in order to compare the peak structural responses obtained through time history analysis when the peak responses obtained through response

Table 1. Ground acceleration time histories. Earthquake


Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Lakes Morgan Hill Superstition Hills Westmorland Whittier Narrows

Location
Chalfant Valley, Mono, CA Chichi, Nantou, Taiwan Coalinga, CA Coalinga, CA Coyote Lake, Santa Clara, CA Imperial Valley, CA Landers, California Loma Prieta Peak, CA Mammoth Lakes, CA Morgan Hill, CA Superstition Hills, CA Westmorland, CA South El Monte, CA

Date
7/21/86 9/20/99 5/2/83 5/2/83 8/6/79 10/15/79 6/28/92 10/18/89 5/25/80 4/24/84 11/24/87 4/26/81 10/1/87

Time
14:42 23:42 23:42 17:05 23:16 11:58 0:05 16:34 4:24 13:16 12:09 14:42

Sensor
CDMG 54099 CWB USGS 1162 USGS 1162 CDMG 57383 USGS 5054 SCE 24 CDMG 57007 CDMG 54099 CDMG 47380 CDMG 01335 CDMG 5169 CDMG 24461

Direction from north ()


0 0 45 135 230 140 275 0 90 0 0 90 180

Copyright 2009 John Wiley & Sons, Ltd.

Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

166

S. SWENSEN AND K. WONG

1.8 1.6 1.4

Acceleration (g)

1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2 2.5 3

Period (s)

Figure 1. Target acceleration response spectrum.

spectrum analysis are relatively comparable between all 12 cases. In order to examine the effects of a design earthquake on a steel frame at the University of Utah location in Salt Lake City, the design response spectrum for this location was obtained. Using the Java Ground Motion Parameter Calculator from the United States Geological Survey, it was determined that the design response spectra for the mentioned location should have an average SS of 1.584 and an average S1 of 0.627 with the constants Fa and Fv both equal to 1.0. Using the methodology outlined in ASCE 7-05 (ASCE, 2006), the target acceleration response spectrum was constructed. This target spectrum is shown in Figure 1. To adjust the chosen ground motion records in achieving unique ground motion records with this target response spectrum, the program RSPMATCH was utilized. Earlier programs performed spectral matching by adjusting the Fourier amplitude spectrum while keeping the phase of the reference time history constant. This approach, however, can be undesirable, because the non-stationary properties of the ground motion records can be drastically changed, and convergence characteristics are often undesirable. RSPMATCH is instead designed to apply an algorithm to the time history that adds wavelets to the time history in an effort to adjust the response spectrum to match the target values. While this spectral matching method is more computationally intensive than simply adjusting the Fourier amplitude spectrum, convergence properties are more favourable and the non-stationary properties of the original time histories are preserved (Lilhanand and Tseng, 1987, 1988). Based on research by Abrahamson (1992), the program produces an adjusted time history, da(t), based on an input adjustment function fj(t). The adjustment function can be modelled after a reverseorder oscillator impulse response or a tapered cosine wave; the selection of the adjustment function depends on the desired form of the output transient time history. Samples of these ground motion adjustment functions and their resulting responses with frequencies of 5 Hz are shown in Figures 2 and 3. In order for the addition of the adjustment function to cause the desired adjustment, it must be placed in the reference time history so that the timing of fj(t) coincides with the peak acceleration response of the time history a(t). Utilizing RSPMATCH, these wavelets are added to the reference time history and the resulting response spectrum is compared with the target spectrum. For the purposes of this project, the resulting time history was set to end analysis when each of the sampled points of the target response spectrum was matched to a tolerance of not more than 5%. A constant damping value of 5% was utilized. Baseline correction was also applied after each analysis step in order to assure that the resulting time history records would not display residual ground displacement. Though RSPMATCH is successful in adjusting the majority of ground motion records to match a given specic response spectrum as long as spectral content and ground motion magnitudes are comparable, some resulting output time histories did not resemble possible real-world earthquake records. This could have occurred in records where the reverse-order oscillator impulse adjustment function was utilized, because the motion in this function stops abruptly (see Figure 2(a)), which is not congruent to the
Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

EVALUATION OF PEAK STRUCTURAL RESPONSES

167

(a) Acceleration (g)


Acceleration (g)
0 2 4 6 8 10
0

(b)

10

Time (s)

Time (s)

Figure 2. (a) Reverse-order oscillator impulse ground motion and (b) response.

(a) Acceleration (g)

(b) Acceleration (g)

10

10

Time (s)

Time (s)

Figure 3. (a) Tapered cosine wave ground motion and (b) response.

1.8 1.6 1.4


Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows

Acceleration (g)

1.2 1 0.8 0.6 0.4 0.2 0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

Period (s)

Figure 4. Response spectra of adjusted ground motions.

behaviour of observed strong ground motions. Because of this reason, those performing this type of response spectrum matching should expect to abandon several adjusted ground motion records even if the desired response spectrum is achieved. Figure 4 shows the elastic response spectra of the 12 ground motion records that were successfully adjusted. It is observed that the adjusted ground motion records for all 12 cases provide acceleration response spectra very similar to the target design spectrum for the University of Utah location. All 12 adjusted records also display peak ground accelerations of about 0.8 g.
Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

168

S. SWENSEN AND K. WONG

4. PROPERTIES OF ANALYSED STRUCTURES Next, time history analyses and response spectrum analyses were performed on three separate momentresisting steel frames for all adjusted earthquake records. For each frame, the steel was modelled to have a yield strength of 248.2 MPa and an ultimate strength of 399.9 MPa. For nonlinear analysis, plastic hinges were simulated at the end of each member. When the yield moment is achieved for a certain member, the hinge will behave in a perfectly plastic manner with a strain-hardening ratio of a = 0. No other nonlinear features were considered. The rst frame is a six-storey three-bay steel frame as shown in Figure 5. Each oor supports a mass of 217.72 Mg, which is distributed with one-third of each storey mass at each interior joint and one-sixth of each storey mass at each external joint. This structure will be referred to as Frame 1. Two 10-storey frames were also examined, as shown in Figure 6. For the rst 10-storey frame, the mass at each storey is 435.45 Mg, which is distributed with one-third of each storey mass at each interior joint and one-sixth of each storey mass at each external joint. This frame will be referred to as Frame 2. Another 10-storey frame was then modelled with the mass at each storey being 544.31 Mg. This storey mass is distributed in a similar manner as that of Frames 1 and 2. This frame will be referred to as Frame 3. Note that except for the nodal masses, Frames 2 and 3 are identical. The three structures were then analysed using SAP2000 to determine the natural periods of each system. The natural periods of each of the three systems determined using modal analysis are shown in Tables 2 and 3.

W27x94 36
W14x342 W14x193
35

W27x94 38
W14x342
37

W27x94 40
W14x193
39

Roof EL=26.21 m

x6
6th Floor EL=21.95 m

W36x135
29

W36x135

W36x135

30 31

32 33

34

x5
5th Floor EL=17.68 m

W36x150 24
W14x257 W14x455
23

W36x150 26
W14x455
25

W36x150 28
W14x257
27

x4
4th Floor EL=13.41 m

W36x210
17

W36x210

W36x210

18 19

20 21

22

x3
3rd Floor EL=9.14 m

W36x210
11

W36x210
12 13 14 15

W36x210
16

x2
2nd Floor EL=4.57 m

W36x210 6
W14x283 W14x500
5 7

W36x210 8
W14x500
9

W36x210 10
W14x283

x1
4 Ground EL=0 m

7.62 m

7.62 m
i

7.62 m

xi - ith Degree of Freedom

- i th Plastic Hinge Location

Figure 5. Frame 1 conguration.


Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

EVALUATION OF PEAK STRUCTURAL RESPONSES

169

W27x94 60
W14x342 W14x193

W27x94 61 62

Roof 63 W27x94 64 EL=42.67 m


W14x342 W14x193

59

x10
10th Floor EL=38.40 m

W36x135 53

W36x135

W36x135

54 55

56 57

58

x9
9th Floor EL=34.14 m

W14x426

W14x426

W14x257

W14x257

51 W36x150 52 W36x150 W36x150 48 47 49 50

x8
8th Floor EL=29.87 m

W36x182 41

W36x182 W36x182 46 44 45 42 43

x7
7th Floor EL=25.60 m

W14x455

W14x283

W14x455

W14x283

W36x210 35

36 37

W36x210 38

39 W36x210 40

x6
6th Floor EL=21.34 m

W36x210 W36x210 W36x210 32 33 30 31 34 29

x5
5th Floor EL=17.07 m

W14x550

W14x311

W14x550

W14x311

27 W36x230 28 W36x230 24 W36x230 25 23 26

x4
4th Floor EL=12.80 m

W36x230 17

W36x230

W36x230

18 19

20 21

22

x3
3rd Floor EL=8.53 m

W14x370

W14x605

W14x605

W14x370

15 W36x260 16 W36x260 12 W36x260 11 13 14

x2
2nd Floor EL=4.26 m

W36x260 5

W36x260

W36x260

6 7

8 9

10

x1
Ground EL=0 m

1 7.62 m

2 7.62 m
i

3 7.62 m

xi - ith Degree of Freedom

- i th Plastic Hinge Location

Figure 6. Frame 2 and Frame 3 conguration.


Table 2. Modal periods for Frame 1. Mode of vibration
1 2 3 4 5 6
Copyright 2009 John Wiley & Sons, Ltd.

Period (s)
1.171 0.422 0.241 0.172 0.134 0.113
Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

170

S. SWENSEN AND K. WONG

Table 3. Modal periods for Frames 2 and 3. Mode of vibration (Frame 2)


1 2 3 4 5 6 7 8 9 10

Period (s)
2.418 0.866 0.517 0.362 0.277 0.234 0.227 0.216 0.197 0.189

Mode of vibration (Frame 3)


1 2 3 4 5 6 7 8 9 10

Period (s)
2.696 0.965 0.576 0.403 0.308 0.260 0.253 0.241 0.220 0.211

5. LINEAR ANALYSIS The 12 adjusted ground motion records were then applied to each of the three structures using SAP2000 to determine the time history response of each case. The individual response spectra shown in Figure 4 were then used to carry out response spectra analysis using SAP2000 on the same three frames. The maximum spectral acceleration and displacement at the roof level were then extracted and compared. The time history analysis used by SAP2000 implements the method of mode superposition (Chopra, 2008) to determine accelerations and displacements. For the linear response spectra analyses, the complete quadratic combination (CQC) method was used to add the contribution from each mode. This method was used because the alternate absolute sum modal combination has been shown to overestimate peak displacements when compared to the CQC method (Chopra and Goel, 2003). The output from the linear time histories and response spectra results indicating the maximum absolute accelerations and relative displacements for each earthquake cases are summarized in Table 4. This data shows that for linear cases, the maximum acceleration and displacement values found using response spectrum analysis are very similar to those found using time history analysis. As expected, however, the maximum responses of the time history analysis cases show more variation than those maximum responses found using response spectrum analysis. The maximum standard deviation of displacement response is 0.78 cm for response spectra analysis, as compared to the maximum standard deviation of 3.98 cm for time history analysis. The maximum accelerations and displacements at the roof level for each case were compared to determine if results from the time history analyses and response spectra analysis yield similar results. The displayed difference in the values indicates the difference between the two values in relation to the time history analysis value. In other words, a negative difference value indicates that the time history value determined is greater than the response spectrum value, while a positive difference indicates that the time history value is less than the response spectrum value. From the results of linear time history and response spectrum analyses, it is noted that the maximum absolute displacement values are fairly similar. All response spectrum values differing from the time history values by less than 12% when compared to the time history displacements, which is as expected due to the approximation in modal combination method used in the response spectrum analyses. Less agreement, however, is seen in the acceleration values. Some response spectrum maximum acceleration values differ from the time history results by more than 30%. It is also noted that for the majority of the cases, the peak accelerations determined by response spectrum analysis are smaller than the peak accelerations calculated using time history analysis. 6. NONLINEAR ACCELERATION RESPONSE SPECTRA The adjusted nonlinear response spectra for a MDOF system were then calculated by extending the nonlinear analysis of SDOF based on ductility demand. For SDOF nonlinear systems, the ductility
Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

Table 4. Linear analysis data.


Six-storey frame linear analysis (Frame 1) Maximum displacement (cm) TH analysis 26.61 24.13 24.95 24.40 22.65 24.95 25.11 23.89 25.19 24.44 26.11 24.99 24.786 1.022 24.54 24.49 24.76 25.34 24.99 24.56 24.62 24.76 24.28 24.54 24.58 24.72 24.682 0.272 7.78 1.50 0.79 3.85 10.30 1.55 1.91 3.64 3.63 0.38 5.86 1.07 1.056 0.991 1.067 0.968 1.048 1.045 0.932 0.963 0.930 0.879 1.187 0.929 1.000 0.085 0.942 0.900 0.923 0.949 0.938 0.940 0.894 0.935 0.919 0.929 0.897 0.844 0.918 0.030 25.93 22.37 22.87 24.18 22.71 23.86 20.76 23.42 23.04 24.43 20.41 22.23 23.016 1.530 10.76 9.12 13.48 1.91 10.47 10.12 4.11 2.87 1.16 5.71 24.46 9.14 7.658 7.591 21.69 20.13 21.32 22.98 21.66 22.15 19.48 21.83 21.22 21.96 19.47 19.58 21.123 1.174 RS analysis % difference TH analysis RS analysis % difference TH analysis RS analysis Maximum acceleration (g) RS analysis 1.056 1.053 1.059 1.062 1.064 1.055 1.057 1.059 1.046 1.055 1.054 1.042 1.055 0.006 10-storey frame linear analysis (Frame 2) 0.712 0.711 0.721 0.716 0.730 0.707 0.717 0.713 0.709 0.717 0.705 0.715 0.715 0.007 10-storey frame linear analysis (Frame 3) 4.73 22.43 26.37 2.20 8.70 9.00 11.09 16.04 14.51 25.79 21.84 7.93 7.874 14.876 46.89 57.13 52.37 53.72 50.23 46.55 51.34 50.17 51.64 53.80 53.20 53.77 51.735 3.004 51.84 51.92 51.31 52.41 51.81 51.67 52.76 51.82 51.16 51.71 51.87 52.13 51.868 0.431 10.57 9.12 2.01 2.44 3.15 11.00 2.76 3.29 0.94 3.88 2.49 3.05 0.568 5.893 0.509 0.593 0.555 0.572 0.584 0.597 0.550 0.647 0.474 0.508 0.639 0.563 0.566 0.052 0.634 0.619 0.602 0.638 0.629 0.638 0.625 0.630 0.591 0.636 0.609 0.632 0.624 0.015 25.90 0.66 6.34 13.85 1.89 17.47 0.65 6.30 3.02 0.90 30.81 6.47 % difference Six-storey frame nonlinear analysis (Frame 1) Maximum displacement (cm) % difference 16.35 10.03 6.77 4.95 4.62 7.15 6.17 6.78 7.89 10.08 4.57 11.90 8.105 3.474

Maximum acceleration (g)

Copyright 2009 John Wiley & Sons, Ltd.


10-storey frame nonlinear analysis (Frame 2) 24.54 4.38 8.59 11.70 7.63 6.83 13.64 2.62 24.73 25.24 4.76 12.26 11.013 9.981 40.11 57.11 51.58 53.70 48.92 48.01 43.10 47.59 43.38 61.59 52.56 44.56 49.350 6.269 45.62 45.75 44.05 46.61 46.57 46.54 43.52 41.20 45.77 46.18 45.78 46.70 45.357 1.646 10-storey frame nonlinear analysis (Frame 3) 0.658 0.664 0.649 0.649 0.665 0.637 0.657 0.655 0.653 0.654 0.649 0.653 0.654 0.007 7.40 27.85 26.45 15.95 6.29 17.82 15.10 17.29 2.69 28.97 22.79 10.47 12.114 14.719 54.44 67.17 55.84 63.56 58.79 58.38 62.77 55.27 63.01 58.96 55.49 58.31 59.332 3.982 56.98 59.46 57.90 58.24 57.99 57.40 58.87 57.87 58.29 57.28 57.41 59.21 58.076 0.781 4.67 11.48 3.69 8.36 1.36 1.68 6.22 4.70 7.49 2.84 3.46 1.55 1.780 5.586 0.465 0.493 0.425 0.520 0.499 0.487 0.429 0.554 0.437 0.437 0.553 0.485 0.482 0.045 0.558 0.589 0.566 0.544 0.578 0.562 0.587 0.574 0.559 0.577 0.574 0.572 0.570 0.013 19.88 19.35 33.01 4.66 15.88 15.26 36.92 3.74 27.84 31.89 3.74 17.81 19.166 11.515 46.34 69.38 54.90 61.77 60.02 58.53 46.38 53.63 42.53 68.14 62.12 53.31 56.422 8.521 45.86 54.34 48.84 51.76 51.63 51.32 52.87 47.66 50.66 50.21 51.24 48.99 50.448 2.319

TH analysis

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

1.425 1.060 1.130 1.233 1.085 1.278 1.050 0.997 1.078 1.065 1.523 1.115 1.170 0.163

EVALUATION OF PEAK STRUCTURAL RESPONSES

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

0.680 0.916 0.980 0.701 0.672 0.777 0.806 0.849 0.619 0.967 0.902 0.663 0.794 0.128

13.73 19.88 14.59 13.22 4.82 3.07 0.97 13.42 5.52 25.01 12.90 4.82 6.822 11.571

Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

171

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

0.613 0.920 0.882 0.773 0.625 0.775 0.774 0.792 0.635 0.921 0.841 0.591 0.762 0.120

1.05 21.68 11.05 16.20 13.97 12.32 13.99 11.13 19.11 26.32 17.51 8.10 8.853 13.524

172

S. SWENSEN AND K. WONG

demand of the system is measured, and then the linear response spectrum is simply reduced by dividing the spectrum by a reduction factor RSDOF (Krawinkler and Nassar, 1990), which is determined by Equation (1): RSDOF ( ) = Fmax,e Fy ( ) (1)

For SDOF systems with long natural periods of vibration, the reduction factor R approaches ductility demand, m; while for SDOF systems with short natural periods, R approaches 1.0 (Cuesta and Aschheim, 2001). For MDOF systems, however, the procedure of calculating the reduction factor is not trivial. Nassar and Krawinkler (1991) noted that for different building types, damping ratios, strength-hardening ratios and ductility demands the reduction factor for a MDOF system can differ greatly from that of a SDOF system with the same ductility demand. They proposed to rst model the MDOF system as a SDOF system. The ductility demands of the SDOF and MDOF systems must be the same and are determined by the relationship in Equation (2):

um max uy

(2)

The average ductility demands for Frames 1 to 3 were determined to be 1.62, 1.77 and 1.88, respectively, after conducting nonlinear time history analysis due to the 12 ground motions. These ductility demands seem reasonable to be utilized for analysis in the Salt Lake City area, since more active seismic zones frequently design steel moment frame structures for ductilities of over 2.0 (Warnitchai and Panyakapo, 1999). The relationship between RSDOF and RMDOF was studied by Nassar and Krawinkler (1991), and the concept is presented in Figure 7. For a SDOF system with a given target ductility mt, the reduction factor is equal to the elastic strength of the system divided by the yield strength of the same system. This relationship has been presented in Equation (1). For mt, this R value is represented as Point 1 in Figure 7. To convert this reduction factor to one that can be used to nd an inelastic response spectrum for a MDOF system, some modications must occur. Instead of dividing the elastic strength of the SDOF system by the yield strength of the system, the elastic strength of the SDOF system is divided by the yield strength of the MDOF system, where the SDOF and MDOF systems share the same ductility. Equation (3) shows the determination of RMDOF: RMDOF ( t ) = Fmax,e Vy ( t ) (3)

This adjusted reduction factor for MDOF systems is represented as Point 2 in Figure 7. In this way, the ratio between the reduction factor for a MDOF system and the reduction factor for a SDOF system with the same target ductility is equal to the yield strength of the SDOF system divided by the yield strength of the MDOF system. Equation (4) displays this relationship:
R SDOF

MDOF
1 2

1.0 1.0 t '

Figure 7. R m relationships for SDOF and MDOF systems (from Nassar & Krawinkler, 1991).
Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

EVALUATION OF PEAK STRUCTURAL RESPONSES

173

RMDOF ( t ) Fy ( t ) = RSDOF ( t ) Vy ( t )

(4)

Based on this method, the reduction factors for Frames 1 to 3 were determined to be 1.16, 1.17 and 1.21, respectively. Because differences in maximum base shear and maximum displacement exist between ground motion records, average values were used for calculating the strength reduction factors. Nassar and Krawinkler (1991) noted that the reduction factors for MDOF systems are frequently much lower than the ductility demands of the same systems. This is particularly true of models that show some soft-storey characteristics (Moghaddam and Mohammadi, 2001). Pushover analysis of Frames 1 to 3 indicate that some soft-storey conditions (i.e., early yielding at column bases) exist in the bottom two storeys. The obtained reduction factors are noted to be much less than the ductility demands of the same structures. It is interesting to note that the reduction factors for the three frames are fairly similar; for all three structures, a value of RMDOF 1.2 was determined. This similarity in ductility values could suggest that for the Salt Lake City area, a reduction factor of 1.2 can often be used in the nonlinear analysis of steel moment frames. Because the MDOF reduction factors are so similar, a value of 1.2 was used to reduce the elastic response spectra of each ground motion record. The resulting nonlinear response spectra based on the MDOF reduction factor of 1.2 are shown in Figure 8. 7. NONLINEAR ANALYSIS The nonlinear response spectra as shown in Figure 8 were then used to obtain peak responses using response spectrum analysis. Again, the CQC was used to combine modal acceleration and displacement values. The results are compared with those obtained using nonlinear time history analysis subjected to the 12 ground motions as shown in Table 5. It is noted that while many of the peak acceleration and displacement values computed by nonlinear response spectrum analysis are similar to the time history results, the agreement is less pronounced when compared to the linear analysis results. For the acceleration values, peak response spectrum values differ as much as almost 37% when compared to acceleration obtained using nonlinear time history analysis. For the peak displacement values, the maximum differences are somewhat smaller in magnitude. Peak displacement values obtained using nonlinear response spectrum analysis only differ from values obtained through nonlinear time history analysis by a maximum of about 27%. It is also observed that for a given model, the peak acceleration and displacement responses sometimes differ by a signicant amount when one ground motion record is compared to another. Since the peak values obtained through nonlinear response spectrum analysis result from reducing similar

1.6 1.4

Chalfant Chi Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows

Ground Acceleration (g)

1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2

2.5

Period (s)

Figure 8. Nonlinear response spectra.


Copyright 2009 John Wiley & Sons, Ltd. Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

Table 5. Nonlinear analysis data.


174
Six-storey frame linear analysis (Frame 1) Maximum displacement (cm) TH analysis 26.61 24.13 24.95 24.40 22.65 24.95 25.11 23.89 25.19 24.44 26.11 24.99 24.786 1.022 24.54 24.49 24.76 25.34 24.99 24.56 24.62 24.76 24.28 24.54 24.58 24.72 24.682 0.272 7.78 1.50 0.79 3.85 10.30 1.55 1.91 3.64 3.63 0.38 5.86 1.07 0.244 4.783 1.056 0.991 1.067 0.968 1.048 1.045 0.932 0.963 0.930 0.879 1.187 0.929 1.000 0.085 0.942 0.900 0.923 0.949 0.938 0.940 0.894 0.935 0.919 0.929 0.897 0.844 0.918 0.030 25.93 22.37 22.87 24.18 22.71 23.86 20.76 23.42 23.04 24.43 20.41 22.23 23.016 1.530 10.76 9.12 13.48 1.91 10.47 10.12 4.11 2.87 1.16 5.71 24.46 9.14 21.69 20.13 21.32 22.98 21.66 22.15 19.48 21.83 21.22 21.96 19.47 19.58 21.123 1.174 RS analysis % difference TH analysis RS analysis % difference TH analysis RS analysis Maximum acceleration (g) RS analysis 1.056 1.053 1.059 1.062 1.064 1.055 1.057 1.059 1.046 1.055 1.054 1.042 1.055 0.006 10-storey frame linear analysis (Frame 2) 0.712 0.711 0.721 0.716 0.730 0.707 0.717 0.713 0.709 0.717 0.705 0.715 0.715 0.007 10-storey frame linear analysis (Frame 3) 0.658 0.664 0.649 0.649 0.665 0.637 0.657 0.655 0.653 0.654 0.649 0.653 0.654 0.007 7.40 27.85 26.45 15.95 6.29 17.82 15.10 17.29 2.69 28.97 22.79 10.47 12.114 14.719 54.44 67.17 55.84 63.56 58.79 58.38 62.77 55.27 63.01 58.96 55.49 58.31 59.332 3.982 56.98 59.46 57.90 58.24 57.99 57.40 58.87 57.87 58.29 57.28 57.41 59.21 58.076 0.781 4.67 11.48 3.69 8.36 1.36 1.68 6.22 4.70 7.49 2.84 3.46 1.55 1.780 5.586 0.465 0.493 0.425 0.520 0.499 0.487 0.429 0.554 0.437 0.437 0.553 0.485 0.482 0.045 4.73 22.43 26.37 2.20 8.70 9.00 11.09 16.04 14.51 25.79 21.84 7.93 7.874 14.876 46.89 57.13 52.37 53.72 50.23 46.55 51.34 50.17 51.64 53.80 53.20 53.77 51.735 3.004 51.84 51.92 51.31 52.41 51.81 51.67 52.76 51.82 51.16 51.71 51.87 52.13 51.868 0.431 10.57 9.12 2.01 2.44 3.15 11.00 2.76 3.29 0.94 3.88 2.49 3.05 0.568 5.893 0.509 0.593 0.555 0.572 0.584 0.597 0.550 0.647 0.474 0.508 0.639 0.563 0.566 0.052 0.634 0.619 0.602 0.638 0.629 0.638 0.625 0.630 0.591 0.636 0.609 0.632 0.624 0.015 25.90 0.66 6.34 13.85 1.89 17.47 0.65 6.30 3.02 0.90 30.81 6.47 8.363 11.338 % difference Six-storey frame nonlinear analysis (Frame 1) Maximum displacement (cm) % difference 16.35 10.03 6.77 4.95 4.62 7.15 6.17 6.78 7.89 10.08 4.57 11.90

Maximum acceleration (g)

Copyright 2009 John Wiley & Sons, Ltd.


10-storey frame nonlinear analysis (Frame 2) 24.54 4.38 8.59 11.70 7.63 6.83 13.64 2.62 24.73 25.24 4.76 12.26 11.013 9.981 40.11 57.11 51.58 53.70 48.92 48.01 43.10 47.59 43.38 61.59 52.56 44.56 49.350 6.269 45.62 45.75 44.05 46.61 46.57 46.54 43.52 41.20 45.77 46.18 45.78 46.70 45.357 1.646 10-storey frame nonlinear analysis (Frame 3) 0.558 0.589 0.566 0.544 0.578 0.562 0.587 0.574 0.559 0.577 0.574 0.572 0.570 0.013 19.88 19.35 33.01 4.66 15.88 15.26 36.92 3.74 27.84 31.89 3.74 17.81 19.166 11.515 46.34 69.38 54.90 61.77 60.02 58.53 46.38 53.63 42.53 68.14 62.12 53.31 56.422 8.521 45.86 54.34 48.84 51.76 51.63 51.32 52.87 47.66 50.66 50.21 51.24 48.99 50.448 2.319

TH analysis

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

1.425 1.060 1.130 1.233 1.085 1.278 1.050 0.997 1.078 1.065 1.523 1.115 1.170 0.163

S. SWENSEN AND K. WONG

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

0.680 0.916 0.980 0.701 0.672 0.777 0.806 0.849 0.619 0.967 0.902 0.663 0.794 0.128

13.73 19.88 14.59 13.22 4.82 3.07 0.97 13.42 5.52 25.01 12.90 4.82 6.822 11.571

Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

Chalfant Chi-Chi Coalinga Coalinga 2 Coyote Lake Imperial Valley Landers Loma Prieta Mammoth Superstition Hills Westmorland Whittier Narrows Average Std. deviation

0.613 0.920 0.882 0.773 0.625 0.775 0.774 0.792 0.635 0.921 0.841 0.591 0.762 0.120

1.05 21.68 11.05 16.20 13.97 12.32 13.99 11.13 19.11 26.32 17.51 8.10 8.853 13.524

EVALUATION OF PEAK STRUCTURAL RESPONSES

175

response spectra by a constant factor, the reduced inelastic spectra are also similar, as shown in Figure 8. This similarity between reduced inelastic spectra is noted in that, unlike the time history results, the peak acceleration and displacement values for a certain model are fairly similar when one ground motion is compared against another, as shown in the standard deviation in Table 5. In other words, inelastic response spectra peak responses will be similar between ground motion records with similar elastic response spectra. This close correlation is not observed when nonlinear time history analysis is performed. It is also noted that both peak accelerations and peak displacements found using the two methods vary to a greater degree in Frame 3 than in Frames 1 and 2. This likely suggests that nonlinear response spectrum analysis results are less reliable when higher ductility demands are utilized. While peak displacement and acceleration values obtained through inelastic response spectrum analysis show less congruence to the time history values than in the elastic cases, this type of analysis may still hold value because it requires much less computing time and storage then nonlinear time history analysis to complete. 8. CONCLUSION It is concluded that for structures that behave elastically when subjected to the considered ground motions, response spectrum analysis provides a fairly accurate approximation of peak acceleration and displacement values. The displacement values obtained through elastic response spectrum analysis are particularly similar to those obtained through linear time history analysis. The accelerations found using response spectrum analysis are somewhat less similar to those found during time history analysis, but still provide a fairly accurate estimate. It is also noted that for linear systems, peak accelerations obtained through response spectrum analysis are usually slightly lower than those found using modal time history analysis. Because of the relative simplicity of response spectrum analysis in comparison to time history analysis and the relative similarity of the determined peak responses, linear response spectrum analysis is a viable alternative to time history analysis. For the site of the University of Utah in Salt Lake City, the nonlinear strength reduction factor was found to be close to 1.2 for all analysed structures. This could suggest that RMDOF values are similar when ground motion parameters are comparable. For structures that are expected to behave nonlinearly during an earthquake event, nonlinear response spectra analysis was utilized according to methods set forth by Nassar and Krawinkler (1991). For the site of the University of Utah in Salt Lake City, the nonlinear strength reduction factor was found to be close to 1.2 for all analysed structures. This could suggest that RMDOF values are similar when ground motion parameters are comparable. It was found that nonlinear response spectrum analysis provides a less accurate estimation of peak accelerations and displacements when compared to time history results. These values, however, may be sufcient for some design purposes. Inelastic response spectrum analysis is also somewhat less viable, because obtaining the strength reduction factor R for purposes of reducing the elastic response spectrum requires the construction of an additional SDOF model. Also, ductility factors and peak elastic base shear values must be determined. It is also observed that greater variation between peak responses found using nonlinear time history analysis and nonlinear response spectrum analysis exists when the ductility values are particularly high. Inelastic response spectra analysis may still be useful, however, because the analysis requires less computing time and storage when compared to nonlinear time history analysis.

REFERENCES
Abrahamson NA. 1992. Non-stationary spectral matching. Seismological Research Letters 63(1): 30. ASCE. 2006. Minimum design loads for buildings and other structures. ASCE 7-05, ASCE, Reston, VA. Chopra AK, Goel R. 2003. A modal pushover analysis procedure to estimate seismic demands for unsymmetric-plan buildings: theory and preliminary evaluation. Report No. EERC 2003-08. Earthquake Engineering Research Center, University of California, Berkeley, CA. Chopra AK. 2007. Elastic response spectrum: A historical note. Earthquake Engineering and Structural Dynamics 36(1): 312.

Copyright 2009 John Wiley & Sons, Ltd.

Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

176

S. SWENSEN AND K. WONG

Chopra AK. 2008. Dynamics of Structures, Theory and Applications to Earthquake Engineering, 3rd ed. Prentice-Hall, Upper Saddle River, NJ, 257259. Cuesta I, Aschheim M. 2001. Inelastic response spectra using conventional and pulse R-factors. Journal of Structural Engineering 127(9): 10131020. Krawinkler H, Nassar A. 1990. Strength and ductility demands for SDOF and MDOF systems subjected to Whittier Narrows earthquake ground motions. In Proceedings of the California Geological SurveySMIP90 Seminar, Sacramento, CA. Lilhanand K, Tseng W. 1987. Generation of synthetic time histories compatible with multiple-damping response spectra. SmiRT-9, Lausanne, K2/10. Lilhanand K, Tseng W. 1988. Development and application of realistic earthquake time histories compatible with multiple damping response spectra. In Proceedings of the Ninth World Conference of Earthquake Engineering, Tokyo, Japan, Vol II, 819824. Moghaddam R, Mohammadi R. 2001. Ductility reduction factor of MDOF shear-building structures. Journal of Earthquake Engineering 5(3): 425440. Nassar A, Krawinkler H. 1991. Seismic demands for SDOF and MDOF systems. Report No. 95, John A. Blume Earthquake Engineering Center, Stanford University, Stanford, CA. Warnitchai P, Panyakapo P. 1999. Constant-damage design spectra. Journal of Earthquake Engineering 3(3): 329347.

Copyright 2009 John Wiley & Sons, Ltd.

Struct. Design Tall Spec. Build. 20, 164176 (2011) DOI: 10.1002/tal

You might also like