| This thesis presents comparisons of different scaling and spectral matching methods for ground motions to perform nonlinear dynamic analysis. This is done to determine (a) which variation of magnitudes could be obtained from one method to the other, (b) identify the advantages and disadvantages of the different methods and (c) recommend which methods should be used to scale or modify ground motions. Eight of these methods are applied to a group of 30 historical ground motions. They are then used for nonlinear dynamic analyses on single degree of freedom structures, a 10-storey reinforced concrete shear wall and a 4-storey steel braced frame structure located in Montreal. A secondary objective is to determine if groups of 1, 3 or 7 ground motions should be used for nonlinear dynamic analyses.;The scaling methods recommended after these analyses are the SIa and MSE scaling methods. The SIa method consists in measuring the area under the response spectra which means between two preselected periods as well as the area under the UHS between these same two periods. The scaling factor is computed to have these areas equal to each other. The MSE method aims at minimising the mean square error between the UHS and the response spectra. These methods show more realistic structural responses with all the structures as well as a low standard deviation. The frequency domain (FD) and time domain (TD) spectral matching methods return standard deviations significantly higher than the ones obtained with the MSE and SIa methods thus they are not recommended for practical use in nonlinear analysis.;To select the appropriate number of accelerograms required for adequate dynamic analysis, results from groups of 1, 3 and 7 ground motions have been compared. It is concluded that, in general, a group of 7 ground motions allows lower seismic forces because this group is allowed to use the average structural response. In some situations, a group of 3 ground motions can return not conservative structural response as well as, in other cases, too conservative structural response. Groups of 1 accelerogram are not recommended because, even with spectral matching methods (FD and TD), the standard deviation is too high to obtain representative results. |
