Sensitivity of building response to variation in integration time step of response history analysis.

摘要

Non-linear time history analysis is a methodology used in earthquake engineering to estimate structural response to a ground motion time history. The methodology utilizes a ground motion acceleration time-history as an input to an analytic model of a structure and solves the structural response at each time step of the ground motion record. This dissertation summarizes the research findings on identifying the appropriate time step for a ground motion acceleration time-history (given any ground motion and structural system) in order to maximize computational efficiency and ensure accurate estimation of the system's response. With increases in number of degrees of freedom of structural systems, an issue commonly encountered in tall structures, the nonlinear analysis using ground motions with small sampling time (high sampling rate) becomes quite time consuming. This study proposes a viable method to reduce the analysis time by downsampling ground motion data. Downsampling is achieved by increasing the ground motion's time step which in turn reduces the overall number of data points. In this method a structure's Frequency Response Function is used to judge the appropriate time step according to the characteristics of a given ground motion. Filtering and downsampling techniques from the field of signal processing are applied to an original ground motion to generate a downsampled ground motion with the goal of obtaining a highly efficient time history analysis without significant error. This dissertation shows a practical methodology and delivers the associated tools for proper down-sampling of ground motion's acceleration time-history to be used in response history analysis. The methodology is not only practical, but also capable of controlling errors in estimates of response. Results of the methodology are judged in terms of a goodness of fit test, comparing observed responses to the downsampled ground motions with similar estimates using original ground motions for typical SDOF and MDOF structural systems. This comparison is measured in terms of closeness of predicting the characteristics of structural response of time history analysis with a large subset of Next Generation Attenuation ground motion database. The response comparison is made with (i) the comparison of maximum displacements in structural response obtained from original and downsampled ground motions; (ii) the cross-correlation coefficient of structural displacement responses. The results show that the proposed method of downsampling earthquake ground motion records is effective and convenient for time history analysis.