Selection and Scaling of Ground Motion Time Histories for Structural Design Using Genetic Algorithms

Abstract

This paper presents a new approach to selection of a set of recorded earthquake ground motions that in combination match a given site-specific design spectrum with minimum alteration. The scaling factors applied to selected ground motions are scalar values within the range specified by the user. As a result, the phase and shape of the response spectra of earthquake ground motions are not tampered with. Contrary to the prevailing scaling methods where a preset number of earthquake records (usually between a single component to seven pairs) are selected first and scaled to match the design spectrum next, the proposed method is capable of searching a set consisting of thousands of earthquake records and recommending a desired subset of records that match the target design spectrum. This task is achieved by using a genetic algorithm (GA), which treats the union of 7 records and corresponding scaling factors as a single “individual.” The first generation of individuals may include a population of, for example, 200 records. Then, through processes that mimic mating, natural selection, and mutation, new generations of individuals are produced and the process continues until an optimum individual (seven pairs and scaling factors) is obtained. The procedure is fast and reliable and results in records that match the target spectrum with minimal tampering and the least mean square of deviation from the target spectrum.