Identification of input ground motion records for seismic design using neuro-fuzzy pattern recognition and genetic algorithms


A data classification system is designed by pattern recognition to preprocess data that can be utilized in a genetic algorithm (GA) that performs search and scaling task of finding strong ground motion (SGM) records for seismic design. Tectonic settings, regional consideration of seismology and site characteristics are taken into account as well as nonlinear structural response and performance-based design requirements for selection of input motion records. The objective of this study is twofold. First, a better understanding of SGM characteristics is attained by applying statistical pattern recognition techniques. Second, the classification of records makes it possible for the GA-based search and scaling methodology, developed in an earlier study by the authors, to present more appropriate input motion records in the design bin for the site under consideration. Better seismic hazard representation would result in reduced uncertainty in demand estimation in the probabilistic performance-based context, which in turn will enhance structural performance and cost efficiency of design.

Publication Title

Proceedings of the 2004 Structures Congress - Building on the Past: Securing the Future

This document is currently not available here.