19 - An Evolutionary Divide-and-Conquer Strategy for Structural Identification

Abstract

Metaheuristics have been applied in dealing with civil engineering tasks such as structural design optimization and time-cost optimization of construction projects. Among others, the concept of genetic algorithm (GA) has shown great potential as a robust and versatile metaheuristic method. The GA method is often used to generate useful solutions to optimization and search problems in many disciplines, using concepts inspired by natural evolution such as selection, mutation, and crossover. Nevertheless, its applicability in identification of large structures is still limited when involving a large number of unknowns. To address this problem, we present an evolutionary divide-and-conquer strategy. The key idea behind this strategy is to divide a large structure into many substructures, each with much lesser unknowns, so that each substructure can be identified more effectively using a recently developed multifeature GA method. The multiple features involve reducing search space throughout a search process based on statistic parameters and employing various mutation operators, allowing broad and narrow searching to occur simultaneously. This strategy also accounts for the interaction effects between substructures using only acceleration measurements. This strategy is also useful to detect local damage in large structures. The performance of this strategy is demonstrated through numerical and experimental studies.