Computationally efficient analysis of cable-stayed bridge for GA-based optimization

Abstract

Optimum design of a cable-stayed bridge structure is very complicated because of large number of design variables. Use of genetic algorithms (GAs) in optimizing such structure consumes significant computational time. Due to nonlinearity, structural analysis itself takes considerable computational time and the genetic algorithm has to perform a large number of iterations in order to obtain global minima. A new approach combining GA and support vector machine (SVM) has been adopted. This drastically reduces the computation time of optimization. The genetic algorithm is employed to obtain the minimum cost of the cable-stayed bridge. Constraint evaluation is done using SVM which is trained by a data base generated through FEM analysis. System level optimization is carried out considering configuration and cross-sectional parameters as design variables. In the present study, optimization was carried out for bridge lengths ranging from 100 to 500 m. Final optimum designs were reanalyzed to check the adequacy of the developed approach.