Author(s)

A. Sepulveda & L. Epstein

Abstract

Application of Bayesian methods in global structural optimization A. Sepulveda**, L. Epstein^ ^Department of Mechanical, Aerospace and Nuclear Engineering, University of California, Los /Inge/as CL4 ^00^, CASL4 ^Department of Biostatistics, The Johns Hopkins University, Baltimore MD 21205, USA ABSTRACT This paper proposes a multistart method to find the global minimum of con- straint problems, and suggests a new scheme, which we call the repulsion algorithm, for selecting initial design points for local searches. Bayesian analysis provides the stopping rules. The method is applied to structural optimization design problems by generating and globally solving a sequence of approximate problems, which is the classical approach in the approximation concepts method. Numerical example problems are presented. INTRODUCTION In structural and multidisciplinary optimization, it is common to find design spaces that are either disjoint or nonconvex. Examples are grillage structures

Keywords