Mixed Variable Structural Optimization: Toward an Efficient Hybrid Algorithm

Abstract

Designing a structure implies a selection of optimal concept and sizing, with the aim of minimizing the weight and/or production cost. In general, a structural optimization problem involves both continuous variables (e.g., geometrical variables, ...) and categorical ones (e.g., materials, stiffener types, ...). Such a problem belongs to the class of mixed-integer nonlinear programming (MINLP) problems. In this paper, we specifically consider a subclass of structural optimization problems where the categorical variables set is non-ordered. To facilitate categorical variables handling, design catalogs are introduced as a generalization of the stacking guide used for composite optimization. From these catalogs, a decomposition of the MINLP problem is proposed, and solved through a branch and bound method. This methodology is tested on a 10 bars truss optimization inspired from an aircraft design problem, consistent with the level of complexity faced in the industry.