A Bilevel Methodology for solving a Structural Optimization Problem with both Continuous and Categorical Variables

Abstract

In industrial structural optimization problems, two kinds of variables are involved : continuous variables, like geometrical parameters, and categorical variables like choices of material and cross-sectional shapes. A key difficulty of this kind of problem is the sensitivity of the computation cost with respect to the number of categorical design variables. In this article, a new methodology is proposed to address this mixed continuous-categorical structural optimization problem. For this purpose, a 10-bar truss test case representative for industrial design is introduced. Promising results have been obtained in terms of computation cost, compared to approaches based on combinatorial algorithms.