Extension of concurrent subspace optimization to structural optimization of product families

Abstract

This paper discusses the problem of structural optimization of product families with predefined platforms. The main challenge lies in the increased design variables and constraints, and providing an optimal tradeoff for individual products performance in the family which are competitive with each other. The Concurrent Subspace Optimization for multidisciplinary problem is extended to product family design with predefined platforms. The main advantage of the proposed approach is that the system level owns the ability to catch the global tendency of the true design space and the number of evaluations required is reduced by using surrogate models. Each subspace optimization problem has the freedom to specify the unique variables for one family member, and the system level optimizes the product platform using the surrogate models created based on subspace optimizations. The process is solved in an iterative way, and the improving surrogate models guide the optimization to the global optimal design. Results from a truss family example with small design space confirm the ability and efficiency of the extended Concurrent Subspace Optimization to address product family problem by compared with ATC approach. Then the proposed method is successfully applied to a family of unmanned aircraft wing structures, which is more complicated and related to practical implementation issues.