A customized bilevel optimization approach for solving large-scale truss design problems

Considerable academic research has been conducted on truss design optimization by standard metaheuristic methods; however, the generic nature of these methods becomes inefficient for problems with many decision variables. This may explain the simplicity of the relevant test problems in the academic literature in comparison with real structures. To address this challenge, this study advocates a customized optimization methodology which utilizes problem-specific knowledge. It improves upon a new bilevel truss optimization method to allow for an arbitrary trade-off between the stochastic upper level and the deterministic lower level search. Numerical simulations demonstrate that for large-scale truss design problems, the proposed method can find significantly lighter structures up to 300 times more quickly than the best existing metaheuristic methods. The remarkable findings of this study demonstrate the importance of using engineering knowledge and discourage future research on the development of purely metaheuristic methods for truss optimization.