Convergence Improvement of Reliability-Based MultiobjectiveOptimization Using Hybrid MOPSO

1. Abstract This study proposes the improvement method of computational efficiency for the hybrid-type multiobjective particle swarm optimization method (MOPSO) that the authors developed for the reliability-based multiobjective optimization (RBMO). The hybrid MOPSO integrates the constraint satisfaction technique using gradient information of constraints with a concept of the single-loop-single-vector method (SLSV). The constraint satisfaction technique consists of two functions: moving the design candidate with constraint violation to the feasible region based on the sensitivity of the violated constraints and to the feasible boundary using bi-section method. Some design candidates, however, take much computation with zigzag iterations until the candidate moves to feasible region. This study proposes the improvement method by eliminating the zigzag iterations based on an idea from the modified SLSV method or the conjugate mean value method proposed for resolving the convergence problem for the reliability-based optimization. The efficiency of the proposed method is demonstrated through several design problems by investigating the convergence and diversity of finding Pareto optimal solutions. 2.