Dual-population constrained multi-objective evolutionary optimization algorithm with repair constraint handling technique for structural optimization

Structural optimization often involves highly complex problems with numerous decision variables and non-convex feasible regions, making it challenging for most optimization algorithms to converge to true Pareto front. Even when convergence is achieved, existing methods typically require thousands of function evaluations, resulting in significant computational cost. This highlights the need for efficient and robust optimization algorithms suitable for real-world engineering applications. In this thesis, we introduce a novel constrained multi-objective evolutionary algorithm, DPCME. The algorithm utilizes two interacting populations that exchange information, enabling efficient exploration of the global search space and helping to avoid convergence to local optima. To further enhance performance, a repair-based constraint handling technique is incorporated, with multiple strategies designed for different scenarios. The proposed algorithm and repair constraint handling technique are systematically tested on three complex engineering benchmark problems: the 72-bar truss, the 120-bar truss, and a chemical tanker structure. Performance is assessed in comparison with state-of-the-art constrained multi-objective optimization algorithms, and the effectiveness of different repair strategies is analyzed. Results demonstrate that DPCME achieves superior convergence and diversity across all test cases, and that the inclusion of the repair constraint handling technique further improves its performance.