DPCME algorithm beats 43 rivals in complex structural optimization
New dual-population method reduces risk of local optima with repair-based handling.
Structural optimization problems often involve hundreds of nonlinear failure constraints and highly non-convex feasible regions, making it difficult to reach the true Pareto front. Traditional algorithms may require thousands of function evaluations, driving up computational cost. To address this, researchers Fardad Homafar and Jasmin Jelovica have developed DPCME (Dual-Population Constrained Multi-Objective Evolutionary algorithm). The key innovation: two interacting populations that exchange information, enabling global exploration and reducing the risk of local optima. A repair-based constraint-handling technique is integrated, with alternative repair approaches evaluated for best performance.
DPCME was rigorously tested on three engineering benchmarks: a 72-bar truss, a 120-bar truss, and a chemical tanker structure. It was compared against 43 algorithms from the latest PlatEMO package, with the 12 best-performing ones selected for detailed analysis. Results show DPCME achieves superior or competitive convergence and diversity across all test cases. The inclusion of repair constraint handling further improves its ability to handle complex constraints, making it a promising tool for real-world engineering design where both performance and efficiency are critical.
- DPCME uses two interacting populations to improve global exploration and avoid local optima
- Tested on 72-bar truss, 120-bar truss, and chemical tanker with hundreds of nonlinear failure constraints
- Outperformed 43 state-of-the-art algorithms from PlatEMO, with repair handling boosting performance
Why It Matters
DPCME offers faster, more reliable structural optimization for engineers, reducing computational cost in complex designs.