DG-MA: A Multi-Objective Optimization Algorithm Based on Improved NSGA-III for Ungrouped Selective Assembly

Traditional selection assembly techniques use a grouping-then-matching strategy, which suffers from inherent limitations, including a high number of wasted parts and poor matching effectiveness when working with parts whose dimensions are not dis-tributed normally. Although ungrouped assembly represents the current mainstream research direction, the vast combinatorial solution space it generates poses significant challenges to the global search and decision-making capabilities of optimization algo-rithms. To address this issue, this study abandons the grouping method and formulates the selective assembly problem as a large-scale combinatorial optimization model based on a global matching strategy, which fundamentally mitigates part wastage. This study proposes a unique hybrid intelligence algorithm, DG-MA, based on the NSGA-III framework to effectively solve the multi-objective optimization of this mod-el. This algorithm is designed to simultaneously optimize both assembly quantity and quality, employing the TOPSIS method to scientifically select the optimal balanced solution from the Pareto front. Comparative experiments demonstrate that, compared to the traditional grouping method, our proposed approach reduces wasted parts by 76.5%, increases the assembly success rate by 8.7 percentage points, and significantly outperforms mainstream optimization algorithms across multiple key performance indicators. Combined with the ungrouped strategy, the proposed DG-MA algorithm provides a thoroughly validated and highly effective solution for large-scale selective assembly problems, significantly enhancing both the quantity and quality of assemblies.