Explainable Machine Learning with Two-Layer Multi-Objective Optimization Algorithm Applied to Sealing Structure Design

This study addresses the challenge of optimizing seal structure design through a novel two-stage interpretable optimization framework. Focusing on O-ring waterproof performance under hyperelastic material behavior, this study proposes a double-layer optimization method integrating explainable machine learning with hierarchical clustering algorithms. The key innovation lies in employing modified hierarchical clustering to categorize design parameters into two interpretable groups: bolt preload and groove depth. This clustering enables dimensionality reduction while maintaining the physical interpretability of critical parameters. In the first layer, systematic parameter screening and optimization are applied to the preload variable to reduce the database, with six remaining data points that constitute one-seventh of the original data. The second layer subsequently refines configurations using E-TOPSIS (Entropy Weight—Technique for Order Preference by Similarity to Ideal Solution) optimization. All evaluations are performed through FEA (finite element analysis) considering nonlinear material responses. The optimal design is a groove depth of 0.8 mm and a preload of 80 N. The experimental validation demonstrates that this method efficiently identifies optimal designs meeting IPX8 waterproof requirements, with zero leakage observed in both O-ring surfaces and motor interiors. The proposed methodology provides physically meaningful design guidelines.