Multi-objective fixture layout optimization for thin-walled parts via FEA and ML-augmented evolutionary algorithm

To improve both efficiency and robustness of fixture design, this paper presents an intelligent framework for fixture layout optimization. First, a modular fixture with reconfigurable components was designed for thin-walled parts, enabling rapid changes in clamping concepts. Next, a Sobol-based greedy sampling (SGS) algorithm was developed for efficient design of experiments (DoE). Based on the layout samples, Python-automated static and dynamic simulations provide training data for ML models. These models then served as surrogates for the objective functions and were integrated with NSGAII to achieve Pareto-optimal solutions. Finally, milling trials with real-time monitoring and surface/tolerance measurements validated the proposed framework.