Experiments and Numerical Optimization of Water-Jet Guided Laser Diamond Machining Based on the Improved NSGA-III Algorithm.

This article first investigates the single-factor effects in water-jet guided laser (WJGL) machining of diamond via experiments, analyzing how processing performance responds to laser energy and machining control parameters to define their optimization ranges. Subsequently, an Optimal Latin Hypercube Sampling (OLHD) is adopted to collect experimental data points, enabling exploration of the interaction mechanisms between process parameters and their compatibility with machining performance indicators. A surrogate model based on Gaussian Process Regression (GPR) with combined kernel functions is constructed to capture the complex nonlinear mapping between process parameters and response metrics. To address inherent uncertainties in the optimization model, an improved NSGA-III algorithm integrating the Expected Improvement dominance partition strategy (EIS) is proposed, using Expected Improvement (EI) to determine dominance relationships under WJGL processing uncertainties and derive matched process parameters. Validation via test functions and machining experiments demonstrate that the proposed method outperforms traditional NSGA-III (T-NSGA-III) with significantly lower prediction deviations. The optimized parameters achieved remarkable performance improvements: cutting depth (Nd) increased by 48.21%, kerf width (Kw) reduced by 1.44%, line roughness average (Ra) decreased by 43.09%, and cutting speed (Cs) improved by 78.40%. This research provides a viable process optimization approach for WJGL technology, enabling high-quality, efficient, and robust diamond machining.