Curvature-aware window identification and variable feed-rate modulation for ultra-precision turning of microlens...

Microlens arrays (MLAs) usually exhibit pronounced local geometric variations near lens boundaries and transition regions, making it difficult to simultaneously balance form accuracy, trajectory smoothness, and machining efficiency in ultra-precision single-point diamond turning (SPDT). Existing methods for controlling boundary-related errors in MLAs often rely on local interpolation or local path densification. Although such strategies can improve local form quality to some extent, they may deteriorate the global smoothness and continuity of the tool trajectory and introduce larger kinematic disturbances in transition regions, thereby increasing the risk of machine-tool vibration and compromising machining stability. To address this issue, this study proposes a curvature-aware window identification and function-based variable feed-rate modulation method for MLA machining, combined with Non-dominated Sorting Genetic Algorithm II (NSGA-II) multi-objective optimization. Different from conventional local interpolation, purely curvature-based resampling, or empirically defined feed-rate adjustment strategies, the proposed method establishes a geometry-driven workflow that links curvature-aware critical-region identification, automatic radial-window extraction, smooth feed-rate modulation, and global accuracy-efficiency optimization. Priority windows requiring denser sampling are first identified automatically from the surface geometry. A smooth feed-rate function is then constructed along the spiral path so that the tool decelerates in critical regions and accelerates appropriately in non-critical regions, thereby coordinating local accuracy control with global trajectory smoothness. Furthermore, the feed-function parameters are globally optimized by simultaneously minimizing the residual-height metric and the total number of cutting points. Experiments on MLA show that the proposed method improves the surface quality of the MLA region while reducing the total number of cutting points from 873,000 to 741,833.