A STEP-NC-based post-process simulation system for closed-loop machining

Traditional pre-process simulations focus on validation tasks such as collision detection and G-code verification; however, they cannot perceive the actual machining status. In response to these limitations, in-process simulation technologies have emerged to facilitate real-time state monitoring. Nevertheless, they remain unable to exploit massive execution data for machining reconstruction or subsequent optimization. This paper presents a STEP-NC-based post-process simulation system designed to achieve workingstep-level machining reconstruction and machining parameter self-evolution. To achieve this, a STEP-NC-based machining reconstruction mechanism is developed, which incorporates workingstep-level data mapping by matching time-stamped execution signals with workingsteps, and features a semantic model that integrates geometric, machining parameters, and execution feedback into a unified representation. On this basis, a workingstep-level Bayesian optimization framework is introduced. Its iterative sampling and surrogate modeling mechanism inherently forms a machining-analysis-recommendation-machining loop to enable machining parameter self-evolution. Experimental results demonstrate that the system reliably retraces the digital history of physical execution with semantic consistency. Moreover, the optimization framework achieves consistent performance improvements across all 12 workingsteps within 30 iterations, with all BO-optimized results exceeding the best initial samples by a mean of 24.33% and up to 68.88% beyond the initial performance range, with convergence observed within 5 directed iterations following initialization in most cases. These findings confirm that the proposed framework effectively bridges the semantic gap between planning and execution, establishing a viable paradigm for closed-loop machining.