The poor machinability of Ti–6Al–4V poses challenges to notch wear prediction during ball-end milling, including complex mechanism modeling and heavy reliance on sensors. Existing data-driven approaches often suffer from limited interpretability and high monitoring costs. To address these issues, this study proposes a hybrid mechanism and data-driven method for predicting notch wear. A geometric model of the ball-end cutting edge was constructed, and the cutting edge micro-element force feature (CFF) and the effective cutting edge length feature (ECEF) were derived using polar coordinate projection and interpolation algorithms. These two features were identified as core drivers of notch wear, with clear mechanistic links. The derived features were then used to train predictive models based on ensemble learning, kernel methods, and artificial neural networks. All predictive models achieved a coefficient of determination () consistently exceeding 0.97, demonstrating robust generalization across multiple modeling paradigms, the Random Forest (RF) model stood out with optimal performance ( = 0.998). Further integrating the random forest algorithm with recursive feature elimination, feature optimization achieved a 50.2% improvement in computational efficiency while retaining 99.84% of the original model’s predictive capability. Finally, based on wear mechanism decoupling and ablation experiments, a cross-scale framework combining geometry, mechanics, and data was established. The results indicated that force-induced thermal cyclic load plays a dominant role in the notch wear process, while abrasive wear acts as an auxiliary factor, and the contribution of the former is 3.6 times that of the latter. This framework offers a new paradigm for wear prediction that is both mechanistically interpretable and practically applicable, with significant potential for high-end manufacturing sectors such as aerospace.
扫码关注我们
求助内容:
应助结果提醒方式:
