A data-driven multi-stage framework for optimization of machining parameters in TC4 turning against cemented carbide tool wear evolution

The high-efficiency machining of titanium alloy (TC4) is severely constrained by intense tool wear. Existing studies have failed to consider the complexity of the dynamic evolution of tool wear and the nonlinear effect on machining performance, which results in poor adaptability in practical production. To address this, is proposed a novel data-driven multi-stage optimization framework to achieve stage-specific optimization of process parameters throughout the entire tool life cycle. Within this framework, the wear evolution stages of TC4 turning tools are more comprehensively delineated, and a mapping relationship among tool wear status, process parameters, and machining performance is established. A hybrid BOBP-NSTOP algorithm, founded on bayesian-optimized neural network (BO-BPNN), NSGA-II, and TOPSIS, is developed to achieve accurate and rapid prediction of dynamic optimal process parameters across the entire tool life cycle. Furthermore, compared to the traditional optimal static process parameters, the dynamic parameters obtained through this framework reduce cutting force and energy consumption while maintaining a high material removal rate. This study provides substantial fundamental data support for the intelligent manufacturing of TC4, confirms the necessity of considering the dynamic characteristics of tool wear, and offers an effective systematic solution for realizing intelligent adaptive machining and process optimization. • A framework for dynamically optimizing parameters to compensate for tool wear. • The wear evolution of TC4 tools extends beyond the three-stage model. • Achieves 95.3% higher computational efficiency while ensuring prediction accuracy. • A comprehensive dataset covering the entire tool wear cycle in TC4 turning. • The necessity of considering the dynamic characteristics of tool wear is demonstrated.