Tool condition monitoring for cavity milling based on bispectrum analysis and Bayesian optimized SVM

Tool wear status seriously affects the dimensional accuracy and surface quality of machined parts. Therefore, tool condition monitoring (TCM) is essential in the milling process of aerospace structural parts due to the utilization of difficult-to-cut materials and complex cutting trajectories. The higher order spectrum (HOS) was first employed to analyze the vibration signals. Then bispectral features extracted from denoised signals were used to characterize the tool wear status. The improved Pearson’s correlation coefficient was used for feature selection in order to reduce the influence of periodic components on the feature selection process. Furthermore, a novel objective function was proposed for the Bayesian optimization algorithm to guide the optimization process of hyperparameters for the support vector machine. This objective function takes into consideration the effect of imbalanced data on the recognition rates. To demonstrate the effectiveness of the proposed method, a milling experiment was conducted to machine a structural part, and vibration signals were collected during the process. Finally, an online TCM model was established based on this. The present study suggests that the proposed TCM system is accurate and robust.