Tool wear state recognition based on WOA-SVM with statistical feature fusion of multi-signal singularity

Tool wear state is the key factor affecting machining quality, machining efficiency, and cutting stability in the cutting process. Serious wear condition will even lead to machining process interruption and machine tool failure. Accurate monitoring of tool wear state has become increasingly important in the intelligent development of manufacturing industry. To monitor tool wear accurately and effectively, a new method based on whale optimization algorithm optimized support vector machine (WOA-SVM) with statistical feature fusion of multi-signal singularity was proposed to recognize the tool wear state. Based on estimating the maximum wavelet transformation module (MWTM), multi-signal denoising and singularity quantitative characterization were carried out. Meanwhile, the probability density transform was performed on the holder (HE) index, and the relevant statistical features were extracted. Random forest algorithm and KPCA algorithm were used for relatively important features screening and dimension reduction fusion of multi-signal singularity features. By establishing the correlation mapping between the fusion features and the tool wear level, a WOA-SVM classification model based on the fusion features was constructed to recognize the tool wear state. The performance of the method proposed was verified based on the milling wear experiment. Results showed that this method can identify the tool wear state efficiently and accurately based on the limited experimental data. Compared with some other classification methods, this method had better classification performance, effectiveness, and feasibility. These findings can be of great significance for evaluating tool condition, replacing tool timely and ensuring machining quality and efficiency.