Tool wear prediction based on PCA-SVM models

With the advancement of the manufacturing industry toward greater intelligence and automation, real-time monitoring of tool wear is critically important for enhancing machining precision, prolonging tool lifespan, and improving tool reliability. This study proposes a tool wear prediction model that integrates principal component analysis with a support vector machine. Initially, features from the time domain, frequency domain, and wavelet packet decomposition are extracted from multi-source signals. These features are then filtered using the Spearman correlation coefficient to retain those most strongly associated with tool wear. Subsequently, principal component analysis is applied to reduce the dimensionality of the filtered features while preserving the most representative components. A support vector machine regression model is employed to predict tool wear, with key hyperparameters, including the penalty factor and kernel width, optimized through a grid search to ensure high prediction accuracy. Comparative evaluations with CNN-SVM and LSTM-SVM models demonstrate that the proposed PCA-SVM model outperforms the others in terms of prediction performance, thereby validating its effectiveness.