Prediction of Tool Wear by Improved Deep Belief Networks Based on Multisensor Signals

Abstract Tool wear online monitoring technology is crucial for enhancing machining efficiency and product quality. This article proposes a tool wear prediction method based on a particle swarm optimization (PSO)-optimized deep belief network (DBN) to improve prediction accuracy. First, features are extracted from sensor signals through time-domain, frequency-domain, and time–frequency-domain methods. Subsequently, a correlation analysis method is employed to identify wear-sensitive features of the tool. Finally, the proposed PSO-DBN model is rigorously evaluated using cross-validation, demonstrating its effectiveness in accurately predicting tool wear across all test cases. Comparative experiments against established models and state-of-the-art approaches show that the PSO-DBN model achieves superior performance, with lower root mean squared error (RMSE) and mean absolute error (MAE) values and a higher R2 score, confirming its precision and reliability in tool wear prediction.