An Innovative Study for Tool Wear Prediction Based on Stacked Sparse Autoencoder and Ensemble Learning Strategy

Accurately predicting tool wear in real time is crucial to enhance the tool prognostics and health monitoring system in computerized numerical control (CNC) machining. This paper proposed a novel integrated deep learning model for predicting the wear of milling tools by fusing multi-sensor features. The raw signals of vibration and cutting force acquired from the continuous cutting cycle were used to extract multi-sensor features throughout the full lifecycle of the milling tools in time, frequency, and wavelet domains, respectively. The sensitive features from these signals were identified through correlation analysis and used as input for the stacked sparse autoencoder (SSAE) model with backpropagation neural network (BPNN) as the regression layer to predict tool wear. SSAE models with different activation function configurations of hidden layers were utilized to construct deep neural network models with different prediction performance, which were taken as primary learners of integrated deep learning model. The intergrated SSAE model based on the stacking learning strategy applied the gradient boosting decision tree (GBDT) regression model with Bayes optimized hyperparameters as the secondary learner to predict tool wear. Compared to the single SSAE model and shallow machine learning models, the proposed method significantly improved both the prediction accuracy and reliability.