Research on Quantitative Monitoring Method of Milling Tool Wear Condition Based on Multi-Source Data Feature Learning and Extraction

Abstract

Milling tool wear monitoring has great significance to guarantee the workpiece quality. However, it is difficult to be quantitatively monitored online. To solve this problem, an online wear quantitative monitoring method (WQM) for milling tools is proposed based on a convolutional neural network (CNN) and principal component analysis (PCA). A wear indicator (WI) is constructed to quantify milling tool wear in this paper. First, the multi-source data collected in cutting process are preprocessed, and the top 30% of them are used to train a CNN. Then, the online monitoring data are input into the trained CNN to obtain deep fusion features (DFF). And, 3 time-domain features and 3 frequency-domain features are extracted from the DFF. Finally, PCA is used to remove redundancy and correlations of the 6 features, and the smoothed first principal component (FPC) is used to construct the WI. It is called convolutional neural network-principal components-based wear indicator (CNNPCWI). It is verified by cutting experiments of 4 milling tools under 3 working conditions. The results show that CNNPCWI is superior in monotonic trend and correlation with number of cuttings than manually extracted features. And it is more conform to the milling tool wear trend, which can be used to quantify the milling tool wear online.