Prediction of tool wear during linear feed milling using multi-sensorial data fusion

摘要

This paper presents a tool wear prediction model based on multi-sensorial data fusion, which integrates both local preserving projection (LPP) and support vector regression machine (SVR). The time-domain features are extracted from spindle current and feed motor torque signals, and the multi-sensorial features are fused by three dimension reduction algorithms of PCA, KPCA and LPP. To compare the fusion effect, SVR model was used to establish the mapping relationship between the fusion feature and the tool flank wear value, and the tool wear value was predicted. The experimental results demonstrate that the effect of feature fusion using LPP method is better than that of PCA and KPCA. In addition, in order to eliminate the motor current fluctuation and unexpected factors in cutting process, wavelet packet transform (WPT) algorithm is used to filter the extracted features to further improve the accuracy of tool wear prediction.