关键词: 刀具磨损, 剩余寿命预测, 特征选择, 小波包分解

Abstract: The real-time detection of drill wear is helpful to the preventive maintenance which will give caution of tool change in drilling production process. For the problem of tool monitoring during automatic production, a strategy of drill bit wear monitoring and remaining life time prediction system is presented for drilling machine with the use of spindle current signal. The spindle current information is acquired by closed loop current sensor, and a moving window technique is used to extract the real parts of data for drilling from the sampled data sequence. The wavelet packet decomposition is used to extract features. Critical features are selected according to their ability of discriminating the wear under Fisher criterion. Logistic regression (LR) combined with auto-regressive moving average (ARMA) model are used to evaluate the failure possibility and remaining life of the drill bit. Experimental results show good performance of the proposed algorithm and it can provide the basis for tool change.

Key words: Feature selection, Prediction of remaining life, Tool wear, Wavelet packet decomposition