At present, many kinds of sensors are used for on-line monitoring of cutting process, tool identification and timely replacement. However, most of the original monitoring signals extracted from the cutting process are time series signals, which contain too much process noise. As the signal noise is relatively low, it is difficult to establish a direct relationship with the tool wear. Therefore, how to obtain the effective information from the online monitoring signal and extract the characteristics that can directly reflect the tool wear from the complex original signal, so as to establish an effective and reliable tool wear monitoring system, is the key and difficult problem in the research of the online monitoring technology of tool wear. Firstly, an experimental platform based on the force sensor for on-line monitoring of tool wear was built, and the signal obtained by the force sensor was used to monitor the tool wear, and the feature information was extracted and fused. The innovation of the project lies in the use of Gaussian process regression (GPR) method to predict the tool wear, the use of feature dimensional rise technology, to reduce the impact of noise, on the premise of ensuring the prediction accuracy, improve the confidence interval of GPR prediction results, improve the stability and reliability of the monitoring process.
On-line tool wear monitoring based on machine learning