Research On Tool Wear Monitoring Based On Multi-sensor Fusion

In the process of machining,tool wear will destroy the surface quality of the workpiece,and tool failure caused by severe tool wear is one of the important factors of cutting equipment failure.If the tool wear can not be accurately monitored,it is unreasonable to estimate the tool life.It will cause premature tool replacement and increase the processing cost.If the cutting tool is not replaced in time before failure serious accidents of destroying the processed workpiece and even damaging the processing equipment could happen.A reliable tool wear monitoring system is essential to the machining process in mass production.In this paper,cutting experiments are carried out in NC machining center.Using Multi-sensor fusion technology.The sensor signals related to tool wear are collected,analyzed and processed,and feature fusion is carried out to build a monitoring model to monitor tool wear.The main research contents are as follows:By milling QT-700 workpiece with cemented carbide disc milling cutter,a complete tool wear process was obtained by milling experiment.According to the tool wear curve,the wear process can be divided into initial wear,stable wear and severe wear.During the experiment,the sensor signals corresponding to each wear amount of the tool were obtained,including cutting force,vibration and acoustic emission signals.The change of signals with tool wear was analyzed,and the results showed that there is a good correspondence between the collected signals and tool wear.The features of each signal were extracted in time domain,frequency domain and time-frequency domain.A total of 175 features are extracted from 7 signals.The change of each signal feature with tool wear was analyzed.The results showed that there is a good mapping relationship between some signal features and tool wear.By calculating the absolute value of the correlation coefficient between each feature and tool wear,the features with low correlation with tool wear are removed,and finally 19 feature values were selected to form the fusion feature vector.A BP neural network monitoring model was constructed to monitor the tool wear.The input items of the model were 19 extracted feature quantities,and the output result was the tool wear.Because the BP neural network is difficult to determine the network structure,the gravity algorithm(GSA)was applied to optimize the initial weights and thresholds of the network,and the GSA-BP monitoring model was established.Training two models,and testing the monitoring accuracy of the two monitoring models.The absolute value of relative error between the predicted value and the actual measured value of the two monitoring models was calculated.The average relative error of BP neural network monitoring model is 4.87%,and the average relative error of GSA-BP monitoring model is reduced to 2.73%,which greatly improves the monitoring accuracy compared with BP neural network monitoring model.In order to verify the universality of GSA-BP monitoring model,a tool and workpiece material were re-selected,and the number of training samples was reduced by half.The average relative error of monitoring results of eight groups of test data was3.28%,which maintained high testing accuracy.In order to be closer to the actual machining requirements,a monitoring model of tool wear with variable parameters was established.Collecting signal data and tool wear under different cutting parameters,changing the network structure,and redesigning the GSA-BP monitoring model,the results show that the designed monitoring model has the ability of monitoring tool wear with variable parameters,and has high testing accuracy.