Research On Tool Wear Condition Monitoring And Life Prediction Based On Multi-sensor Information Fusion

Tool wear is a prominent phenomenon in metal cutting processes,particularly in the cutting of difficult-to-machine materials.It not only compromises the accuracy and quality of machined parts,but also leads to a reduction in the service life of the tool.In severe cases,tool wear can even impact the precision of the machine tool or cause damage to the tool itself.Therefore,it is of great practical significance to carry out research on monitoring of tool wear condition and prediction of tool remaining life.In this paper,a tool wear monitoring experiment was conducted by using multiple sensors to collect monitoring data,and information mining was applied to the data.A method utilizing bidirectional long short-term memory networks for monitoring tool wear conditions and predicting remaining tool life was proposed.By analyzing the tool wear process,the milling experiment of 300 M steel which is difficult to process was designed.The monitoring data of vibration signals and force signals are selected for feature analysis and extraction from three aspects: time domain,frequency domain and time frequency domain,and 17 sensitive features are selected by mutual information method to form the optimal special collection.A tool wear monitoring model based on bidirectional long short-term memory network(Bi LSTM)was established and its monitoring effect was tested.The average of the model reached 97.8%.Compared with BP neural network,support vector machine(SVM),and long short-term memory network(LSTM),the recognition accuracy of the Bi LSTM network was improved to some extent.Based on the tool milling lifecycle data,a hybrid model combining Parallel Channel Rest Net and bidirectional long short-term memory network(PCResNet-BiLSTM)was established to realize automatic feature extraction and complete the relationship mapping from the original monitoring data to the tool wear.Furthermore,Bi LSTM model was established on the basis of the wear amount after multiple tool runs.The wear trend before tool scrapping was predicted by 50 cutting strokes in advance,and the remaining cutting stroke from the current time to the tool scrapping time is output,that is,the remaining life.The results show that the tool wear trend error(MAE)was within 5μm,and the cutting stroke error was within 10 strokes,which can meets the actual machining requirements.The tool wear management system was developed.All the researches are visualized based on the APP Designer module in matlab software,and the functions of data analysis and processing,tool condition monitoring and remaining life prediction are realized.Independent application software could be generated after the packaging was completed.It provides a technical reference for the practical application of intelligent monitoring in tool wear.