Tool Wear Condition Monitoring And Remaining Life Prediction Research

In the machining cutting process,the tool is an extremely important part of the whole process,because the degree of wear of cutting tools will directly affect the efficiency of production and processing,as well as the surface quality of the finished cutting products,machining accuracy,etc.,when the cutting tool wear has been relatively serious when not replaced in a timely manner,will cause the cutting tool breakage,increasing production costs and even lead to cutting and processing of finished products When the cutting tool wear has been relatively serious,but not replaced in time,it will cause the cutting tool breakage,increase the production cost and even lead to the scrapping of the finished cutting products and the damage of the processing machine.Therefore,this paper focuses on tool wear monitoring and proposes a dual-channel input long and short-term memory network + convolutional neural network(CNN-LSTM)for high-speed milling tool wear monitoring and a new deep neural network,i.e.,multi-scale recurrent convolutional neural network(MSRCNN)for high-speed milling tool remaining life prediction.The research in this paper is as follows.(1)Milling tool wear forms and wear mechanisms are studied;the trends of multiple sensor signals in the open data set,the relationship between multiple sensor signals and tool wear values are analyzed;a high-speed milling field test is designed,and a large amount of vibration signal data is collected for tool wear state monitoring by a laser vibrometer in the field test;the milling tool wear signal data preprocessing process extracts relevant time and frequency domain features from the original signal in the pre-processing process of milling tool wear signal data;then the features with less interference and sensitive to tool wear were finally selected for the subsequent model by using the distance-based feature screening criteria.(2)For the tool wear monitoring experiments,three models of one-dimensional convolutional CNN neural network,BP neural network and random forest are used to establish the tool wear status classification model,and a two-channel input CNN-LSTM tool wear status monitoring model is proposed,and the experimental prediction results of different models are compared to prove that the two-channel input CNN-LSTM tool wear status monitoring model The accuracy is better than the other three models,which lays the foundation for tool wear condition monitoring.(3)For the tool wear monitoring experiments,this chapter uses CNN and RNN measures to establish the tool remaining life prediction model respectively,and proposes a new deep neural network based on multi-scale recurrent convolutional neural network(MSRCNN)for high-speed milling tool remaining life prediction,illustrates the multi-scale recurrent convolutional neural network(MSRCNN)model structure,parameter settings,training tuning,and finally the experimental results show that the proposed MSRCNN has obvious advantages in terms of accuracy and convergence compared with the existing CNN-based prediction measures and RNN prediction measures models,and the method breaks through the limitations of the convolutional neural network prediction model in this field and provides a theoretical basis for tool remaining life assessment.