Tool Wear And Life Prediction Based On Deep Learning And System Development

The transformation of Intelligent Manufacturing in the manufacturing industry will further integrate the new generation of information technology with the traditional manufacturing industry.In the future factory environment,all kinds of sensors will continuously collect monitoring signals related to equipment.As the basic equipment of manufacturing industry,tool wear will affect the product quality and production efficiency.Therefore,the recognition of tool wear status,the prediction of tool wear amount and tool remaining life are of great significance to improve product quality and production efficiency.In view of the above problems,this paper studies how to use tool condition monitoring information to accurately predict tool wear and tool remaining life.The main research contents are as follows:(1)the preprocessing of tool condition monitoring signal is studied,including removing null value,detecting and removing outliers,removing polynomial trend.(2)The feature extraction method of tool condition monitoring signal is studied from three aspects: time domain feature extraction,frequency domain feature extraction and time frequency domain feature extraction.In the time domain,the statistical feature information of the signal is taken as the time domain feature of the signal.In the frequency domain,the frequency domain feature of the signal is extracted by amplitude spectrum and power spectrum analysis,and the time frequency domain feature of the signal is extracted by empirical mode decomposition.(3)Considering the risk of high false discovery rate in multiple hypothesis testing as feature selection method,Benjamin-Yekutieli method is introduced as feature selection method,which not only ensures the effectiveness of selected features,but also controls the false discovery rate.(4)A two-stage prediction method based on deep learning is proposed in this paper.The prediction method unifies the classification of tool wear status and the prediction of tool wear value in the same method,which can simultaneously predict the current wear status classification and wear value of the tool.The prediction model of tool remaining life is established by using dropout deep feedforward network.The experimental results show that the performance of the proposed two-stage prediction method is better than that of support vector machine in terms of wear status classification and wear value prediction.In addition,compared with the tool remaining life prediction model based on support vector machine,the remaining life prediction model based on dropout deep feedforward network can better fit the real remaining life curve.Finally,based on Matlab GUI,Java and SQL Server database,this paper develops a tool wear and remaining life monitoring system to achieve data acquisition and analysis,tool wear and remaining life prediction,visual monitoring and other functions,and improve the level of enterprise information management.