Research On Tool Wear Monitoring Based On RFE Algorithm

In the field of mechanical manufacturing,cutting tools are the focus of many researchers.In the machining process,the state of tool wear will have a great impact on the workpiece quality.In case of serious tool wear,it will directly affect the normal operation of processing equipment.Premature tool replacement will lead to the increase of processing cost;Failure to replace the cutting tool in time will lead to unqualified workpiece processing quality and even damage to the processing equipment.Therefore,for the whole manufacturing industry,the research on tool wear monitoring is very necessary.By analyzing the mechanism and process of tool wear and the relationship between tool wear and sensor signal,the tool wear state is divided into three stages,and the signals to be collected are determined.Aiming at the problem that the original sensor data can not directly reflect the change of tool wear state in the machining process,the signal processing technology is used to preprocess the original signal.Remove the invalid signals collected during feed and withdrawal in the cutting process,and eliminate the singular points in the original signal by filtering method.In the time domain,13 features are extracted from each signal;In the frequency domain,10 features are extracted from each signal;In time-frequency domain,wavelet packet energy entropy is extracted for each signal.Aiming at the problems of high data dimension and complex data matrix.Logical regression,classification and regression tree,linear regression and linear discriminant analysis are used as the base model of recursive feature elimination algorithm.The optimal number of selected features is determined by recursive feature elimination based on cross validation.The recursive feature elimination of different base models is used to rank the extracted features.Through the optimal feature number and feature ranking,four groups of optimal feature sets are finally obtained.The extreme k-nearest neighbor algorithm and the best nearest neighbor algorithm are used to train the extreme k-nearest neighbor model respectively.The algorithm model is verified by ten fold cross validation and validation set.The comparative analysis shows that the recursive feature elimination algorithm based on classification and regression tree combined with extreme random tree algorithm can effectively monitor tool wear.A complex tool state information monitoring platform is developed based on Python and Lab VIEW.The platform is divided into four modules: basic information,status detection,data processing and tool process management.It can write and query the basic information of machine tools,operators,materials and tools.At the same time,it can collect data online in real time,warn and process data offline in the later stage.