Research On Online Monitoring And Prediction Method For Wear Condition And Wear Value Of End Mill Cutter

As a typical intelligent manufacturing equipment,intelligent machine tool can independently decide the execution flow of each link in the manufacturing process.It can monitor,diagnose and repair all kinds of deviations in the production process,provide the optimal solution for processing and production,calculate and forecast the remaining life of cutting tools,spindle,bearing and guide rail,and provide the remaining service time,replacement time and current status.Therefore,self sensing,self-learning,self-adaptive and self-optimization of machining state of intelligent machine tools are one of the key technologies that need to be solved urgently,and the on-line monitoring technology of tool wear is an important part of it.The on-line monitoring of tool wear status and wear amount can provide decision-making basis for timely tool replacement,radius compensation,tool path optimization and failure warning of intelligent machine tools.The prediction based on online monitoring expands the time scale for formulating timely,effective,reasonable and accurate implementation strategies.All of these have important engineering application value and theoretical research significance.As one of the most widely used cutting methods,milling has the characteristics of various processing conditions,variable cutting parameters,discontinuous cutting process,complex and strong randomness of milling cutter wear mechanism,which makes it difficult to form an accurate and universal tool wear online monitoring and prediction theoretical support.In order to solve the problem that the accuracy and generality of the existing tool wear monitoring technology of high-grade CNC machine tools are restricted each other,this paper takes the integrated three edge end mill with complex morphology as the research object,and takes the accuracy judgment of milling cutter wear status and the adaptability of monitoring or prediction methods as the research goal.The research is carried out from four aspects: the judgment and measurement of end mill wear state,on-line monitoring of wear state,on-line estimation of wear amount,and on-line prediction of wear state and wear amount.The main work and achievements are as follows:(1)The paper analyzes the wear pattern of integral three edge end milling cutter in cutting process,and puts forward a comprehensive judgment method for the wear state of milling cutter based on the combination of width of flank wear and acreage of rake wear.Aiming at the problem that it is not easy to obtain the image of milling cutter wear,a rapid measurement device for flank wear and rapid acquisition of positioning support for rake surface images are developed.By means of image preprocessing,image registration and image difference,a fast calculation method for the edge defect area of the rake face is proposed,which solves the problem of obtaining the milling cutter wear image and calculating the wear amount.By comparing the performance of two indicators,the defect area of rake wear and the width of flank wear,in judging the wear state of the milling cutter,combined with the analysis of its influence on the integrity of the cut surface of the workpiece,the rationality and effectiveness of the judging method are verified,which makes up for the defect that the accuracy of monitoring is reduced by using a single method to determine the tool wear state.(2)From the perspective of data-driven on-line monitoring of tool wear status,an online tool wear state classification and recognition model based on "single-layer sparse auto-encoder + classifier"(SSAE network model)is established to adaptively extract the fusion signal features of cutting force and cutting vibration,which overcomes the dependence of expert experience in extracting cutting signal features.In view of the deficiency that the SSAE network model must supervise the training process of tool wear experience data,an unsupervised tool wear condition online monitoring method is proposed based on sparse auto-encoder(SAE).The feature of the fusion signal of force and vibration is extracted adaptively,and the input signal is reconstructed to form error sequence.By mining the internal relationship between the reconstructed error series and tool wear,the threshold and criteria for judging tool wear state are determined,and the corresponding online monitoring strategy is designed.The online monitoring of initial wear,stable wear,severe wear and failure of milling cutter is realized.The limited conditions and manual intervention in the process of on-line monitoring tool wear state of high-grade CNC machine tools are reduced.The effectiveness of the method is verified by several milling experiments in two different milling environments.(3)From the perspective of online monitoring of tool wear driven by the fusion of data and physical models,a cutting force model under end mill wear is established.According to the analysis of the influence of tool wear on the force model parameters in the cutting process,the wear-related parameters are integrated into five force model coefficients to simplify the cutting force model.Taking the force signal collected in the cutting process as a bridge,the force model coefficients are separated from the milling force model.Through the correlation analysis between the force model coefficients and the wear trend of the rake and the flank of milling cutter,the flank wear estimation model is established to avoid a large number of force model parameter calibration experiments and calculations.On this basis,an unsupervised online tool wear estimation method is proposed.This method can be better adapted to the new cutting environment.Before estimating the flank wear value,only three small distance cutting experiments with blunt cutter are needed to correct the estimated model parameters,which improves the adaptability of the online monitoring model or method for tool wear of CNC machine tools.The practicability of the proposed method is verified by experiments on different cutting conditions under two tool types.(4)On the basis of unsupervised on-line monitoring of wear state and wear amount,time series analysis method is used to reveal the characteristics,laws and stability of monitoring time series.Then,the methods of identification,order determination,parameter estimation,and adaptability testing of the prediction model are explored,and an on-line iterative prediction method of tool wear state and wear amount based on ARIMA is proposed.Continuously update the training set composed of historical value and current value.The wear state and wear amount of milling cutter in the future can be predicted with a certain step length.Thus,the time scale for intelligent machine tools to formulate timely,effective,reasonable and accurate execution strategy is expanded.The rationality and feasibility of the prediction method are verified by several experiments under different cutting tool types and cutting parameters.