Research On Real-time Growing Segment Regression Model Of Tool Wear In Machining Process

In practical engineering,some phenomena have the features of timeliness and phased change.The timeliness of these phenomena will lead to the prediction model being effective only in the short term,thus requiring continuous updating of the models,and if the phased change are not taken into account in establishing the prediction model,it may increase the complexity of modeling calculation,influence the prediction effect and even lead to the failure of the model.Tool wear in the process of machining has obvious phase features of initial,normal,rapid wear and failure phase,and in the process of batch precision machining,the accurate prediction of tool wear is an important prerequisite for adjusting and controlling processing parameters to ensure stable processing quality and a key part of intelligent processing technology.Therefore,in order to realize online real-time prediction of tool wear,it is of great engineering practical significance to explore the method of establishing time-predictive model with time features to improve the prediction accuracy of tool wear.Based on the phase features of tool wear,this paper studies the real-time prediction model of tool wear.(1)Through the analysis of feature curves of timeliness and obvious phase change in engineering practice,this paper proposes a modeling method for real-time growth segment regression(RGSR)model.The idea of modeling is to collect the minimum sample from the starting point to establish the initial prediction model,and then to implement on-site detection,validation of model satisfaction,modify prediction model,segment judgment and segment point determination,and segment prediction model establishment,and cycle through these links,the modified model or prediction model with segment features varying with time is established respectively.(2)The segment judgment and segment point determination are the core content and key links of RGSR model modeling.By studying the method of segment judgment,an absolute error(AE)region segment method based on Gauss process regression(GPR)is proposed,that is,by analyzing the AE of curve fitting based on GPR model,the segment judgment is performed by using the method that the segment point is at or near the AE peak point.The method can judge whether the process change has entered a new phase in time and find out the segment point of its phase change,through various application examples,the validity of the AE region segment method is verified.(3)Through the tool wear experiments of AISI 1045 and stainless steel,the RGSR model was demonstrated by GPR and polynomial regression.When constructing the RGSR model with GPR,the AE region segment method is adopted as the method of segment judgment and segment point determination.When constructing RGSR model with polynomial regression,the prediction error and the order difference before and after model modified are used as the basis of segment judgment and segment point determination.The experimental results show that prediction model established by segment can effectively improve the prediction accuracy and effect,which is more in line with the tool wear process in actual machining.Moreover,because of the small sample size required by the segment prediction model,the computational complexity is lower than that of the whole prediction model,which can reduce the order of the model when using polynomial regression predictive modeling.