Research On The Prediction Of Tool Wear State By Using The Support Vector Machine

Nowadays, the process of parts has achieved full automatic production mode,in this mode, how to ensure the precision and surface quality of machined parts is very important,which will directly affect the processing cost and processing efficiency. The appearing of Tool wear monitoring technology can solve such problems in machining, so the important practical significance of the tool wear state detection is self-evident. According to the characteristics of machining condition, this paper selected cutting force and acoustic emission signal as a monitor signal,the cutting force increases with increasing friction between the tool and the workpiece, tool wear can be characterized the obvious.In the process of cutting,the acoustic emission signals from the cutting zone, it contains a wealth information of tool wear and acoustic emission signal of high frequency(100KHz-1MHz),can be very good to avoid the environmental noise pollution area of low frequency.This paper to detect tool wear state do the following work:1.Tools for cutting force signals were analyzed using time domain and wavelet packet method,so the Time domain features and wavelet packet decomposition of each frequency band energy feature can be getted.The acoustic emission signal empirical mode decomposition(EMD) decomposition, the true intrinsic mode function(IMF) to get the percentage component of the energy with the original signal energy as feature values.The Principal component analysis(PCA) dimension reduction method, is adopted to reduce the dimensionality of feature vectors, to remove the correlation, reduce data redundancy.2.Build support vector machine classifier model to realize the accurate classification of once after each pass tool wear. Further build support vector machine regression model to realize the tool wear prediction of once after each pass tool wear.Using the GUI interface of MATLAB, designed the intuitive and convenient man-machine interactive interface, so as to realize the simple operation of the tool wear state detecting system and prediction of tool flank wear value system.Through a series of research work in this paper, the signal feature extraction and support vector machine classifier and regression model was combined for the tool wearstate detection,so as to get the expected results and provides a new method for the further implementation of tool wear condition monitoring.