Research On Milling Tool Wear Monitoring Method Based On Multi-Kernel Function Support Vector Regression

Milling is a common method of metal cutting,the milling tool wear will cause dimensional accuracy and surface quality of the workpiece to be reduced.High-precision milling tool wear monitoring can provide decision support for tool change decisions and tool compensation during machining,thereby improving workpiece quality and processing efficiency.The current research of milling tool wear monitoring mainly extracts the tool wear features from the statistical features of the data.It lacks comprehensive consideration of the data’s multi-dimensional features and time series features,and cannot fully reflect the milling tool wear conditions,thereby affecting the performance of milling tool wear monitoring.Therefore,this paper takes the tool wear in milling processing as the research object,and studies the tool wear monitoring method based on multi-Kernel function support vector regression(MKSVR),aiming at the single dimensional,multi-dimensional and time series characteristics of the monitoring signal.The specific research is mainly in the following aspects:Firstly,the monitoring signal of milling tool wear is analyzed based on the characteristics that the monitoring signal has single-dimensional characteristics,multidimensional and time series characteristics;The data input layer,preprocessing layer,feature extraction layer,feature selection layer and regression prediction layer are designed in detail.Then,the key technologies of the construction feature space are studied,including MKSVR-oriented milling tool wear feature extraction method and feature selection method based on collaborative filtering.The former performs feature extraction through time-domain frequency domain and wavelet domain analysis,densely connected convolutional neural network and long-short-term memory neural network;the latter uses variance,correlation coefficient,maximum mutual information coefficient and random forest to select and form the extracted features Milling tool wear feature space.Finally,on the basis of the above research,a milling tool wear prediction model based on MKSVR is proposed to realize the mapping of the feature space to the milling tool wear value to predict the milling tool wear.By comparing and analyzing with other milling tool wear prediction models,the effectiveness of the MKSVR-based milling tool wear prediction model is verified.