Tool Wear Detection Method Based On Convolution Neural Network

Milling is a machining process that uses rotating tools to remove workpiece materials,and is a commonly used machining method in mechanical manufacturing.The degree of wear of milling cutters during the machining process directly affects the quality of the workpiece,while the degree of wear of milling cutters also indirectly affects production safety,efficiency,and economic costs.Therefore,it is of great significance to study the wear status of milling cutters during the machining process.This article proposes the method of using convolutional neural networks to extract features from two-dimensional images and surface texture images of milling workpieces converted from time series physical quantities related to milling cutter wear,and achieve classification of milling cutter wear degree.The main research content of the paper is as follows:First of all,the research methods and feature extraction methods of tool wear at home and abroad are analyzed,and the convolution neural network method is adopted for feature extraction in this paper.The mechanism of tool wear is analyzed;A detailed introduction was given to several models of deep learning,especially convolutional neural networks;Introduced the evaluation criteria for deep learning models;Laying the foundation for subsequent research.Secondly,a method for detecting the wear degree of milling cutters based on time series data and GAF-CNN is proposed.Use the Gram angle field to convert the timeseries physical quantities related to wear collected during milling into two-dimensional space,and improve se resnet by applying the improved se resnet to the PHM2010 and Nasa milling cutter wear datasets.The experiment has proven that the improved model performs better than the original model,and the comparison with other models also proves that the method has better performance.Thirdly,a lightweight convolutional neural network milling cutter wear degree detection method based on texture image is proposed.The experiment of texture image acquisition on the surface of milling workpiece is designed,and it is proposed to use the gray level co-occurrence matrix to extract texture features from the texture image.Because the amount of data obtained is small,the data set needs to be expanded by data enhancement.In order to facilitate the deployment of the tool wear classification network on mobile devices,the lightweight convolutional neural network ShuffleNetV2 is improved and applied to the surface texture data set of the milling workpiece to realize the classification of the wear degree of the milling cutter.The experimental results show that the accuracy of this method to identify the wear degree of milling cutter is high,and the comparison with other models also shows the superiority of this method.Finally,the recognition effects of the two proposed methods are compared.Because the lightweight convolutional neural network milling cutter wear degree detection method based on texture image takes less time to process the image,and the model size is smaller.It is concluded that the lightweight convolutional neural network milling cutter wear degree detection method based on texture image has more advantages.