Method For Monitoring Wear Status Of End Mill Based On Watershed Algorithm

In order to promote China’s transformation from a large manufacturing country to a strong manufacturing country,and to narrow the gap between China and other strong manufacturing countries such as the United States and Japan,it is imperative to improve the level of intelligence and automation in our manufacturing industry.The "Made in China2025" plan proposes to develop high-grade CNC machine tools with functions such as depth perception and intelligent decision-making,and the wear status of the tool directly affects the efficiency and machining accuracy of the CNC machine.Therefore,the development of a tool condition monitoring system that can be integrated into the machine tool is of great significance for the improvement of the level of intelligent manufacturing and automated manufacturing.In order to achieve accurate and rapid monitoring of tool wear status,a machine vision-based tool wear status monitoring method was studied,and the main research contents were as follows:1.In view of the limitations of the commonly used area segmentation methods such as susceptibility to noise interference,the decision was made to use the watershed algorithm for the area segmentation of the tool wear image,then the basic idea and correspondence of the watershed algorithm processing images were studied and the mechanism of the watershed algorithm was elaborated by using the submersion simulation model as an example,meanwhile,the applicability of the watershed algorithm for the segmentation of the tool image was determined after analyzing the distribution characteristics of the pixel grayscale values and gradient values of the tool wear image,and the over-segmentation problem was found and analyzed in the processing results of the watershed algorithm,then it was clarified that the noise and background texture in the image were the causes of the over-segmentation problem.2.In order to eliminate the noise and background texture components in the tool wear image,the morphological component analysis algorithm was proposed to decompose the tool wear image sparsely,then the components of the tool wear image and the morphological differences between them were clarified after analyzing the tool wear image from the morphological point of view,thus,the applicability of the morphological component analysis algorithm for the sparse decomposition of the tool wear image was determined,and the dictionary selection method for each component was investigated,finally,curvelet transform and local discrete cosine transform were selected as the dictionaries for the target tool image and background texture image,and the noise and background texture components were successfully decomposed from the tool wear image.3.The effectiveness of the above algorithms was verified through a series of image experiments and a set of tool inspection application system based on FPGA + DSP architecture was developed,meanwhile,the system was successfully integrated in the CNC machine tool as a separate functional module,which realized the functions of image acquisition,image segmentation,feature extraction and status recognition in the processing cycle,finally,the practicality of the system was verified through the engineering applications in the processing site.