Research On On-machine Tool Wear Detection System Based On Machine Vision In Milling Process

With the widespread application of computer technology,the mechanical manufacturing industry is developing in the direction of intelligence,digitalization and networking.In particular,the rapid development of aerospace,automotive and other manufacturing industries places higher requirements on the processing quality and efficiency of modern manufacturing industry.However,the tool wear status affects the cutting efficiency and the quality of workpieces.So it is necessary to build an efficient tool wear condition detection system.It can obtain the tool wear status and wear loss during processing,optimize the cutting parameters,reasonably select tool change time,improve tool utilization and cutting efficiency,and save costs.This paper studies on-machine detection system of milling tool wear based on machine vision.The main research contents are as follows:The form and process of tool wear were studied.The weighted average graying method,adaptive hybrid filter noise reduction method,contrast linear stretch method,and OTSU adaptive threshold segmentation algorithm were used to preprocess the tool wear image.Closing operations and area filling were used to close the holes and complete the edges of worn area binary images.Based on the improved Zernike moment method,the pixel-level edges obtained by Lanser operator were relocated to obtain contour edges with sub-pixel accuracy.According to the wear characteristics of ball-end mill and end mill,different image processing algorithms were designed to extract tool wear loss.A set of milling tool wear image on-machine acquisition device was designed.Different image dynamic acquisition schemes were designed for end mill,ball-end mill and disc milling cutters,the spindle speed and acquisition interval were determined.The wear detection experiment was performed on the three milling cutters mentioned above.By comparing with the measured value of opticalmicroscope,the average error rate of the maximum flank wear value was 2.95%,which verified the reliability of the system.The basic structure of CNN and related technology of network parameter optimization were studied.An automatic recognition method of milling tool wear type based on CNN was proposed.The network structure was designed,and this recognition model constructed was trained and tested with the images collected by experiments.The performance of the model was evaluated by comparing with the recognition accuracy of VGGNet-16 model.The tool wear on-machine detection system software was developed.Use the MFC application module in Visual Studio 2015 development tool and combined with the SDK development kit that come with the camera to develop the wear image acquisition module and software window function,combined with Open CV library,Halcon library and deep learning technology to mixed programme for developing tool wear image processing module and wear status intelligent identification module.