Research On Milling Cutter Wear And Failure Detection Method Based On Machine Vision

The cutting tool occupies a very important position in the machinery manufacturing industry.The quality of the cutting tool condition not only directly affects the cutting quality of the workpiece,but also affects the life of the CNC machine tool.In the process of processing and production,if the tool condition can not be monitored in a timely and effective manner,it will lead to poor workpiece processing quality,the interruption of the production process,and even cause a series of problems such as material waste and machine tool failure,which will cause huge losses to enterprises.Therefore,it is very necessary to study the tool monitoring method in order to realize the tool condition monitoring.In this thesis,the coated carbide milling cutter is taken as the research object.By using the method of machine vision,the wear of the milling cutter is directly measured,and the failure of the cutter is predicted based on the surface texture of the processed workpiece.The specific contents are as follows:Firstly,the characteristics of tool wear and wear process are analyzed,and it is clear that the wear mainly occurs on the main back face of the milling cutter.Determine the milling cutter blunt standard to provide the discriminant basis for tool failure.The variation law of surface texture of the machineed workpiece is analyzed to provide theoretical basis for tool failure detection.According to the structure characteristics of the milling cutter,the image acquisition scheme is designed.Reasonable selection of hardware equipment,and write dual-camera image acquisition software,so that the image data acquisition more convenient.A CCD industrial camera and a 75 mm lens were used to capture the milling cutter image,and the workpiece was photographed by a microscopic camera to achieve data acquisition.Then,machine vision method is used to measure the milling cutter wear directly.In view of the stop position of the tool tip is random,taking the bottom image of milling cutter as the research object,a method of secondary mask is proposed to extract the straight line of the blade,and the least square method was used to fit the straight line so as to calculate the rotation Angle of the milling cutter.After determining the position of the tool tip by rotating positioning,taking the side wear image of milling cutter as the research object,the wear characteristics of ROI region of tool tip were extracted by image processing method.The method of minimum outer rectangle based on image rotation was used to measure the tool wear directly,which solved the problem that the outer rectangular edge could not fit the edge of the blade.Finally,according to the variation of workpiece texture before and after tool failure,machine learning method was used to detect tool failure.The collected workpiece texture image is preprocessed to make the texture image clearer.The method of optimizing GLCM by wavelet was used to extract the image texture features so as to get more texture information.The extracted texture features are reduced by PCA to reduce the redundancy of the data,and the BP neural network and GA-BP neural network were trained with the dimensionality reduction data.Then the CNN neural network is trained with these workpiece texture images,and the experimental data are used to verify the three models.Through comparison,it is found that the CNN neural network model has a better effect on tool failure prediction,the detection accuracy of verification set and test set reached 98.46% and 92.3% respectively.