Research On Full Surface Defect Inspection Of Bearing Rings System Based On Machine Vision

Bearing is the foundation of the modern industrial parts and components.It has been hailed as a "joint" of mechanical equipment.Bearing ring is an important component of the bearing and it can produce all kinds of defects on the surface due to various reasons in the process of production,which not only has a harmful effect on the bearing ring appearance but also affects the precision equipment and greatly reduces the service life of the machine.At present,most of the bearing enterprises in China are still in the stage of manual visual inspection,which is of low efficiency,high cost and often defective products miss inspection.In view of the problems faced by companies in bearing defect detection,this paper designs a set of bearing ring full-surface defect detection system based on machine vision,which can realize real-time,reliable,rapid and accurate detection of workpiece.The main research contents of this paper include:(1)Overall scheme design of the detection system.Analyze the working condition of bearing ring production line and put forward the overall scheme of the design of the detection system on the basis of the design requirements,which includes image acquisition system,mechanical drive system,control system and detection software.The software technology is determined where the detection software is based on LabVIEW platform and Python script and the controlling system adopts Mitsubishi PLC.At the same time,the working process of the whole system is designed.(2)Main structure design of the detection system.The image acquisition system is designed and selected where the annular shadowless light source and the bowl light source are selected for lighting.The mechanical transmission system is designed through analysis and calculation,including the design and selection of grasping mechanism,overturning mechanism and turntable mechanism.The pneumatic circuit of the electronic control system is designed and the I/O ports of Mitsubishi FX3 U PLC are allocated.The assembly of each part of the design is carried out and the platform of the detection system is built.(3)Research and application on all surface defect detection algorithm.The detection algorithm relies on OpenCV vision development package.By comparing the performance of various gray scale transformation and filtering pretreatment methods,exponential transformation and bilateral filtering are selected as the main algorithm in the pretreatment process.For the upper and lower end faces of the ring,Hough transform is compared with the fitting outer circle of least square method,and the latter is selected to construct the end face mask to extract ROI region(interested region).Then the end face defects are identified.For the inner and outer sides of the ring,the subregion of the side is segmenting based on coordinate search to locate the ROI region more accurately.A mask is constructed by contour search method to extract the ROI region.Defects on the inner side are identified based on the whole ROI and defects on the outer side are identified based on the subdivision of the ROI region.The detection software application is designed on LabVIEW platform based on the detection algorithm.The humanized detection interface is set to display the detection status in real time.The image acquisition module,image processing module,PLC communication module and other submodules involved in the overall software framework are designed.(4)Experimental research on detection system.The function test of the testing system is carried out,including the mechanical and electrical function test and the function test of the testing software.The results show that the detection system can meet all the design requirements and some technical indexes.Then the defect identification experiment is carried out through the detection system.The important parameters of the detection program including BT,γ and S are determined by experiments on the software platform,by which the performance of the algorithm is initially verified.A new detection strategy is designed to meet the requirements and the best parameter combination is selected through experiments.The comprehensive accuracy of the detection system is greater than 95% and both the false detection rate of good products and the missed detection rate of defective products could be controlled within 3.5%,which further verified that the designed detection system could meet the design requirements.There are 71 figures,18 tables and 95 references in this paper.