Research On Bearing Surface Defect Detection System Based On Machine Vision

As a basic component widely used in mechanical equipment,the quality of bearing will seriously affect the stability of equipment operation.In recent years,with the vigorous development of China’s manufacturing industry,the demand for bearing products in various industries is increasing,and the requirements for bearing appearance quality are also increasing,especially in bearing import and export trade.Although China’s machining technology has achieved a high level,it is inevitable to produce some damage in the batch production of bearings.At present,enterprises still use the methods of human eye detection and manual sorting for the sorting of defective bearings.This method not only has the problems of heavy workload and low efficiency,but also has high missed detection rate and false detection rate due to different sorting standards and visual fatigue of workers.In view of the above problems,this paper designs a set of bearing surface defect detection system based on machine vision technology to replace manual visual sorting and realize the automatic detection of bearing surface defects.The main research contents and achievements of this paper are as follows:（1）In order to realize the full inspection of bearing surface defects,according to the defect types and distribution positions of bearing surface,a multi station combined light source lighting scheme is designed to complete the image acquisition work,and the selection of industrial camera,optical lens and visual light source and the design of lighting mode are carried out for each inspection station.（2）According to the system detection requirements,an automatic sorting system is designed and built.The hardware system is mainly composed of feeding device,feeding device,image acquisition device,turnover device,sorting device and stacking device,and the overall operation is controlled by PLC.（3）Corresponding detection algorithms are designed for different detection stations,including Edge Detection,EDCcircles Circle Detection,EDLines Line Detection,Region Segmentation and Extraction,Blob Analysis and other methods based on traditional image processing algorithms.In addition,this paper proposes a pit defect detection algorithm based on SFCS-YOLO v3 bearing dust cover surface,which achieves 97.50%detection accuracy,and a bearing image oil drop removal algorithm based on two-stage neural network（Attenitve GAN and AE-WGAN）is proposed,which can reduce the false detection rate of oil drops on the bearing surface by 93%.（4）This paper develops a set of bearing surface defect detection software based on QT creator platform.The software integrates the functions of image acquisition,image processing,result visualization,data statistics and parameter configuration.Finally,after testing,the detection accuracy of the bearing surface detection system designed in this paper is 95.80%,and the detection time of a single bearing is about 2.14 s.