Research On Detection And Classification Of Bearing Outer Ring Surface Defects

Bearings are extremely common in all walks of life,and they are important components of modern electromechanical equipment.In the bearing production process,the outer ring surface quality inspection still relies on manual inspection,subjective factors are strong and the efficiency is poor.Machine vision-based defect detection system has the advantages of high efficiency,high degree of automation,non-contact detection.It is very suitable for quality inspection of bearing outer ring surface.A set of bearing surface defect detection and classification system was designed.At the same time,feature selection and classification were studied.The workload and innovations of this thesis are described below:1)Designed for the hardware and software of the defect detection and classification system.Firstly,the structure of the 6204 bearing and the types of defects commonly found on the outer ring are analyzed.The goal of the system design is established.Then the whole system is designed according to this goal.Then,the hardware part of the system is designed.It is mainly divided into two parts: image acquisition subsystem and work piece transfer subsystem.The scheme is given for the selection of devices in each subsystem.Thirdly,the software algorithm part of the system is given the scheme,and image engineering is used as the algorithm framework.The algorithm of this paper is designed and explained in order.Finally,the algorithm is validated by classification experiments.2)The hardware part designed in this paper,namely the image acquisition subsystem and the work piece transfer subsystem,can realize automatic image acquisition.At the same time,the double pneumatic manipulator can simultaneously realize the loading and unloading of the bearing and improve the system operation efficiency.Innovations in terms of aspects.3)In the feature selection and classification of the existing bearing outer ring defect detection,the subjectivity is too strong when feature selection,and there is no relationship between feature selection and classification.A feature selection algorithm combining correlation coefficient discrimination,eigenvector combination,and wrapping method is proposed,which is referred to as CEW algorithm.The algorithm steps are as follows.Firstly,a feature set of the research object is established,and thecorrelation coefficient between the features is used as a criterion to eliminate strong correlation features.Then,the mixed scatter matrix between classes is used as the criterion,combined with the feature search strategy of increasing l minus r.Search strategy for rough selection of features can enhance the correlation between features and classes,and weaken the correlation between features.Finally,based on the performance of Support Vector Machine(SVM)classifiers,fine selection of features and selection of optimal feature combinations are selected.This is the final classification feature.