Research on algorithm for detecting commutator surface defect based on YOLOv5
ZHANG Xiao-li
MA Yi-chen
CANG Yu-ping
DONG Shao-fei
HAO Na
HE Si-si
Abstract：During manufacturing and using mechanical parts,defects may occur on the surface of mechanical parts,during repeated use of components,their minor defects may expand or even cause damage to the com-ponents,leading to malfunction in the system to which they belong.This article takes typical industrial component commutators as the research object and proposes a detection method of component defect based on deep learning algorithm.In this research,based on KolektorSDD data set,firstly,Mosaic data enhancement method was used to rotate and crop the data of commutator defect data set,and then expand the data set and build the data set.Secondly,the constructed data set was divided into training set and testing set,and through YOLOv5 convolution neural networks training model,a commutator surface defect recognition model was established.Finally,the model is used to test the data in the testing set.The results show that the average ac-curacy rate(mAP)and positive sample recall rate(Recall)of the performance evaluation index of the training model are over95%,and the detection accuracy of the commutator surface defects based on YOLOv5 algorithm can reach to 90%.
Keywords：commutator surface defectdeep learningconvolutional neural networkYOLOv5 algorithm
Publication Date：2023-11-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 21-27 )