Research On Quality Evaluation Of Resistance Spot Welding Based On Image Processing Multi-feature Coupling

Resistance spot welding is a commonly used method of connecting plates in industrial production,and is widely used in automobile body manufacturing,transportation,aerospace and other fields.The nugget of spot welding is enclosed between the workpieces,and the internal conditions of the solder joint cannot be viewed during and after welding,which increases the difficulty of the quality evaluation of the solder joint.Traditional solder joint inspection methods mostly evaluate the quality of a single feature of solder joints,and it is difficult to accurately evaluate the quality of solder joints.Therefore,based on the solder joint image and the imprint image,this paper extracts multiple features for welding quality evaluation,establishes a multi-feature coupling model,and improves the accuracy of welding quality evaluation.First,the camera,lens and light source were selected according to the actual working conditions of the welding,the image acquisition system was built,and the system calibration was completed.The welding experiment is designed,the image of solder joint and imprint is collected,and the nugget diameter of solder joint sample is obtained by destructive experiment.The data set of solder joint-imprint-nugget diameter sample is established,and the median filter is determined as the image preprocessing method.Secondly,the surface feature distribution of the solder joint image is analyzed,and the surface texture characteristics of the solder joint such as HOG feature,gray level cooccurrence matrix feature and LBP feature of the solder joint surface image are extracted to evaluate the surface quality of the solder joint.Designed the imprint image processing flow,and optimized the imprint contour fitting algorithm,reducing the fitting error from 9.34% to2.40%.Correlation analysis of imprint characteristics and feature screening were carried out.Six characteristics of imprint aspect ratio,density,equivalent diameter,ellipse area,average axial length and pitting area were selected to predict the diameter of the solder nugget.Then,based on the characteristics of the solder joints,a solder joint surface quality model and a nugget diameter prediction model are constructed.The surface quality model uses two engines,SVM + HOG model and CNN model.The detection accuracy rate can reach97.51%.The nugget diameter prediction model uses SVR,the prediction error of the sample is between ±5%.model.The accuracy of the supplementary model for comprehensive quality evaluation of solder joints can reach 93.85%,and the error of the nugget diameter prediction is between ±6%,and most of the sample deviations are distributed between ±4%.Finally,based on the multi-feature coupling model,a welding quality inspection platform was built,and a visual interface was designed for it.In the welding laboratory,the welding quality inspection platform was tested.The platform has good performance in identifying surface defects of solder joints,and the error between the predicted result of the nugget diameter and the actual situation is within ±8%.