Research On Intelligent Positioning And Detection Of Body-in-white Welding Spots Based On Machine Vision

Spot welding is the most important connection mode in the welding process of body-in-white.Its quality directly affects the quality and even safety performance of the vehicle.At present,the main welding spot detection method of the domestic main machinery factory is manual extraction by the portable ultrasonic flaw detector.This method has many problems,such as low detection efficiency,low coverage,and easily influenced by human factors,which cannot meet the increasingly stringent demands of market and consumers for automobile quality.In view of the above problems,it is urgent to develop a set of body-in-white welding spot on-line detection equipment with high reliability and high detection efficiency.The automatic positioning and detection of the welding spots can be carried out to realize the coverage of all the key welding spots 100%,so as to ensure the vehicle quality and improve the automatic and intelligent level of automobile production and manufacture.This paper mainly studies the key technologies in the development of intelligent detection equipment for body-in-white.This project is funded by the Science and Technology Department of Guangxi Province,and is jointly tackled by Hunan University and SAIC-GM-Wuling Automotive Co.,Ltd.The main contents are as follows:Firstly,the factors affecting the positioning during on-line detection of white-in-body welding spots are analyzed,and the necessity of using visual servo control theory is expounded.In view of the nonlinear and strong coupling characteristics of the robot vision system,it is impossible to compensate the positioning error by using the traditional modeling method.Therefore,the machine vision is combined with the depth learning algorithm,and the Support Vector Machine(SVM)regression theory is used to compensate the visual error.The Particle Swarm Optimization(PSO)algorithm is used to optimize the parameters of the SVM model,so as to further improve the positioning accuracy of the robot vision system and provide theoretical support for the precise positioning method of solder joints.Secondly,the visual positioning method of industrial robots is studied.Aiming at the problem that the positioning precision of traditional teaching playback robot cannot meet the demand of ultrasonic probe,a welding spot positioning method based on machine vision and SVM regression is proposed.First,the control strategy of machine vision positioning is established based on the actual conditions of welding spot on-line detection,and the initial positioning of the welding spot is carried out by measuring the deviation between teaching position and expected position of the robot by the machine vision system that combines laser ranging information.Second,a SVM regression model based on PSO is established to compensate for the vision positioning error.In the end,different optimized SVM parameter methods and commonly used error compensation methods are compared through experiments.It is proved that the SVM error compensation model based on PSO has better effect on the prediction accuracy and response time,which can further improve the positioning accuracy of machine vision system and achieve the accurate positioning of welding spots.Then,research on image processing algorithms for core steps of machine vision positioning.First,in order to make the welding spot clear and improve the image quality,the design of the imaging system and the preprocessing of the image ROI extraction,dryness and enhancement are carried out respectively.Then the effects of different traditional edge detection operators on welding spot processing is compared and analyzed,and the Canny edge operator with better anti-interference ability and better processing effect is selected.In the end,the center coordinates of welding spots are extracted by Hough circle transform,which provides technical support for the correct guidance of the teaching robot to correct deviations,and also provides algorithm support for the development of visual positioning software systems.Finally,a machine vision intelligent positioning experiment platform is built to verify the positioning method proposed in this paper.First,the design and selection of the hardware system and the development flow and main functions of the softw are system are described.Then the white-in-body welding spot visual positioning experiment platform is set up,and the experimental steps are designed in detail.In the end,according to the national metrological technical specifications on the evaluation of the characteristics of the instrument,the accuracy of the system positioning is verified.At the same time,the efficiency of the online detection and the consistency of the positioning are tested.The experimental results show that this method can meet the positioning accuracy requirements of ultrasonic probes,which can be used for on-line detection of white-in-body welding spot quality.