Research On The Classification Of Weld Surface Defect In Header Pipe Joint Using Machine Vision

As header is the main part of the tube boiler,the role of the header is to guarantee the evenness of mixing and the working of heating boiler.The weld between the cylinder and the pipe works with the pressure of the internal medium and the stress caused by temperature difference,is the weak link of the whole boiler.In manufacture,the short pipe joints are mostly made by automatic.Due to the instability factors of the manufacture,the weld surface would inevitably appear defects,such as incompletely filled groove,burn through,welding misalignment,overlap and so on.These defects can easily create the malignant accident such as boiler shutdown.In order to ensure the safe operation of the boiler,this paper has carried out the research on the welding inspection technology.The research content mainly includes(1)The visual automatic detection system of welding in header pipe joint.According to the requirements of the weld detection,we set up the welding inspection system based on endoscope and CCD camera.In order to realize the automation,the mechanical transmission and motion control module are designed.And the software of welding inspection based on LabVIEW software was developed.On the basis of study,the automatic system is used for picture acquisition of the welding surface.(2)The research of the classification of welding surface defect.Analyzed the texture features of different weld defects,the method of characterization of weld surface defects based on image texture is studied,And Back-propagation artificial neural network method is used for defect classification.Researched the parameters of the factors of the input parameters and the neural network structure on the defect classification performance.We optimized the best neural network structure and input parameters.Based on above research,the optimized network is applied to the classification of welding surface defect in the header pipe joint.(3)The research of welding surface depth information extraction method based on binocular vision.The binocular vision technology has been studied.According to the situation,we use a single camera to acquire image to replace the stereo camera.Verified the measurement error of the binocular vision system by standard block.Based on above research,the binocular vision technique is applied to the depth extraction of welding surface defect in the header pipe joint.According to the characteristics of the weld image,the method of image preprocessing of the weld were studied.The above work made beneficial exploration for the automatic classification of weld surface defect in header pipe joint.