Study Of Key Issues For Non-contact Detection Of Magnetic Flux Leakage Method And Identification For The Weld Defect

Welding is modern industrial base,and welding equipments easily fail in the course of fabricating and service.The quality of the weld plays a vital role on the equipment safety,and thus the periodic inspection of the weld is one of the important measures to ensure the safe operation of equipment.At present,it is usually using Radiographic Testing,Ultrasonic Testing and Magnetic Memory Testing et al.for on-line inspection of the weld at home and abroad.Compared with the above methods,the magnetic flux leakage(MFL)technology to detect defects in the weld,because the technology is more sensitive to the crack,stoma and other defects in the weld structures of the ferromagnetic material,and thus comparing with the other methods,MFL has incomparable advantage on the research costs and the accuracy of the analysis.MFL detection of weld defects has the technical feasibility,but there are two key technical problems.One is the structural characteristics of the weld,the magnetization direction parallel to the traveling direction of the detection system is no longer available.The other one is using MFL in the detection of the weld defect,MFL signal generated by the weld contour superimposing with the defect signal.There are some limitations that the two key technical problems lead to the traditional signal processing thinking of the weld defect detection and identification work.This article is to solve the practical engineering problems as the starting point,MFL detection system of butt weld has been developed,and the visual identification of weld defect has been achieved by the help of the image processing technology.According to butt weld MFL principle,combining the numerical analysis and experimental research method,a new magnetization system is designed whose magnetization direction is perpendicular to marching direction.The relationship between the characteristics of the weld defect,the weld structure and MFL spatial distribution are analyzed qualitatively and quantitatively.And by the gradation linear transformation,the vertical components of MFL data of two typical defects(rectangular slot and cylinder,one representative of the rectangular slot for the area defects crack and the other representative of the cylinder for the volume defects stoma and corrosion pit)are converted into the corresponding gray value to generate the grayscale image.The grayscale image can be visually shows the weld and defect information.Then by the MFL image of the weld defect as the example,statistical properties of MFL image of the weld defect are analyzed.On this basis,study MFL image of the weld defect from the main characteristics of the image such as chroma,texture and shape.Then by the gradation-color conversion method,the generated grayscale image is converted into the pseudo--color image to enhance the visual effect of weld defect characteristic information.Afterwards,using gray-gradient co-occurrence matrix extract the texture features of MFL image of the weld and two typical defects.Then select the characteristic quantity of weld image by the hierarchical clustering method,then analyze the selected the characteristic quantity using of the k-means clustering.And clustering results are showed as visualization graphics in order to achieve the classification and identification of typical weld defect.For the crack is the main defect affecting the mechanical properties of the weld,departure from the basic theory and algorithms of mathematical morphology,combined with MFL image features of different location crack of the weld.Through a combination of histogram equalization,the Otsu method binary,morphological remove small objects,and morphological edge detection operator,construct an adaptation for the edge detection algorithm of MFL image of the weld crack defect to extract the shape characteristics of the weld crack.According to the results of edge detection,comparing with traditional edge detection algorithm,the structure of the algorithm has the characteristics of high detection accuracy in the weld crack image edge detection,and the results achieve visual identification of different location crack of the weld.