Research On Defect Detection And Wall Thickness Measurement Of Zirconium Tube Groove Based On Machine Vision

Zirconium tube is one of the important components of nuclear fuel element in heavy water reactor,its function is to load nuclear fuel.There are plugs at both ends of zirconium tube to seal the nuclear fuel,therefore,there is a 120° groove at both ends of zirconium tube to ensure the reliability of welding.Because the zirconium alloy is soft,the groove will inevitably suffer damage or pollution in the process of processing and transportation,resulting in defects,which will seriously affect the safety of nuclear reaction.With the continuous development of the nuclear power industry,the demand for zirconium tubes is increasing year by year,and the production quality requirements are also increasing.The efficiency and accuracy of the traditional manual sampling inspection have been unable to meet the demand.In recent years,machine vision inspection technology is frequently used in the field of defect detection and size measurement because of its high efficiency,accuracy and non-contact,which is suitable for the zirconium tube vision inspection in this paper.Based on this,the machine vision method is used to study the defect detection of zirconium tube groove and the size measurement of zirconium tube wall thickness.The subject of this paper is selected from the major science and technology project of Changsha(approval number: kq1804005)and Key projects of international scientific and technological innovation cooperation with the government(approval number: 2017YFE128400).Firstly,The paper focuses on the imaging scheme design of zirconium tube inspection,and then according to the technical requirements,the paper designs the zirconium tube wall thickness measurement algorithm and zirconium tube groove defect algorithm.The efficiency and accuracy of the proposed algorithm are verified by experiments at last.The main contents of this paper are as follows.1)Imaging system design: According to the causes and attachment position of many defects in the zirconium tube groove,the lighting mode based on the compound light source of coaxial light and annular light is designed.By adjusting the brightness of the two light sources,two kinds of zirconium tube end images are taken respectively.The top image is used to detect the dirt and scratches on the groove,and the chamfer image is used to detect the foreign body on the groove and to measure the wall thickness.In order to determine the corresponding relationship between the image and the actual size,the imaging system model is analyzed and the imagingsystem is calibrated by the imaging system calibration method.2)Algorithm for measuring the wall thickness of zirconium tube design:According to the appearance of the inner and outer wall edge of zirconium tube in the chamfer image,a variety of edge extraction algorithms are used to extract the inner and outer wall edge.On the basis of the sub-pixel edge extraction algorithm based on quadratic curve fitting,an improved gradient calculation method is proposed according to the edge shape of zirconium tube in this paper.The circle fitting algorithm based on the least square method is used to fit the extracted sub-pixel edge points to obtain the inner and outer diameter size of the zirconium tube and then calculate the wall thickness.3)Algorithm for groove defect detection design: According to the characteristics of the top image and chamfer image,a ring unfolding algorithm based on bilinear interpolation is proposed,which expands the circular zirconium tube into a rectangular image in order to increase the proportion of ROI and reduce the amount of calculation.Based on the similarity between the columns of the unfolded image,a defect detection algorithm based on "Average Template" is proposed.The algorithm compares the similarity of the gray values of column pixels and then and replaces the suspected defects pixels based on threshold setting,to design the average template of the defect-free state of the original groove image.The algorithm locates the suspected defects through the difference between the template and the original image.Finally,according to the actual criterion rules,the recognition of different kinds of defects in the top image and chamfer image is detected.