Eddy Current Pulsed Thermography Nondestructive Testing Method Based On Low Rank Total Variation Regularized Tensor Decomposition

With modern industry is in constant development,metal materials are widely used in the manufacturing of various production equipment,but some equipment runs in a variety of harsh environments for a long time,and the surface or internal parts of the equipment are prone to corrosion,cracks and other defects.In serious cases,it will lead to equipment failure,resulting in economic losses and even security accidents.Eddy current pulsed thermography is widely used as a nondestructive testing technology.In its application,an important step is to extract the characteristic information of defects.In the existing research,some feature extraction algorithms and image processing algorithms are used in the field of nondestructive testing.However,the defect information is susceptible to strong background noise and artifact interference caused by heat conduction.The existing methods have not fully solved the above problems in the actual testing.In this paper,a tensor decomposition model and its algorithm based on low-rank total variational regularization are proposed to process and analyze the data of eddy current pulsed thermography,thus realizing effective identification and location of defects.The research work in this paper is mainly as follows:(1)The eddy current pulsed thermography system was established to conduct thermal imaging detection for different types of test samples and collect thermal imaging data.The distribution characteristics of the collected data in time dimension and space dimension are analyzed,and the tensor theory is used to explain the data distribution.(2)Based on the low-rank and spatial continuity characteristics of thermal imaging data,the low-rank tensor decomposition model and three-dimensional total variational tensor decomposition model are studied,and the two tensor decomposition models are matched with the characteristics of thermal imaging data.On this basis,the tensor decomposition model and algorithm based on low-rank total variational regularization are proposed.In this algorithm,tensor kernel norm is used to constrain the low-rank background part of thermal imaging data,and three-dimensional total variational norm is used to constrain the artifact components formed by heat conduction in high temperature regions in the data,so as to realize more accurate defect feature recognition and extraction.(3)By conducting experiments on a variety of different types of test samples and using the method proposed in this paper and other existing tensor decomposition methods to process thermal imaging data,different methods are compared from the visual detection effect and quantitative evaluation index,and the advantages and limitations of each method are analyzed.The final results show that compared with other defect detection methods,the proposed method has certain advantages in visual detection effect and quantitative evaluation index performance.Specifically,the proposed method preserves the shape and contour of defect information more completely,and the recognition rate of defects and the signal-to-noise ratio of defects are higher than other data processing methods of the same type.