Research On Defect Detection Of Ferromagnetic Materials By Pulsed Eddy Current Thermal Imaging Based On Faster R-CNN

Ferromagnetic material parts are widely used in large-scale mechanical equipment,aerospace,pipeline transportation and other fields.In the long-term service process,due to the influence of operating environment,manufacturing process and use mode and other factors,ferromagnetic material parts are easy to produce different degrees of surface or internal damage,which affects the operation and service life of equipment,leaving hidden dangers,or even damage Causing industrial accidents.Therefore,it is very important to detect the ferromagnetic parts in time.In order to build a non-contact and non-invasive defect detection model,this paper uses the infrared thermal imaging technology in the nondestructive testing technology to establish the infrared thermal image defect detection test platform.Through studying the defect detection technology and method based on the infrared thermal image,the nondestructive testing analysis of the defects of components or equipment is realized.This paper is devoted to the establishment of the infrared thermal imaging test platform and the construction of the infrared thermal imaging defect detection model.The research is focused on the thermal image denoising,defect recognition and positioning,edge segmentation and other problems and difficulties in the infrared thermal imaging of ferromagnetic materials.The research contents are as follows:(1)An optimized K-SVD dictionary learning algorithm is used to deal with the noise in the infrared image of ferromagnetic materials.Firstly,the initial dictionary is fixed,and then the sparse coefficient is obtained by using the orthogonal matching pursuit algorithm;secondly,the initial dictionary is updated by using the sparse coefficient through the iterative algorithm,and the redundant dictionary adapted to the target signal is learned from it.Experimental results show that the method can effectively reduce the interference of noise in the image,which is conducive to the next step of defect recognition and location.(2)Aiming at the problems of slow detection speed of surface defects and too much image interference information in the infrared thermal imaging detection of ferromagnetic materials,an improved infrared image intelligent detection model of ferromagnetic materials based on Faster R-CNN was adopted to improve the image processing ability of Faster R-CNN.The improved Faster R-CNN uses VGG-16 network for transfer learning,fuses multiple layers of feature graphs in the network.And the selection scheme of anchor point box of RPN network is optimized.The experimental results show that the model can accurately detect the defects and determine the specific location of the defects,and the average detection accuracy of four kinds of defects with different lengths reaches 96%.(3)In order to solve the problem that the edge feature information of the identified defect is not prominent in the image,the edge detection method based on mathematical morphology is adopted to segment the edge of the identified defect.Based on the ratio of information entropy,this method uses operators in different directions to detect the edges of defect images,and then gives weights according to the useful information of each edge detection image,and then fuses the images,so that the edge contour of defect images can be segmented accurately.In this paper,infrared thermal image is used as a non-destructive testing method,and the infrared thermal image non-destructive testing platform is established to solve the problems of thermal image segmentation and defect detection positioning.The defect detection and analysis model of ferromagnetic materials is established,which can detect the defects of the equipment and provide guarantee for the safe operation of the equipment.