Research On Defect Detection Method Of Ferromagnetic Material Infrared Thermal Image Based On Cluster Analysis

Industrial equipment is affected by factors such as operating environment,manufacturing process,and usage methods.Its ferromagnetic components are susceptible to damage to the surface or inside to varying degrees,affecting equipment operation and service life,leaving hidden safety hazards,and even causing industrial accidents.For this reason,it becomes very important to carry out non-contact and non-intrusive detection of the ferromagnetic parts of the equipment in time.In order to build a non-contact,non-intrusive defect detection model,this article uses infrared thermal imaging technology in non-destructive testing technology to establish an infrared thermal image defect detection test platform to study the infrared thermal image defect detection technology and method to achieve the defect of components or equipment Analysis of non-destructive testing.The thesis focuses on the establishment of infrared thermal image detection test platform,the establishment of infrared thermal image defect detection model and other research content,focusing on the problems and difficulties of thermal image feature extraction,edge detection and classification model optimization in infrared thermal image of ferromagnetic components.The contents are as follows:(1)A GGCM(Gray Gradien Co-occurrence Matrix)texture feature extraction and crack defect recognition method for infrared thermal images is proposed.Extracting the characteristics of infrared thermal images is the key to establishing a classification model of ferromagnetic materials.Ferromagnetic materials are affected by the use method and forging process.The local temperature abnormality causes the infrared thermal image to be prone to local color unevenness,and it is difficult to extract image features.To this end,this paper uses GGCM to extract the 15 texture features of the infrared thermal image,and constructs the infrared thermal image feature set for establishing a defect detection model based on ELM(Extreme Learning Machine).Experiments show that the proposed model can effectively characterize the infrared thermal images of ferromagnetic materials and correctly identify defects in thermal images.(2)The infrared thermal image defect recognition model based on ELM isaffected by the random selection of ELM parameters,resulting in unstable model performance.Aiming at the above problems,an infrared thermal image crack defect recognition method based on ELM structure optimization is proposed.First,the feature set of the infrared thermal image of the ferromagnetic material is K-means,and the sample entropy is used to determine the optimal number of clusters,which is used to optimize the structure and parameters of ELM,that is,the optimal number of clusters,cluster centers,and cluster radius,Used to optimize the number of hidden layer nodes and activation function parameters of ELM.Experimental results prove that the method can effectively improve the recognition performance of the model.(3)Segmentation and edge detection of the target in the infrared thermal image have the problems of image mis-segmentation and image edge features being easily disturbed.A method of infrared thermal image segmentation and defect edge detection based on K-means is proposed.First,single-scale Retinex(single-scale Retinex,SSR)is used to enhance the infrared thermal image;then the standard deviation is used to determine the optimal number of K-means clusters,and the enhanced infrared thermal image is processed by K-means.Image segmentation;Corrosion and expansion in mathematical morphology are used to remove useless information from the segmented image,and Canny operators are used to detect defective edges in the infrared thermal image.Experimental results show that the proposed method can accurately segment defects in infrared thermal images and detect their edges,and has strong practicability and effectiveness.(4)Establishment of eddy current infrared thermal image nondestructive detection system.The platform uses eddy current equipment to heat the ferromagnetic components or equipment to obtain infrared thermal images,and uploads the infrared thermal images to the upper computer for analysis via the network.The system consists of hardware acquisition and host computer analysis software.The hardware acquisition part uses GYMCU90640 infrared dot matrix temperature measurement module as the data acquisition module.The STM32F103C8T6 chip is the control module and ESP8266 is the signal transmission module.The thermal image is subjected to denoising processing,feature extraction,classification and recognition,and finally the infrared thermal image of the ferromagnetic material parts orequipment can be detected remotely in real time.In this paper,infrared thermal image is used as a non-destructive detection method,which solves the problems of thermal image mis-segmentation and optimization of defect detection models.A non-destructive infrared thermal image detection platform is established to realize the defect detection and analysis of ferromagnetic specimens.