Research On Identification Of Internal Defects In Graphite Electrodes Based On X-ray Radiographic Inspection

Graphite electrodes are widely used in electric furnace steelmaking and electric discharge machining due to their excellent electrical and thermal conductivity,processability and high temperature resistance.The service life of the electrodes directly affects the operation of the overall production process.Detection is an effective means to control the quality of the electrode,but the characteristics of the graphite electrode make it more difficult to detect the internal quality of the electrode.Currently,the traditional percussion method and cutting section method are extensively used in graphite electrodes detection,but they are incomplete.The X-ray inspection method can intuitively display the size and shape of the internal defects of the material,and easily determine the defect.It has been widely used in material quality inspection.However,the traditional manual evaluation of radiographic film has the disadvantages of large subjective influence and heavy workload.Therefore,X-ray inspection method combined with image processing and automatic defect recognition was proposed,based on the characteristics of graphite electrodes and their internal defects.The main contents and results are as follows:Firstly,graphite electrodes exposure curve was producted.According to the parameters of the self-made graphite electrode exposure curve,a X-ray inspection process was developed for pre-made pores,cracks,and graphite electrode blocks with inclusion defects.The qualified graphite electrode ray negatives were converted into digital images.Then,electrode ray images processing method was studied based on the features of graphite electrode ray images and the applicability of various image preprocessing algorithms.After comparing various image denoising and intensification methods,the median filtering and Laplacian sharpening operators were selected to preprocess electrode ray images,which have improved the overall quality of the electrode image.Feature extraction of electrode defects was performed.Canny's algorithm was selected for the segmentation of electrode defects after comparing the detection results of different edge operators.To increase the accuracy of edge extraction,the traditional Canny's algorithm was improved.The feature is the basis for defect identification.After analyzing the image characteristics and image characteristics of the negative film of the electrode defect,a set of characteristic parameters that accurately reflect the nature of the electrode defect was determined.Featureparameter algorithms were estimated on basis of defect marking,tracking and filling,and the calculated feature values constitutes a feature parameter were provided to the classifier as the basis for defect identification.Unsupervised training and supervised fine-tuning were combined in the deep belief network to automatically identify electrode defects.The influence of different factors on the model recognition rate was analyzed.The parameter setting and sample allocation were optimized,obtaining a higher accuracy rate of defect recognition.The results show that through the production of graphite electrode exposure curves,reasonable exposure specifications and processes can be selected to obtain qualified inspection films.Besides,through computer image processing,electrode defect characteristics can be accurately extracted from the image as a basis for defect identification.In addition,deep belief network is reliable for automatic recognition electrode defects,with the highest recognition rate reaching 99.33%,showing good results in accuracy and stability,and finally provide theoretical reference for the computer automatic detection of graphite electrode defects.