The Research On Fault Identification Method For Basin Insulators Based On X Ray

With the advantages of small maintenance,good equipment integration and high reliability of power supply,Gas Insulated Metal-enclosed Switchgear(GIS)have become a wide range of large switching equipment used in domestic 110 kV and above voltage grade substations.The basin insulator is an important part of GIS supporting the busbar and the isolation chamber,which is related to the safety of the whole equipment.The basin insulator is a composite material formed by the combination of epoxy resin and alumina.The temperature and the curing of the process will cause the inconsistency of the aggregation and differentiation of the two materials,which makes the internal stress difference obvious.If high pressure is added,cracks or bubble defects will be formed;at the same time,in the process of installation and use,partial metal fall off or equipment aging may also form a foreign body defect.The installation and disassembly procedure of the complete set of GIS is complex.How to efficiently and accurately detect the running state of its internal equipment without disassembly of GIS is the current research hotspot.The basin insulator is one of the key internal equipments of GIS.It is one of the key points and difficult points to install between GIS air chambers.In this paper,after a lot of analysis and research on the defect detection method of the basin type insulators,a method for detecting the defects of the basin type insulators based on the combination of X ray and improved convolution neural network(CNN)is proposed.In this paper,the X Ray Digital Radiography(X-DR)detection technology is used to test the basin insulator for several times,and a large number of basin insulator ray images are collected.Considering the characteristics of low contrast and many kinds of noise,a method of X-ray basin-type insulator image denoising based on improved block matching 3D is proposed.To overcome the pseudo Gibbs phenomenon resulted from the wavelet hard threshold algorithm used in collaborative filtering of block-matching 3D method and get more image details,this paper presents an improved wavelet threshold denoising method.An improved Kalman filtering method based on anisotropic diffusion is proposed to overcome the ringing effect of block-matching 3D filtering.It achieves clear edges and more details.After pretreatment,the image quality has been greatly improved,it provides a good image basis for subsequent fault identification,but it is still not easy to identifythe types of defects such as cracks,bubbles and metal foreign objects in the basin insulators.Although the crack defect has a certain length,its depth is shallow and the diameter is small,which caused the image contrast not obvious.The area of the bubble and metal foreign body is small,and it is easy to be blocked by other objects,so it is difficult to recognize.In order to extract different types of features for different types of defects,the workload is huge and the steps are tedious.But the convolution neural network can learn the feature of the image automatically and save a lot of feature extraction time,and the accuracy is more advantageous than other methods.In this paper,a defect recognition method based on improved CNN is proposed.The fixed convolution kernel of the traditional CNN is replaced by a deformable convolution kernel.The deformable convolution kernel can detect more regions of interest and make the identified features more typical.The traditional activation function linear correction unit(ReLU)is replaced by parameteric ReLU(PReLU),which can not only improve the silent phenomenon,but also speed up the convergence speed.Simulation results show that compared with support vector machine(SVM)and traditional CNN,the improved CNN has less computation cost,faster convergence rate and more accurate identification of basin insulator defects.The experimental results show that the fault detection method based on the combination of X ray and improved CNN can be used to detect the internal equipment without disassembly and contact with the GIS equipment.It is a safe,efficient and accurate method for detecting the defects of the basin insulators.