Novel Near-Field Microwave Testing Method for Outdoor Insulators

Abstract

Outdoor insulators are very vital in providing energized overhead lines mechanical support and electrical insulation from grounded structures. Insulators can be divided according to the material they are made of into glass, porcelain and polymer (Non Ceramic) insulators. After the 1970's, Non Ceramic Insulators (NCI) became very popular because of their ease of manufacturing, good contamination performance (due to hydrophopecity) and low installation cost. When manufactured or stored, insulators may be subjected to defects such as air voids, cracks in the fiberglass reinforced polymer (FRP) rod and contamination inclusion in the silicone rubber matrix. So, testing the insulator before the installation process and ensuring that it is defect free is very important to provide safety for workers and to avoid power outages. Unlike porcelain and glass insulators, it is hard to detect defects in NCI due to the absence of intermediate electrodes and the close distance between the sheds. In a previous study, a new technique was proposed that depends on near-field Microwave non destructive testing (NDT) using open ended rectangular waveguide sensor. In this technique, the NCI was considered as a multi layer structure where any defect in these layers will affect the reflection coefficient value measured by the sensor. The microwave NDT technique showed good potential in detecting different types of defects in the NCI. One of the short coming of the previous work is the lack of the ability of the system in identifying the type of defect inside the NCI sample. In this work, different defects in both silicone rubber and fibre glass core were detected. Artificial neural networks (ANNs) were applied in order to detect and specify the type of defect in the outdoor insulator. Moreover, stepwise regression was used in extracting the important features, where the features inspected in this study are Fast Fourier Transform (FFT) components, statistical features, and the details energy of the wavelet decomposition. The detection rate of defects in the silicone rubber layer was found to be 92.5%. Moreover, the type of defect in the silicone rubber matrix was successfully specified (air void or metal) where a classification rate of 95% was reached. In the case of fiberglass core defects, the detection rate was 90% and the classification rate between air void in the fiberglass core and cracked core samples was 97.5 %.