Research on GIS Mechanical Defect Diagnosis Method Based on Gramian Angular Field and Lightweight Neural Network

Gas insulated switchgear (GIS) mechanical defects are the main cause of equipment failure, and accurate and efficient identification of defect types is crucial to ensure the safety of equipment operation. Aiming at the high complexity of GIS mechanical defect signals, which is difficult to select feature parameters and depends on expert experience for a long time, this paper proposed a GIS mechanical defect diagnosis method based on gramian angular field (GAF) and lightweight neural network. Firstly, the one-dimensional defect vibration signals were transformed into a two-dimensional spectrum using GAF, which contained a wealth of defective features that did not need to be extracted manually; then the GAF maps were inputted into the lightweight neural network SqueezeN et, and a network model for the classification and identification of GIS mechanical defects was constructed; finally, based on the real GIS mechanical vibration simulation platform that has been built, the vibration signals of three types of typical mechanical defects, namely poor contact of disconnector, loose bolt on the disconnector base and loose bolt on the bus bar conductor base, were collected and analyzed in terms of map conversion and diagnosis. The results show that the diagnostic method proposed in this paper can recognize mechanical defects with an accuracy of more than 95%, which provides a new way to accurately and reliably diagnose GIS mechanical defects.