Graph Computing Based Electric Power Equipment Defect Grading with Multi-scale Mechanism

Power equipment inspection aims to discover and eliminate equipment defects and potential safety hazards timely and ensure the safe operation of equipment. Now a large number of graph structure data have emerged in the equipment inspection field to improve the organization, management, and cognitive ability of knowledge. As a key step in the inspection business, equipment defect grading is used to determine the severity that affects the safe operation of equipment, so the accuracy and generalization ability of the model are required to be high. The pooling operation of traditional graph neural networks has the problem of information loss. Therefore, this paper introduces the multi-scale mechanism to improve the accuracy and generalization ability of the model and realize the accurate grading of equipment defects. By conducting experiments on NCI-1 and CEPRI_EQUIP datasets and comparing the proposed algorithm with baseline methods, the significant performance of the proposed algorithm is verified.