Insulation failure localization of running rails based on improved multi-scale convolution intra-class adaptive model in urban rail transit

Abstract As the return path of traction current in DC traction power supply system (DC TPSS), running rail maintains certain insulation with the ground. Due to the environment and prolonged operation, insulation failure frequently occurs in the running rails. The amplitude of stray current (SC) and rail potential (RP) in DC TPSS will significantly increase with insulation failure, affecting the reliable operation of both DC TPSS and trackside equipment. Firstly, a dynamic simulation platform for the RP of the return system with insulation failure is established in this paper. Secondly, to tackle the challenge of insulation failure localization under varying conditions, an improved multi-scale convolution intra-class adaptive model is proposed. The model includes data augmentation, an improved residual network, and a multi-scale feature extractor. Transfer learning is achieved by reducing the marginal distribution distance using maximum mean discrepancy. Finally, experiments are carried out using the proposed failure localization method on datasets. Experimental results demonstrate that the model achieves an average localization accuracy of 98.47% across multiple insulation failure localization tasks. The proposed model can accurately localize insulation failures under varying working conditions.