Design and Verification of Simulation Data Generator for Rail Transit Switch Machine

Abstract

Because the fault data of rail transit switch machine are difficult to obtain and the site fault is difficult to reproduce, it is difficult to diagnose or predict the switch machine. In this paper, the power fault data of S700K switch machine is divided into creeping fault and mutation fault, and a simulation data generator for generating massive fault data is developed. The generators involve the synthesis of minority over-sampling techniques and generative confrontation networks. Finally, the long-term memory neural network is used to predict the generated gradual fault data to verify the authenticity and reliability of the simulation data generator. The experimental results show that the generated data can predict the future power trend of the switch machine, which proves the authenticity and feasibility of the simulation data generator.