Noise Currents Measurement Method for Rail Vehicles Based on Broad Learning System

Abstract

In order to ensure the electromagnetic compatibility between train and trackside communication system, metro projects at home and abroad have strict limits for the noise current generated by train during operation. At present, current sensors and filters are commonly used in the test of noise current of metro vehicles. This method increases the number of installed devices, and the device have poor portability, and are prone to damage during handling. In order to improve test efficiency and reduce costs, this paper proposes a vehicle noise current measurement method based on the broad learning system (BLS). Firstly, the current data samples obtained by flexible current loop and data recorder are divided, and then Fourier transform (FFT) is performed on the divided samples. Finally, the BLS model extracts the frequency signal as the input to obtain the corresponding noise current value. The experimental results show that the average error of the noise current value obtained by the BLS model is around 4%, which meets the experimental requirements. Moreover, the testing accuracy of this method is significantly better than existing methods: weighting method, BP neural network, particle swarm optimization and artificial bee colony, the effectiveness of the proposed testing method has been verified.

Keywords

• rail transit
• electromagnetic compatibility
• vehicle noise current
• broad learning system
• fourier transform