Compressed Sensing Based Distributed Sampling Method for Vibration Signal from Train Rotating Parts

Abstract

Sensor network based on-line fault diagnosis for rotating parts of the train bogies is a significant technology that is making a great impact on rail transit operation. But the signal sampling method conforming to Nyquist theorem can cause great trouble because it will produce too large amount of data to be transmitted and stored via the poor bandwidth sensor network. Aiming to this, this paper proposes a novel distributed sampling method of vibration signal from train rotating parts based on compressed sensing theory. The system consists of two parts: analog-to-information converter based signals sample and orthogonal matching pursuit algorithm based signals reconstruction. The experiment results show that the proposed method can recover signal over 0.9 correlation degree while the compression ratio is over 5. This mean that the method proposed in this paper will great reduce the transmission bandwidth and storage. At the same time, the compressive signal also contains all the failure feature information as the original signal. The experiments also show that the proposed method is better than traditional method in the robustness to noise and packet loss.