Real-Time Rail Fault Diagnosis Based on Vibration Signal Analysis and Second-Order Sinusoidal Model

Abstract

To ensure the reliability of rail transit, it is necessary to diagnose and monitor rail faults. Although the first-order sinusoidal signal model can be used for rail diagnosis, its accuracy is too low. This paper proposes a second-order sinusoidal model to solve this problem. First, the parameters of the second-order sinusoidal model are optimized to approximate the average signal via the least-squares batch learning. Next, with the rail vibration signal model based on the second-order sinusoidal signal model, information related to the rail average signals, which includes the amplitude modulations and the phase modulations, is extracted and analyzed, and the process of rail crack generation is determined. The second-order sinusoidal model extracts the rail characteristics of the amplitude modulation and the phase modulation, reflects the rail fault information and monitors the rail breaking process. Finally, with the experiment (Fig. 0: on the right) and actual rail data, rail fault diagnosis are demonstrated, which are beneficial for the safety of rail transit.