Data-driven condition-based monitoring of high-speed railway bogies

In this paper, a method is proposed to monitor the running stability in a high-speed railway bogie and to detect and resolve different faults that may occur in bogie components critical to vehicle stability, particularly increased conicity caused by wheel wear and degradation of yaw dampers. The method is based on the analysis of the lateral accelerations of the bogie frame using the random decrement technique (RDT) to extract the free response of the bogie. The output of the RDT algorithm is then analysed using the Prony method to identify the characteristic exponents of the system, eventually allowing to define the stability margin of the bogie and the frequency of the hunting motion. These data are fed into fault classification algorithms, to obtain information on the condition of the yaw dampers and on wheel–rail conicity. The paper also presents the implementation of the method in a prototype condition-monitoring unit installed on an ETR 500 class high-speed train. Results from numerical experiments and from high-speed line tests are presented, showing the capability of the method to separate the case of a bogie with new wheel profiles from a condition with worn profiles.