Bogie Speed Estimation and Signal Source Separation via Rail Vibration Analysis

As an important part of high speed trains, any fault or failure of the bogie may cause fatal disaster and/or expensive lost. Considering on-train condition monitoring system is still restricted by many limits, bogie condition monitoring via rail vibration may be very valuable. For the rail vibration signals exited by bogies are short termed, non-periodic, non-stationary, not pre-separated corresponding to wheels, and with time-shifting of phases, they must be separated corresponding to each wheel for fault location. In this paper, to separate and reconstruct the vibration signals from two bogie wheels, a measurement mixing model is established and some assumptions about bogie speed, measure accordance, etc. are given. Based on envelope correlation analysis of signals from multi-sensors, the bogie speed and measurement windows are estimated so that the rail vibration signals can be separated corresponding to each bogie wheel, which allows the contact dynamics of the wheel-track to be characterized accurately for the purpose of diagnostics. A simulation and an experiment have been applied to verify this method. Some result as well as discussion will be given.