Feature Extraction of Point Rail of High-Speed Railway Turnout Through Oscillatory Behavior Based Signal Decomposition

Turnouts, due to their special structure, are prone to failure. A valuable way to monitor their health conditions is by analyzing wheel-rail force signals, as they are rich in information about the health conditions of turnouts. The key step is to accurately position the force signal segments generated by the turnout, especially taking into account the difference between the location of turnout-generated force signal in the measured signal and the turnout location recorded in documents. Considering that the wheel-rail force impact caused by the point rail is often more obvious, this paper, thus, proposes a point rail generated feature extraction method based on oscillatory behavior based signal decomposition (OBSD). The OBSD uses Morphological Component Analysis (MCA) and the Tunable Q-factor Wavelet Transform (TQWT) to transform the analyzed signal into distinct low and high oscillatory wavelets. Since the choice of parameters involved in OBSD affects the results of the decomposition, a systematic discussion of the parameter selection for wheel-rail force signal processing is presented, and the convergence index (CI) for the parameter selection is given. The position of the turnout can be determined by identifying the features of the wheel-rail force signal caused by the point rails. The efficiency of this method was confirmed by actual measurements of wheel-rail force signals.