An Adaptive Wavelet Library to Detect Surface Defects in Rail Tracks Using a Laser Ultrasonic System

Abstract

This study is concerned with locating surface defects that occur in rail tracks. Ultrasonic Rayleigh waves were used to investigate the rail track surface. To generate and sense these waves a fully non-contact laser ultrasonic transduction system was employed. The laser-generated signals are in general more susceptible to environmental noise in comparison with signals generated by other methods. Meanwhile, the quality of signals received from one point may vary in each time of measurement. Continues Wavelet Transform (CWT) is a practical tool in dealing with complicated signals. In this regard, CWT works better if its mother wavelet is carefully selected based on the nature of the analyzing signal. Seeing that, a library of mother wavelets was tailor-made for studying laser-based Rayleigh waves in rail tracks. Mother wavelets were designed based on characteristics of the incident wave packets after extensive measurements on rail tracks. For analyzing a signal, initially, the first biggest wave packet that is the incident wave is recognized. Absolute cross-correlation is then used to pick a mother wavelet from the library that has the maximum resemblance with the incident wave. Using such an approach, the irrelevant wave packets can be easily discarded, and surface defects are exposed.