Analysis of rail defects signals by the matlab wavelet toolbox programme

Nichoga V.*, Vashchyshyn L.**, Saldan O.*

*Lviv Polytechnic National University
**Karpenko Physico-Mechanical Institute of the NAS of Ukraine

Rail networks across the world are getting busier with trains travelling at higher speeds and carrying more passengers and heavier axle loads than ever before. The combination of these factors has put considerable pressure on the existing infrastructure, leading to increased demands in inspection and maintenance of rail assets [1]. Nowadays, rails are systematically inspected for internal and surface defects using various non-destructive evaluation (NDE) techniques. The most common of which are ultrasonic and magnetic flux leakage (MFL) methods. The article is focused on the analysis of defectoscopic signals received using the magnetic wagon-defectoscope of Lviv Railway (MFL method) by the continuous wavelet transform (CWT).
The most important question in all methods NDE - is selection of information about defects from defectoscopic signals received during the checking railway (defectogram). At present, wagon-defectoscope operator’s experience is essential for making the correct decision about technical condition of rails. To improve the operator’s work efficiency, which is mainly based on expert visual assessments, it is necessary to automate the analysis of recorded signals. That is why we turned to the powerful tool of digital signal processing such as wavelet transform (WT). The main WT applications - analysis and processing signals, non-stationary in time (such as defectoscopic signals) when the analysis should include not only the  signal frequency characteristics, but also information about some local coordinates, which reveal themselves in one or other frequency components.
WT divides into continuous (CWT) and discrete (DWT). DWT focuses on speed (by sampling values of scale (frequency) and time (samples)) and the possibility of a complete reconstruction of the signal after analysis (by orthogonal wavelet functions), which led to its use, mostly for denoising and compression of signals. As for CWT, it requires only one wavelet function - mother, whose creation is not problematical (unlike DWT) and allows to identify those signals that are similar to the mother wavelet.