Vittorio Belotti, Francesco Crenna, Rinaldo C. Michelini, Giovanni B. Rossi
WAVELET SIGNAL PROCESSING APPLIED TO RAILWAY WHEELFLAT DETECTION
Aim of this research is the development of a reliable automated procedure for the detection of wheelflat faults in railway diagnostics. This kind of fault occurs when the train wheel slides on the rail. The wheelflat produced by sliding is a relevant cause of acoustic pollution, passenger discomfort, damage to rails and boogies. A good wheelflat detection diagnostics is common interest, both, for railway and vehicles operators.
The diagnostic method presented in this paper is based on the wavelet analysis of accelerometric signals obtained by an experimental campaign on a test train. The application field for wavelet analysis is new, and the results hitherto obtained show a better and simpler issues than classical time-frequency based analysis.
From a statistical standpoint, the experimental data are exhaustive base to achieve relevant diagnostic estimates in the train speed range between 10 and 100 km/h. The wavelet analysis processing method allowed to distinguish a wheelflat defected wheel up to 94% of the considered cases. Moreover, the same accelerometer signal giving wheelflat diagnostic is used to measure the train speed by a proper processing method leading results within a 2% by respect to commonly used commercial sensor.