Fault diagnosis of locomotive gearboxes based on adaptive fractional Fourier transform

To address difficulties in extracting transmission system fault information from operational locomotives on railways, this paper proposes a diagnostic method tailored to variable-speed conditions by analyzing current signals from traction motors for locomotive gearbox faults diagnosis. As traction motor speed changes linearly, the fault characteristic frequencies of the transmission system also vary linearly but at multiple rates, differing from the motor power supply frequency and gear rotational frequency. Although the fractional Fourier transform (FrFT) is well suited at analyzing linear frequency modulation signals, selecting its optimal order complicates its application. To overcome this, this paper proposes to construct a new hybrid entropy to uniformly characterize the spectral peak sharpness in FrFT and the degree of energy concentration in the frequency domain, and to use it as the fitness function to determine the optimal order through exhaustive search. Experiment results from industrial implementation show that the proposed method can effectively diagnose transmission gear faults in tractor motors.