Rail–wheel contact stress assessment using acoustic emission: a laboratory study of the effects of wheel flats

This article presents part of a wider laboratory study of the acoustic emission (AE) generated during rail/wheel interaction. The particular experiments reported here use simulated wheel defects with a view to developing methods of in situ wheel flat diagnosis using rail-mounted sensors. The analytical approach recognizes the similarity between rolling bearing defects 11 and wheel defects and exploits the principle of demodulated frequency resonance 23 to recognize specific defect frequencies associated with the number, intensity, and disposition of the wheel flats. A set of experiments were carried out on a scaled test rig consisting of a single wheel running on a circular track, during which AE was recorded at a fixed point on the track. The wheel had a set of three asymmetrically distributed flats machined onto it and the rolling speed and axle load were varied. A previously developed time-based analytical model 17 validated against control experiments with an undamaged wheel, was used to reconstruct the normal pattern for a wheel running round the track and to provide a means of removing the effect of attenuation as the defect pattern moves away from and back towards the stationary sensor. Because the wheel flats contact the track at a known frequency, associated with wheel rotational speed, they introduce a pulsatile aspect to the signal, although at a frequency much lower than the frequency of the carrier (AE) wave. Accordingly, once the signal had been corrected for attenuation, it was enveloped to a suitable frequency range, and a frequency-based pulse-train model was used to devise an approach to matching the measured spectra to the expected pulse train spectra in a way similar to the concept of defect frequencies in bearings 24 . For all the enveloped time-series over the full range of conditions, the pattern of flats was clearly visible and reasonably reproducible between consecutive rotations, but with sufficient variation to warrant a spectral approach. The pattern of harmonics was, as expected, dependent on the number, intensity (size), and circumferential disposition of the wheel flats, as well as the axle load, while the rotational speed of the wheel controlled the fundamental frequency. Some frequencies lower than the fundamental were found and these were associated with track rotational frequencies and unwanted frequencies due to normal wheel rolling. These lower frequencies were removed using a high-pass filter 22 to focus on wheel flat diagnosis. It is concluded that, with appropriate calibration and modification, this approach could be used for the diagnosis of wheel defects (as distinct from track defects) in real railway wheels using either track-mounted or wheel-mounted sensors.