Multiple Sensors and Partial Calibration for On-Board Measurement of Rail Acoustic Roughness: Results of Rolling Tests

The indirect measurement of roughness using vibro-acoustic sensors on board rolling stock is promising, as shown by recent research on this topic. However, a number of improvements can be made to the existing methods to extend their range of validity. In this paper, the performance of the multi-sensor method developed in the framework of the French research project MEEQUAI is presented. The estimation of the effective roughness is performed for different sensor families using a least squares method in the frequency domain, from the signal spectra in third octave and modelled transfer functions corresponding to the detected track layout. A partial calibration method is also proposed. The calibration gain is relative and allows the variability of the transfers with the track support to be retained, thanks to the models. The method was tested in France with sensors installed on a test train running at several speeds on a track section of a railway test center as well as on two track sections of the national railway network with different supports.