12. Methods to detect and predict rapid localized deterioration of track irregularity

  • An effective method has been developed to correct the position of track irregularity waveforms with an accuracy of ± 25cm along the railway track, by comparing frequently measured track irregularity waveforms.
  • By computing the difference between the corrected waveforms, rapid deterioration of local track irregularity can be detected. After this, a method to predict changes of ±1mm, up to fifteen days ahead, was developed.

The collapse or failure of structures beneath a track can lead to roadbed caving or tamping insufficiency during track maintenance, which in turn can cause rapidly deteriorating track irregularity requiring immediate attention (Fig. 1). In order to detect the signs of rapid deterioration, RTRI developed a small, lightweight track inspection device that is mounted on commercial vehicles for high-frequency track condition monitoring. However, the waveforms collected on a daily basis vary in position along the railway track by a few meters. Therefore, although differences in waveforms measured on different days can be compared, it is very difficult to automatically identify which locations are suffering from rapidly deteriorating irregularity (Fig. 2).

A method was therefore developed using characteristics of high-frequency track monitoring data to allow automatic detection of rapidly deteriorating track irregularity by position correction: while shifting the waveform phases the phase with the largest correlation coefficient between waveforms was found, making it possible to reduce positional shift to within ±25cm along the railway track. By computing the difference between the corrected waveforms, a method was developed to automatically extract locations with rapidly deteriorating track irregularity using discrepancies in measured waveforms from different measurement days (Fig. 3).

Next, a method was devised to predict future track irregularity in locations where rapidly deteriorating track irregularity was detected: after each new measurement Bayesian inference is applied to consecutively update predicted track irregularity. The accuracy of the method makes it possible to predict gaps of ±1mm up to fifteen days ahead. This allows for example, early detection of locations subject to rapidly deteriorating track irregularity, facilitating the planning of appropriate maintenance work (Fig. 4). This newly developed position correction method was also integrated as a new function in the track maintenance and management system ‘LABOCS'.

Fig.1 Example of localized rapid deterioration of track irregularity
Fig. 2 Example of position correction under previous method
Fig. 3 Cross correlation method used to correct position and extract locations subject to rapidly deterioration of track irregularity
Fig. 4 Example of consecutive update of prediction
PAGE TOP