A methodology for track geometry estimation using inertial measurements: compensation of bogie filtering

In this paper, a methodology for the estimation of track longitudinal level based on the double integration of acceleration measurements is presented. The method is meant to be implemented on vehicles in-service along main line, where trains run at various speed depending on the line characteristics. The system relies on the information coming from a single sensor, mounted at the centre of the bogie frame. A suitable strategy to account for the filtering action introduced by the bogie, that makes the system blind to specific wavelengths, is proposed. To monitor the track conditions, the standard deviation and the peak value of the longitudinal level computed over a 100 m windows are adopted as synthetic indicators. Linear regression models relating the indicators sampled from the direct measurements taken by a diagnostic vehicle and those computed by double integration of the acceleration signals are realized. Both regression models, fed with the data gathered during a long-term monitoring, are finally adopted to estimate the track longitudinal level from the instrumented commercial vehicle, showing satisfactory results.