RAMS for track

An approach, which assist track geometry maintenance decision making by improving the quality of the measurement data and prediction of the degradation.

Goal

The main aim of the project is to develop an approach, which assist track geometry maintenance decision making by improving the quality of the measurement data and prediction of the degradation.

Project status and results

Maintaining the RAMS parameters of railway system within acceptable thresholds at lowest possible cost is of the utmost importance for the railway infrastructure managers. Track geometry deterioration has negative consequences on safety, availability, and travelling quality. To restore the track geometry parameters to their design values it is essential to employ effective maintenance strategy. Reliable and complete track geometry data are the foundation to implement an efficient and effective maintenance planning and scheduling. Traditionally, condition of track geometry have been evaluated using the information of aggregated quality indices like standard deviation of a track section. However, these indices cannot provide precise information on the location of single defects. In this regard, recently, many studies have devoted to predict the occurrence of single defects. Prediction of single defects’ occurrence is a significant prerequisite for predictive models and maintenance modelling. It should be noted that to predict the occurrence of the defects, both the position and the time of occurring defects are of utmost importance.

Generally, track geometry measurement data suffer from positional errors. Therefore, collected track geometry measurements in different inspection runs need to be pre-processed before it is used to model geometry degradation and implement a condition-based maintenance strategy. To position the accurate location of the measurement data, it needs to precisely estimate the distance travelled by the measurement car from a milepost. In reality, the accuracy of travelled distance estimations is affected by wear condition of the wheel, wheel slip and slide on the rail, environment condition, and the calibration of the wheelset and the optical encoders equipped on the wheelset. In addition, in collecting the track geometry measurements, the sampling positions are not equal in different inspection runs. These influential factors cause to shift, stretch or compress the measurement data with non-constant and random distances between any two successive sampling points. On the other hand, maintenance operation and track geometry degradation distorts the inspection measurement data. In this regard, accurate alignment of the inspection measurement can reduce the positional error and improve the quality of the measured data.

Therefore, there is a need to implement an alignment method that both align the different inspection measurement data, precisely, and maintain the original shape of the datasets. In addition, it is preferable that the alignment method run fast with less memory.

The aim of this project is to find an accurate and efficient method for the alignment of track geometry measurements to reduce their positional errors. In this regard, the application of five alignment methods i.e. Correlation Optimized Warping (COW), Recursive Alignment by Fast Fourier Transform (RAFFT), a combined method using RAFFT and COW, Dynamic Time Warping, and Cross Correlation Function (CCF) on the railway track geometry measurement data are evaluated. The first three methods divide the datasets into small segments and align the segments. The other two methods consider the whole dataset for alignment. Based on the achieved results, it seems COW, DTW, RAFFT, and combined method can align datasets when one of them is stretched or compressed with respect to the others. In addition, the results showed that CCF is usable when there is a fix shift between different datasets and is inefficient when one dataset is stretched or compressed with respect to the other. When the amplitude of the measured data of different inspection runs are changed, DTW changes the shape of the aligned dataset and attempts to mimic the amplitude of reference dataset, which is not desire here. Therefore, DTW have a high precision in aligning measurements at the cost of warping the aligned dataset. Furthermore, based on the results, COW is inefficient in finding the shifts at the start and end of the datasets. However, RAFFT has a precise alignment at most of the times without changing the shape of the aligned dataset, it has some sporadic warping in some sections of the aligned datasets. Generally, the results showed that the combined method is the most efficient method in aligning track geometry measurements.

Using these approaches, one can reduce the positional errors in track geometry measurement data and improve the accuracy of the inspection measurement data. Therefore, it will strengthen the analysis and prediction of the track geometry degradation.

Sponsor: Trafikverket/JVTC
Researchers: Mahdi Khosravi (PhD candidate), Alireza Ahmadi (PL)
Duration: 2019-2023