RAILWAYS and infrastructure managers regularly measure track geometry to identify irregularities and inform maintenance work. This data is traditionally retrieved by dedicated track inspection vehicles, which typically conduct inspections at three-monthly intervals on conventional lines.
Recent developments in measurement devices for track inspection and their deployment on commercially-operated trains have increased the frequency of measurements. This in turn has improved the quantity and level of detail of the data available to infrastructure managers.
However, with more data available, using this information effectively has become a greater challenge. As a result, the team at Railway Technical Research Institute’s Track Technology Research Laboratory set out to identify ways to effectively use this data to the benefit of track maintenance.
Conventional track inspection cars measure three points on the rail and calculate longitudinal or alignment track irregularities based on the relationship between the position of these three points, as shown in Figure 1. On the other hand, the “inertia versine track measuring device” used by equipment installed on commercial trains and vehicles calculates an irregularity only by measuring one cross section based on the well-known relationship between irregularity and acceleration.
This method has resulted in a significant reduction in the size of the equipment (Figure 2), enabling installation on the bogies of a commercially-operated vehicle, and more recently on car bodies. The equipment in use includes rail displacement detection devices for rail location identification and a gyro used to detect the spatial location and attitude, or orientation, of the equipment and the accelerometer.
By using commercially-operated vehicles, infrastructure managers can retrieve inspection data every day making it possible to identify track irregularities quickly and accurately. They can also evaluate the margin of error in the predicted value as well as consider the impact of seasonal fluctuations on the variation of the measured irregularity.
Furthermore, it is possible to identify changes that may cause track irregularity earlier. By evaluating and comparing track irregularity more accurately both before and after maintenance work, it is possible to use this method to evaluate ballast conditions, leading to a more accurate prediction of how the ballast condition will be improved by maintenance work.
The prediction method, which is based on frequent measurements and uses large quantities of historical data, is capable of predicting future track irregularity as a probability distribution.
Figure 3 shows an image of the prediction. When the fluctuation in the historical data is small, the predicted values are limited to a small range. However, when the fluctuation is large the predicted values disperse across a larger range. Since the predicted value can be considered a probability distribution, it is possible to evaluate the reliability of the predicted value and to determine the probability that the irregularity value exceeds the upper limit. Therefore, maintenance works can be carried out in a more preventative manner by prioritising the spots which display a higher probability of exceeding the upper limit of irregularity.
It is also possible to show the impact of seasonal changes on underlying structural conditions. Figure 4 shows an example of fluctuation in the longitudinal level irregularity on a concrete bridge. This figure indicates the monthly average of the vertical irregularity. And while the amount is rather small, it fluctuates cyclically in one year while increasing and decreasing repeatedly. This indicates that track irregularities fluctuate seasonally at this spot.
Based on the understanding that seasonal irregularity change is not negligible, we have devised a prediction method in which the irregularity value is divided into components of seasonal and non-seasonal fluctuations.
Figure 4 also indicates the result obtained by applying this method to two years’ of historical data in order to predict the longitudinal level of irregularity for the next eight months. As Figure 4 shows, the predicted value and the actual measurement value are sufficiently close to each other to make it possible to predict any irregularity with high accuracy.
Ballast tamping is conducted during track maintenance to minimise track irregularities, and is often carried out at specific intervals. The prediction method makes it possible to evaluate the level of ballast degradation to a high degree of accuracy and to devise a cost-effective ballast replacement plan based on careful observations of time-series track irregularity data, thus eliminating unnecessary work.
Figure 5 shows how much improvement can be achieved in the period after maintenance work is conducted for lots A to C. Here the longitudinal level irregularities before the work are similar to each other, and for lots A and B, the amount of improvement is sustained even six months after the maintenance work takes place. However, for lot C, the amount of improvement gradually decreases. This demonstrates that the impact of the maintenance work is low, and that a deterioration in the ballast degradation is a potential cause of this effect.
These new features of data analysis for high-frequency inspection led us to revise our track irregularity maintenance planning system (MTS). We subsequently applied this new system to the data obtained from the high frequency inspections, comparing it with information retrieved from the traditional method and found that approximately 10% of the locations requiring further inspection were different from each other.
This is caused by the significant variation in the amount of data available for processing in both systems. It is expected that by using high-frequency inspection data effectively, the quality of the track irregularity maintenance plan will improve significantly compared with practices based on data obtained from conventional low-frequency inspections.
*This article was written by: Masashi Miwa, laboratory head, Track Technology Research Laboratory, Track Technology Research Division, and Eiji Yazawa, Hironori Sano, and Tsuyoshi Yamaguchi from the Track Technology Research Laboratory.