Reliability of the implemented predictive analytics of the state of the railway track in JSC Russian Railways, directions for improving the predictive analysis

Abstract

The capacity of the Eastern polygon of roads is systematically increasing: in 2013 it amounted to 97,8 million tons and increased to 180 million tons in 2024, however the requirements of the consignors of goods force to increase the capacity limits. By 2032, the goal is to further increase the throughput capacity to the milestone of 270 million tons. The annual growth in freight traffic on the Eastern Polygon increases the cost of repairs and maintenance of the track infrastructure. It is becoming increasingly difficult to maintain the track infrastructure, since in order to implement a strategy to increase cargo flow, not only inter-train intervals are optimized, but also the time of technological «windows», therefore, it is unacceptable for managers of track maintenance to make mistakes when planning work, including those in prioritizing troubleshooting. But without understanding the actual condition of the track, as well as reliable predicted consequences, it is not possible to perform the task of switching to railway track maintenance based on the actual condition of the full range of predictive analysis. JSC «Tvema Company» commissioned by JSC «Russian Railways» has created an information and analytical system for diagnostics and monitoring of railway infrastructure, which is part of a Unified Corporate Infrastructure Management System. It has developed a section of predictive analysis of the geometric parameters of the railway track, but the results obtained after a few months in most cases have not been confirmed. The implemented predictive analysis technique does not yet provide accurate results, since it does not cover the full range of the three areas of analysis, and the multifactorial system of the degradation process of the track condition is also not taken into account. Predictive analysis is a complex process of analyzing both the current state of the track infrastructure and the expected consequences. Using only mathematical models for predicting malfunctions without using calculations of the interaction of rolling stock and railway track, it is impossible to obtain reliable data, since not all malfunctions have retrospective series of the degradation process, and there are also many situations where malfunctions are still absent, but will form in the near future due to the structural features of the track and operating conditions. In predictive analysis, it is important to consolidate logical situations related to the peculiarities of the influence of geometric parameters of the track on the resulting power overloads under the influence of rolling stock in order to further reduce the power factor and prolong the life of the railway track. The direction of improving predictive analysis is currently an urgent topic.