Improving control over operational characteristics of subway rolling stock by using retrospective passenger flow estimation

This study investigates the process that forms operational loads on the electric rolling stock within a city’s underground railroad system. The task addressed relates to the impossibility to directly measure passenger flows on separate sections of the network because most control systems only record passenger entry and exit events without tracking their full route. This creates analytical gaps and makes it impossible to accurately assess the actual load, which is critical for technical diagnostics as well as maintenance. To solve this problem, a method for estimating passenger flows has been devised, based on stochastic modelling of passenger movement on an oriented graph of the transport network. An approach to load estimation has been proposed, the distinctive feature of which is the rejection of determining a single most probable route for each passenger. Instead, the conditional "weight" of a passenger is probabilistically distributed among the set of all possible routes that could lead him/her from the potential entry station to the actual exit station within time constraints. Method verification by simulation on a conditional network showed high accuracy of the results (the average relative error did not exceed 0.5%). The distribution of errors is symmetrical, close to normal, and concentrated around zero, which indicates the absence of systematic deviations. The accuracy is attributed to the fact that the probability distribution makes it possible to level out the uncertainty of the passenger’s choice of a specific route and obtain an objective integrated assessment of the load at the level of individual routes. The scope of practical application of the results includes technical monitoring systems for rolling stock, adaptive repair planning, optimization of timetables, as well as improvement of transport safety in the context of using anonymous means of payment for travel