STATISTICS OF FUEL LEVEL SEGMENTS WITH INTERPOLATION OF GAPS
DOI:
https://doi.org/10.31891/2219-9365-2026-87-39Keywords:
fuel level, time series, linear interpolation, Gaussian-weighted smoothing, data segmentation, measurement data quality, shunting locomotiveAbstract
The paper presents the results of a comprehensive statistical analysis of telemetry data collected from shunting diesel locomotives by on-board monitoring systems. The object of the study is the time series of fuel level, running speed and diesel-generator set power; the subject is the set of methods for restoring the integrity of such series and for their subsequent segmentation. The work focuses on the processing of measurement data. The problem of recovering missing values in time series by means of linear interpolation is considered. A continuous data array of 4,898,881 minute-by-minute measurements collected from a fleet of six locomotives is analyzed. The discrete rate of change of the fuel level and a Gaussian-weighted moving-average filter are applied jointly to control non-physical signal jumps. Separately, 97,494 samples were found to lie below the known threshold fill level of the tank; these samples were replaced with interpolated values. It is established that interpolation of micro-gaps lasting up to 10 minutes, of which 325 were recorded, restores the continuity of the fuel level series, whereas for 53 macro-gaps with a maximum recorded duration of 195 days end-to-end interpolation is not applied. The statistical characteristics of the fuel level distribution are determined: the skewness of the series ranges from -0.510 to -0.247 and the kurtosis from 2.14 to 4.35. After the integrity of the series was restored, the array was divided into 67,855 operational segments, among which 280 refuellings were identified. Micro-gaps account for only 0.022 % of the array. Control experiments showed that recovering a much larger share of samples (2.0 % of the array — points below the known threshold fill level of the tank) changes the number of segments by 0.015 % and the number of refuellings by only one; a hypothetical direct recovery of the micro-gaps themselves (0.022 % of the array) changes the number of segments by 0.079 % and the number of refuellings likewise by only one.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Даніїл ІВАЩЕВ, Володимир ГЕРАСИМОВ, Анатолій КОРОЛЬ

This work is licensed under a Creative Commons Attribution 4.0 International License.


