Railway track condition change detection: A data-driven approach to track geometry and component degradation modeling

Venancio Da Silva Ramos, Jose Augusto

Permalink

https://hdl.handle.net/2142/127517 Copy

Description

Title

Railway track condition change detection: A data-driven approach to track geometry and component degradation modeling

Author(s)

Venancio Da Silva Ramos, Jose Augusto

Issue Date

2024-12-11

Director of Research (if dissertation) or Advisor (if thesis)

Edwards, John Riley

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

Railroad, Track degradation, Track geometry, Track components, Big Data, Ballast.

Abstract

The railroad track system is a critical transportation asset that is responsible for transferring wheel loads from rolling stock to the roadbed. To ensure safe and efficient operations, U.S. Class I railroads conduct frequent inspections, generating a substantial amount of track health data. With the growing adoption of data science tools, these datasets offer new opportunities for more sophisticated maintenance and safety analyses. While many railroads leverage data trending for geometry prediction, they often overlook the impact of evolving component condition and unrecorded maintenance, which are crucial for accurate degradation modeling. This thesis evaluates the relationship between track geometry degradation and ballast profiles across curved and tangent track segments using data from a primary corridor on a U.S. Class I railroad. A stochastic approach revealed a significant correlation between degradation, initial profile conditions, and the Ballast Health Index (BHI), with faster deterioration observed in areas with poor initial geometry and high BHI values. Additionally, cross-correlation was employed to address the challenge of identifying unrecorded maintenance activities, proving effective in detecting track changes and improving data quality metrics for linear degradation models. These findings provide a quantifiable method for assessing degradation under varying conditions, enhancing maintenance prioritization and demonstrating that cross-correlation can identify maintenance events and support track monitoring and maintenance planning.

Graduation Semester

2024-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/127517

Copyright and License Information

Copyright 2024 Jose Augusto Venancio da Silva Ramos

Owning Collections