Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.

Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.

Author ORCID Identifier


Campus-Only Access for Five (5) Years

Document Type


Degree Name

Doctor of Philosophy (PhD)

Degree Program

Civil and Environmental Engineering

Year Degree Awarded


Month Degree Awarded


First Advisor

Carlton L. Ho

Subject Categories

Civil Engineering | Geotechnical Engineering


This study addresses the complex challenge of estimating Track Degradation Rates (TDR) to improve the scheduling of preventive maintenance for railway tracks. The unpredictable nature of TDR necessitates a detailed analysis of a large dataset, which includes historic passenger track geometry data from 2011 to 2021, integrated with advanced Ground Penetrating Radar (GPR) and Light Detection And Ranging (LiDAR) data. The integration of these technologies is pivotal in examining 280 miles of passenger revenue track, allowing for a deeper understanding of subsurface and drainage conditions, which are critical factors in track degradation. LiDAR data, in particular, is instrumental in extracting key properties of ditches, such as depth, distance, and condition, all of which significantly impact track integrity. To address the issues of high TDR, which lead to increased maintenance costs and reduced ride quality, this study proposes various Machine Learning (ML) models to predict both short-term and long-term behavior of railway tracks under different conditions. For short-term predictions, the study employs classifiers trained to define the probability of encountering yellow tag defects in subsequent inspections with high accuracy. In contrast, long-term behavior prediction involves the use of both stochastic analysis and ML regression models. The stochastic models focus on estimating expected TDR based on subsurface and drainage conditions, while the ML regression models aim to predict TDR values using 12 extracted features with high accuracy. The results of this comprehensive analysis are pivotal in enhancing the scheduling of geometry maintenance for railway tracks, thereby providing reliable data for decision-making in various maintenance activities, including ballast cleaning and ditching. This approach, combining extensive data analysis with advanced ML techniques, marks a significant advancement in the field of railway maintenance, ensuring greater efficiency and safety in track management.


Available for download on Saturday, February 01, 2025