Description
TitleArtificial intelligence-aided prediction of broken rail-caused derailment risk
Date Created2020
Other Date2020-01 (degree)
Extent1 online resource (xviii, 189 pages) : illustrations
DescriptionBroken rails are the leading cause of freight train derailments in the United States. The American railroad industry annually spends billions of dollars on track inspection, maintenance and repair. Accurate prediction of broken rail risk is critical for the railroad industry to further improve operational safety, efficiency and the state of good repair.
This dissertation research focuses on predicting the risk of broken rail-caused derailment via Artificial Intelligence (AI) empowered by the fast-growing “big data” in the railroad industry, related to network-level track characteristics, maintenance activities, traffic and operation, as well as condition monitoring. The intended contributions of this research include:
•Development of a novel, customized Soft Tile Coding based Neural Network model (STC-NN) to predict the spatial-temporal probability of broken rail occurrence for any given time horizon. This proposed AI algorithm shows superior performance over several alternative algorithms in terms of solution quality, computational efficiency, and modeling flexibility.
•An analysis of the relationship between the probability of broken rail-caused derailment and the probability of broken rail occurrence. New analyses are performed to understand how the probability of broken rail-caused derailment may vary with infrastructure characteristics, signal type, weather, and other factors.
•Development of an Integrated Broken Rail Derailment Risk Model for predicting location-centric broken rail-caused derailment risk on the network-level. Predicting and identifying “high-risk” locations can ultimately lead to significant safety improvement and cost savings.
The major conclusions of this research include:
•The proposed STC-NN algorithm can predict broken rail risk for any time period (from 1 month to 2 years), with better performance for near-term prediction than long-term prediction. The algorithm slightly outperforms Extreme Gradient Boosting, Logistic Regression, and Random Forest, and is also much more flexible.
•Appropriate network segmentation is important for prediction accuracy. Our proposed dynamic segmentation scheme shows a significant improvement over the fixed-length segmentation scheme.
•Segment length, traffic tonnage, number of rail car passes, rail weight, rail age, track curvature, presence of turnout, and presence of historical rail defects are all found to be among influencing factors for broken rail occurrence.
•Signaled track in the cold season has the lowest ratio of broken rail-caused derailments to broken rails, while non-signaled track in warm weather has the highest. Moreover, lower FRA track classes (Class 1, Class 2) have higher ratio of broken rail-caused derailments to broken rails, compared with higher track classes Class 3 and Class 4.
•A longer, heavier train traveling at a higher speed is associated with more cars derailed per broken rail-caused derailment.
This work uses enterprise-level big data for over 20,000 miles of track from a major freight railroad in the United States. The new methodology, algorithm, and analysis results can potentially be implemented for railroad rail asset management, in support of both short-term inspection and maintenance prioritization as well as long-term capital planning and resource allocation.
NotePh.D.
NoteIncludes bibliographical references
Genretheses, ETD doctoral
LanguageEnglish
CollectionSchool of Graduate Studies Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Statistical Profile