Predictive Analytics in Positive Train Control: Advancing Railway Safety and Efficiency

On September 12, 2008, United States congress enacted the Railway Safety Improvement Act (RSIA08; P.L. 110–432),which mandated Positive Train Control (PTC) implementation on various passenger and freight railroads. The RSIA was resultant from a fatal collision between a Metrolink passenger train and a union pacific freight train, which resulted in 26 fatalities and 145 injuries reportedly due to human error. PTC is a communication-based train control system designed to minimize human error by enforcing stringent safety protocols during train operations. Despite its extensive adoption, there is a paucity of academic literature on PTC system architecture and its critical failure modes. This study aims to address this gap by analyzing PTC system principles, regulatory requirements and key failure modes that affect system reliability. It proposes the utilization of Artificial Intelligence (AI) to develop a predictive model for forecasting train delays due to system failures. Term frequency—inverse document frequency (Tf-idf) natural language processor (NLP) feature extractor was used to convert text data from historical incident record. Four machine learning algorithms were used for training and validation of the datasets, and based on the performance, RandomForestRegressor was selected as the best performing model with the highest explained _variance _score and the lowest errors across most metrics. The model was subsequently deployed for delay prediction, providing a decision support tool for rail operators. The findings of this study can inform railway authorities and academic researchers on strategies to enhance PTC system safety and reliability, thereby contributing to improved operational efficiency in railway transportation.