Two vexing issues face railroad management and policy makers: How to make inspection and maintenance decisions to minimize service disruptions and delays to customers; and how to use available data to quickly grasp patterns in grade crossing accidents to improve allocation of resources and intervention strategies. This research aims to provide railroad decision and policy makers with improved tools to address these two problems. The Maintenance of Way (MOW) research used data provided by a freight railroad. Further, railroad personnel were Subject Matter Experts (SMEs) to specify context for the data, including specific terminology used by railroad personnel and identifying whether the maintenance activity was major or minor. The grade crossing analyses utilized two publicly available databases: FRA Highway-Rail Crossing Database and FRA Highway Rail Accidents Database. Our first analyses derived conditional probabilities from latent Dirichlet allocation (LDA) text mining for input into a maintenance optimization model for scheduling surfacing activities to minimize service disruptions. Our second set of analyses demonstrates two tools -ggplot and Tableau - to identify dangerous crossings and visualize accidents by driver gender and vehicle type. The results demonstrate topic modeling can be applied to maintenance data to identify where maintenance activities will most likely impact customers. Further, the results demonstrate two readily available Big Data visualization tools can be utilized by railroad personnel and policy makers to identify dangerous accident locations across the US and vehicle and driver characteristics that may provide insights into possible intervention strategies for reducing the frequency of those types of accidents.
展开▼
