Managing disruption of rail-truck hazmat networks: a machine learning–optimization approach

Rail-truck intermodal networks serve as major freight infrastructure, transporting both regular and hazardous material. Accidents and infrastructure failures pose a significant threat to these networks due to associated losses to life, the environment, and the economy. Dealing with these risks is challenging due to the physical and economic scale of the problem. Developing efficient disaster management plans is thus operationally and economically quite challenging. We propose an optimization and machine learning methodology for this problem. In this methodology, impact-based categorization and classification of unknown service legs or intermodal terminals are done via appropriate clustering and classification models, while for the optimization of the shipment plans, a bi-objective model is developed that employs network criticality measures as determined in the machine learning phase. The methodology was applied to a rail-truck intermodal network in the United States. The results indicate that post-disruption consideration should be incorporated into the transportation planning problem; machine learning algorithms can efficiently categorize network elements with high accuracy; and efficient pro-active post-disruption management can avoid a significant increase in cost and associated risks.