Analytical Approaches to Risk Assessment and Management for Railroad Transportation of Hazardous Materials

Abstract

Hazardous materials (hazmat), such as crude oil and gasoline, are harmful to humans and the environment because of their toxic ingredients, but their transportation is integral to sustain our industrial lifestyle. In North America, a significant portion of hazmat shipments is moved via the railroad network. Rail hazmat incidents are rare though the consequences could be catastrophic. The low probability–high consequence nature of such events mandate that a risk-averse plan be implemented for routing hazmat shipments. We propose a value-at-risk (VaR) and conditional value-at-risk (CVaR) methodology to route rail hazmat shipments, using the best train configuration, over a given railroad network with limited number of train services, such that the transport risk is minimized. Freight train derailment records of the Federal Railroad Administration (FRA) were analyzed to model the behavior of railroad accidents, and to estimate their conditional probabilities. The proposed methodologies were used to study several problem instances generated using the realistic network of a railroad operator in Midwest United States, and to demonstrate that they are superior to the existing risk measures in the literature in regard to providing risk-averse routing of hazmat shipments and being versatile enough to yield various routes based on the risk preferences of decision makers. Next, we propose a CVaR model, as a risk-averse routing plan for multiple rail hazmat shipments and multiple origin-destinations pairs, such that the total transport risk in the railroad network as measured by CVaR is minimized. However, it may happen that certain links and yards of the railroad network tend to be overloaded with hazmat traffic and risk. To overcome this issue, we also promote equity in the spatial distribution of risk. Therefore, the main problem is to find minimum risk routes, while limiting and equitably spreading the risk in any zone where the railroad network is embedded. The problem is mathematically formulated, and a heuristic algorithm is proposed for its solution, which takes into consideration the risk load limits on arcs and transferring yards and spreads the risk equitably throughout the network. Moreover, a lower bound based on a Lagrangian relaxation of the mathematical formulation is also provided. Finally, several computational experiments are developed using the above realistic railroad network.