Thumbnail Image


Publication or External Link






With the increasing natural and human-made disasters, the risk of an event with potential to cause major disruption to our transportation systems and their components also increases. It is of paramount importance that transportation systems could be effectively recovered, thus economic loss due to the disasters can be minimized. This dissertation addresses the optimization problems for transportation system performance measurement, decision-making on pre-disaster preparedness and post-event recovery actions planning and scheduling to achieve the maximum network resilience level.

In assessing a network's potential performance given possible future disruptions, one must recognize the contributions of the network's inherent ability to cope with disruption via its topological and operational attributes and potential actions that can be taken in the immediate aftermath of such an event. A two-stage stochastic program is formulated to solve the problem of measuring a network's maximum resilience level and simultaneously determining the optimal set of preparedness and recovery actions necessary to achieve this level under budget and level-of-service constraints. An exact methodology, employing the integer L-shaped method and Monte Carlo simulation, is proposed for its solution.

In this dissertation, a nonlinear, stochastic, time-dependent integer program is proposed, from operational perspective, to schedule short-term recovery activities to maximize transportation network resilience. Two solution methods are proposed, both employing a decomposition approach to eliminate nonlinearities of the formulation. The first is an exact decomposition with branch-and-cut methodology, and the second is a hybrid genetic algorithm that evaluates each chromosome's fitness based on optimal objective values to the time-dependent maximum flow subproblem. Algorithm performance is also assessed on a test network.

Finally, this dissertation studies the role of network topology in resilience. 17 specific network topologies were selected for network resilience analysis. Simple graph structures with 9~10 nodes and larger network with 100 nodes are assessed. Resilience is measured in terms of throughput and connectivity and average reciprocal distance. The integer L-shaped method is applied again to study the performance of the network structure with respect to all three resilience measures. The relationships between resilience and average degree, diameter, and cyclicity are also investigated.