| dc.description.abstract | The increment of extreme weather events is creating larger, and more frequent problems among infrastructure systems worldwide. In this thesis, the focus is on traffic networks by the analysis of a recent concept, the resilience. This concept evaluates the impact that perturbations create on traffic networks from the beginning of the perturbation until the total recovery. Due to the novelty of the concept in the Transport area, this thesis aims to develop new mathematical tools in order to quantify, and understand this concept in traffic networks.
Following the previous objective, this thesis provides the reader with the following
contributions:
Literature review. A literature review about existing definitions, and methodologies
to evaluate resilience in transport networks is presented, together with an analysis of the existing traffic assignment models, and travel cost functions. In addition, a literature review of methodologies for the identification of critical, and vulnerable links is included.
A dynamic restricted equilibrium assignment model. A methodology to evaluate resilience in traffic networks is presented, based on a new dynamic restricted equilibrium assignment model, which evaluates not only the evolution of the cost level during the whole evolution of the perturbation, but also the stress suffered by the network users due to the changes in the travel conditions. In addition, formulations to evaluate the perturbation, and the recovery resilience are introduced.
A bounded link travel cost function. A new link travel cost function which
explicitly considers the effects of weather events in traffic networks is introduced, including not only a parameter to determine the intensity of the hazard, but also a parameter to include the local vulnerability of each link when the perturbation occurs.
A mapping, and a bi-phase sensitivity analysis. A bi-phase sensitivity analysis
of the parameters included in the proposed method to evaluate resilience, including local approach (OAT), and a global approach (Latin Hypercube) is presented. The statistical approach implemented allows its use in complex models by reducing the number of points needed for the analysis, with the consequent time saving.
Methodologies to identify critical, and vulnerable links. Novel methodologies
to identify, and rank the links of traffic networks by their vulnerability, and also by their criticality are presented. The methods are based in the analysis of the Fisher Information Matrix, and its eigenvalues and eigenvectors. In addition, this methodology is extended for the identification of vulnerable, and critical areas of a traffic network.
Practical applications. The proposed methodologies are tested in examples, and
real traffic networks in order to show their performance, and characteristics. In addition, the presented examples allow the validation of the results, and the associated computational requirements. | en |
