Modeling heterogeneity in transportation infrastructure deterioration

One of the key elements for managing transportation infrastructure is to accurately capture and predict the performance of the facility through well established deterioration models. A sound deterioration model should incorporate 1) physical principle that reflects the deterioration mechanism; 2) relevant variables affecting the deterioration process; and 3) rigorous statistical approach to estimating the model. This dissertation aims at addressing these critical issues with focus on highway pavements. Data collected from in-service pavement sections are adopted to capture the real-world pavement deterioration process. A widely used pavement performance indicator, riding quality in terms of International Roughness Index (IRI) is used. A nonlinear model with a hierarchical parameter structure is formulated to effectively account for both observed and unobserved heterogeneity. The model is estimated through an econometric technique, Maximum Simulated Likelihood estimation. Simulation is employed to solve the computationally challenging problem of multi-dimensional integration. Engineering implications based on estimation results are discussed. The findings are not only consistent with engineering judgment but also helpful to reveal and enhance understanding of the pavement deterioration mechanism. Furthermore, the proposed methodology provides flexibility to obtain both parameters reflecting deterioration for all units and each individual unit of the population. The second part of the dissertation establishes and evaluates optimal maintenance policy on the basis of realistic deterioration models. The optimal policy is obtained so that the total cost, agency plus user cost, is minimized. A steady state resurfacing problem is investigated in the case study. In particular, the effect of model accuracy related to unobserved heterogeneity on total cost is discussed. This study makes a contribution to transportation infrastructure management and design in the following sense. From a management viewpoint, the proposed methodology with hierarchical parameters can accommodate both network and project levels of management. It also facilitates decision making for budget planning and resource allocation. From a design viewpoint, model estimation results can be used to update the current AASHTO pavement design equation by incorporating other critical factors.