Uncertainty in Infrastructure Deterioration Modeling and Robust Maintenance Policies for Markovian Management Systems

Infrastructure management systems help public works agencies decide when and how to maintain, repair, and rehabilitate infrastructure facilities in a cost effective manner. An integral part of an infrastructure management system is a model describing how the different infrastructure facilities to be managed deteriorate over time and with use. Many sources of error limit the ability of management systems to accurately predict how built systems will deteriorate. This dissertation introduces and examines different techniques for considering error and uncertainty in deterioration modeling within an infrastructure management system. Techniques used include robust optimization and adaptive control. In the context of robust optimization, both MAXIMIN and Hurwicz decision criteria are studied. Computational studies involving the simulated management of pavement systems illustrate the strengths and weaknesses of the proposed approaches. These studies involve short-term, limited horizon planning, as well as indefinite-term, infinite horizon planning. Single facility infrastructure management problems are presented alongside more complex problems involving the management of a network of an arbitrarily large number of related facilities. It is found that both robust optimization and adaptive control formulations have certain comparative advantages. Some discussion is included of the possibility of combining the robust and adaptive frameworks to create a new hybrid approach. Regardless of what approach is used, this work makes clear that consideration of uncertainty in deterioration modeling during decision-making can alter ‘optimal’ maintenance strategies selected and change the potential user and agency costs of infrastructure management.