Optimal Life-Cycle Costing with Partial Observability

Abstract

Optimal costing of structures requires taking into account lifetime management (inspection and maintenance) costs in addition to initial construction cost. Recently published papers on lifetime costing have used Markov decision process models to incorporate management policies into structural reliability analysis. In this paper, an optimization model is described based on partially observable Markov decision processes. The model depicts varying levels of uncertainty inherent in differing inspection techniques. The costs and inherent uncertainties of these techniques are related to environmental degradation from fatigue and corrosion processes. The mathematical approach leads to a management policy that explicitly includes frequency and type of inspection and extent of repair. The model is demonstrated in an example involving the management of a steel girder highway bridge.