Estimation of infrastructure performance models using state-space specifications of time series models

We consider state-space specifications of autoregressive moving average models (ARMA) and structural time series models as a framework to formulate and estimate inspection and deterioration models for transportation infrastructure facilities. The framework provides a rigorous approach to exploit the abundance and breadth of condition data generated by advanced inspection technologies. From a managerial perspective, the framework is attractive because the ensuing models can be used to forecast infrastructure condition in a manner that is useful to support maintenance and repair optimization, and thus they constitute an alternative to Markovian transition probabilities. To illustrate the methodology, we develop performance models for asphalt pavements. Pressure and deflection measurements generated by pressure sensors and a falling weight deflectometer, respectively, are represented as manifestations of the pavement's elasticity/load-bearing capacity. The numerical results highlight the advantages of the two classes of models; that is, ARMA models have superior data-fitting capabilities, while structural time series models are parsimonious and provide a framework to identify components, such as trend, seasonality and random errors. We use the numerical examples to show how the framework can accommodate missing values, and also to discuss how the results can be used to evaluate and select between inspection technologies.