A Risk- and Performance-based Infrastructure Asset Management Framework

Abstract

Asset managers who maintain the infrastructure can improve the efficiency of their practices by making use of decision-support frameworks. These models use performance data to help maximize the outcomes of maintenance strategies and financial allocations. In this research a new decision-support framework for asset managers was developed and tested. This framework improves upon existing models in two respects: (a) it provides an effective and practical method of accounting for uncertainty/risk, and (b) it includes a method for predicting asset performance over time in situations where there is limited historical data.

Outcome-based scenario analysis was chosen as the most effective approach to model risk in asset management. The proposed framework presents managers with “best-case,” “most-likely case,” and “worst-case” scenarios, which are defined by applying quantile regression analysis to the asset-performance data. For situations in which there is a lack of adequate historical performance data, an elicitation model was developed based on the Delphi technique. This approach provides a rigorous method for estimating asset performance using information solicited from a panel of experts. The elicited data was aggregated by a Bayesian hierarchical model and the Markov Chain Monte Carlo algorithm.

A case study was conducted to demonstrate the applicability of the decision-support model. While the proposed framework is generic and could be used for any type of asset, this study involved pavement condition on the city streets of Bryan, Texas. The results indicated that using a traditional deterministic model (rather than a scenario-based approach) could lead to significant over- or under-estimation of the budgets required to achieve certain asset-performance results. This demonstrates the urgent need for asset managers to use a practical model that can provide them with information about uncertainty and risk in asset-performance assessments. The case study also demonstrated the effectiveness of the data-elicitation technique, as the results of this approach were shown to be commensurate with historical information about pavement performance collected by the city of Bryan. The success of this approach in approximating historical performance trends provides evidence for its usefulness in situations where such historical data is unavailable.