ABSTRACT
A water utility would like to model condition of a key strategic asset using performance data to inform future maintenance decision-making. Currently, routine maintenance at water treatment works are informed by an assessment of component condition based on time-based inspections. We propose a method for using performance data gathered by the utility for legislative purposes to inform maintenance decision-making. This model framework is grounded in the theory and methods of stochastic processes and elicitation of expert judgment, but aligned with the practicalities of the industrial decision-making context. After describing the asset to be modelled, the data available and the principles of the Hidden Markov Model developed, we illustrate an application to three water treatment works. The credibility of the modelling assumptions to accommodate the limitations of the real data are discussed.
