Parametric Analysis of Value of Information for Monitoring Infrastructure Components

Abstract

Information collected by monitoring systems can provide a significant economic benefit to the operation and maintenance of infrastructure components only under specific conditions. Not only that information has to be precise and useful for relevant decision problems, e.g. for scheduling maintenance and repairs, but the decision maker has to be able to process that information and react in due time. Moreover, the detected phenomena do not have to be predictable without the monitoring system; otherwise the information would be redundant. Formally, all these considerations can be embedded in the Value of Information (VoI) metric, a utility-based metric for assessing the impact of data in decision making under uncertainty. For computing the VoI, we predict the outcomes of sensors, the reaction of controllers and the overall cost, and compare it with the expected cost without using the monitoring systems. In this paper, we investigate the relation between the VoI and key features of the monitoring systems, of the component deterioration and of the decision-making process, including sensor availability, precision, degradation rate, damage predictability, reaction time, repair costs, and economic discount factors. Leveraging previous work, we model the maintenance process as a Partially Observable Markov Decision Process (POMDP), and we compute the VoI of long-term monitoring. The parametric analysis allows us to identify the optimal amount of information to be collected (without over or under-instrumenting components) and, e.g., to compare the VoI of frequent imprecise measurements with that of rarer but more precise ones. Overall, our framework is useful for identifying conditions when the benefit of monitoring is high, and it can be applied to a broad range of infrastructure systems.