Quantifying the value of structural monitoring for decision making

Abstract

A life-cycle approach to infrastructure design and management involves decisions pertaining to operation, maintenance, intervention, and rapid response measures. Such an approach may only be conceived when formulated on the basis of observations during the life-cycle of these systems. Structural Health Monitoring (SHM) offers a tool to such an end, with sensors employed to generate information on the state of structural systems, which may then be exploited to derive performance indicators. A fundamental, practical question regarding monitoring of structural systems is however the quantification of any gains, monetary or otherwise, for infrastructure owners if they choose to install a monitoring system to their structures, in place of, or in addition to, other available choices, such as structural inspection visits. This essentially comprises a Value of Structural Health Monitoring (VoSHM) problem, which poses important mathematical and computational challenges related to several infrastructure system uncertainties, stochastic observations and their Value of Information (VoI), and any uncertain action outcomes. In this work, we implement optimal stochastic control approaches for infrastructure management in the form of Partially Observable Markov Decision Processes (POMDPs), which inherently possess the notion of the VoI into their formulation and, in fact, automatically utilize it at every decision step for decision-making. In addition, we show that based on POMDPs the VoSHM can be efficiently estimated, allowing for informative decisions by the structural owner, based on quantitative metrics in relation to the expected benefits of the SHM system. A representative application is shown in this regard for a multi-component engineering system, showcasing the wide applicability and effectiveness of the suggested approach and its practical merits.