Observation-based Decision-making for Infrastructure

Abstract

Abstract. The aim of this review paper is to summarize available options for decision-support toward the optimal management of infrastructure systems under consideration of uncertainties. Thereafter, we elaborate on a particular variant, namely Partially Observable Markov Decision Processes (POMDP), as a flexible framework for the incorporation of information from observations (visual inspections, non-destructive testing and monitoring) in the context of optimal inspection and maintenance planning. Examples from the literature are presented, demonstrating the potential of planning under uncertainty and the links to the Value of Information (VoI) from Structural Health Monitoring (SHM).

Value of structural health information in partially observable stochastic environments

“…}, and {(3,3,3)}, penalties are -5.0, -10.0, -14.0, and -18.0, respectively, where vector (i,j,k) denotes component condition level combinations, i.e., (3,3,1) indicates that there are 2 components in condition level 3 and one component in condition level 1.…”

“…These new possibilities come with relevant questions related to the actual value and necessity of increased quality measurements or continuous SHM information in facilitating optimal actions. SHM frameworks seek to determine appropriate mappings from raw response measurements to condition and performance indicators, which can, in turn, support decision-making towards costeffective intervention and maintenance actions that increase safety and mitigate risks [3]. Decision-making pertains to the type and sequence of actions that are selected in order to optimize some overarching predefined life-cycle objective.…”