Decision-making based on Structural Health Monitoring: a Reliability-based Framework

The average age of strategic infrastructure in the Western world is steadily increasing, raising concerns about inspection, maintenance, and replacement costs. An accurate assessment of structural conditions is crucial to optimize resource allocation. Nowadays, advances in technology and computational methods have made structural health monitoring (SHM) increasingly attractive to researchers, prompting the scientific community to explore innovative methods that also take advantage of artificial intelligence. The recent collapse of bridges, especially the Morandi Bridge in Genoa in 2018, and the publication of specific guidelines for the management of existing bridges have further emphasized the need for effective monitoring systems to ensure the safety of bridges. However, a consolidated and widely applicable framework for monitoring data interpretation is still lacking. In this work, I propose an SHM framework based on the decision theory built on Bayesian networks (BNs). The structure of BNs enables an effective representation of the monitoring variables involved in the interpretation process and their interconnections. In the decision-based SHM perspective, structural reliability provides a crucial link between SHM information and decision-making. In the proposed BN-based framework, structural reliability is assessed as a function of monitoring observations. Under the assumption of linear models and normally distributed random variables, the relationship between monitoring observations and structural reliability can be expressed with an easy-to-define closed-form solution. This solution allows for a clear interpretation of the contribution of each information source, including prior knowledge. The proposed SHM framework contributes to define a rational process for the use of monitoring information, whether obtained from on-site testing or from sensor-based systems. The rationale behind the framework suggests how monitoring systems should be designed. Nowadays, monitoring system design is mainly based on heuristics derived from the experience of operators and practitioners. However, in part due to the prescriptive nature of existing guidelines, operators often place excessive confidence in on-site tests and continuous monitoring. In the process of designing monitoring systems, the uncertainties involved should be considered in relation to prior knowledge and expected benefits. It is widely agreed that a monitoring system is effective if it influences decision-making. As a consequence, structural reliability must be sufficiently sensitive to monitoring observations. Building on this consideration, I develop a general reliability-based method to evaluate monitoring system performance, as well as a formal approach for defining monitoring thresholds within a decision-theoretic framework.In addition, this work presents an application of the proposed framework to the design of non-destructive tests (NDTs) on post-tensioned prestressed concrete bridges. In this context, a metrological validation of several NDT techniques is carried out within the Alveo Vecchio project.