Optimal Structural Health Information approaches for the efficient classification and management of structural systems

As the existing infrastructure ages while industrial and societal demands continue to rise, asset managers and operators look towards approaches to better manage and utilise the structures. To aid in the process, the asset managers and operators acquire additional information to gain knowledge about the structure and guide actions. However, collecting information and performing actions are associated with costs and consequences that may not always be justified by their benefit. Therefore, before acquiring information or implementing an action, the benefit and impact of the decisions should be quantified for efficient and sustainable allocation of resources.
This thesis focuses on normative decision making for optimising structural health assessment and achieving efficient management and classification of existing structures. The thesis demonstrates utilising decision analytical approaches in identifying optimal strategies by considering probabilistic models for structural performance, knowledge (information) and performance (actions) in conjunction with economic and life-safety consequences applicable throughout the structure’s service life. The work incorporates modelling and integration of information from multiple sources into structural reliability models. Specifically, probabilistic models and approaches to utilise linear elastic proof loading, proof load monitoring, laboratory testing and Unmanned Aerial Vehicle (UAV) -based damage detection information for structural reliability updating are developed. Three decision situations are considered: 1) Proof load testing of an existing bridge with tailored monitoring of the test, 2) Re-classification of existing bridge’s capacity using in-situ and laboratory testing, and, 3) Information acquirement from Unmanned Aerial Vehicles for integrity management of an offshore wind turbine support structure. The former two scenarios have basis in a classification system adopted by the Danish Road Directorate. The classification is determined based on the structural traffic live load carrying capacity, and can be related to load rating procedures adopted in other countries.
The diverse decision scenarios and information and action choices considered require utilising an extension to the classical Bayesian decision analysis framework. Specifically, decision analysis for proof load test planning required modelling proof loading as an action guided by monitoring information, with explicit modelling of the uncertainties associated with the test outcome. The extension also made it possible to distinguish decision scenarios based on the state of the information acquirement as well as action implementation. The present work contributes towards co-formulation of the decision value analyses concept and classification of decision value analyses types depending on the sets of optimal information strategy and actions, and the states of information acquirement and action implementation. Efficiency of various information acquirement and performance management strategies is assessed by quantifying the Value of Information, Value of Actions or Value of Information and Actions, respectively, depending on what leads to an expected utility gain.
The performance of structural systems is modelled probabilistically accounting for physical, statistical and model uncertainties. Structural Health Information is modelled accounting for the information type, costs and precision. Here, approaches for modelling of laboratory testing and proof loading and its utilisation in updating structural reliability are introduced and developed. Structural integrity and performance management actions e.g. repair, proof loading, and, reclassification, are modelled accounting for the associated uncertainties and consequences.
The applicability of the approaches is demonstrated with numerical case studies illustrating the modelling, information integration, structural reliability updating, and decision optimisation. A novel case study on efficiency assessment of Unmanned Aerial Vehicles for Structural Health Monitoring is presented with a Value of Information analysis. Further, different strategies for acquiring information using UAVs are modelled with explicit consideration of the relative precision and uncertainties of the information and associated costs.
The thesis contributes significantly to the modelling sophistication for load testing. A novel approach to model and integrate proof loading in structural reliability models is developed to optimise proof load monitoring, choice of stop criteria, and proof load level selection. The approach enables balancing the risks associated with proof loading with the service life performance benefits of the tested structure. It is demonstrated that load testing can be combined with information from monitoring during the test and integrated into structural reliability models and decision scenarios, including the scenario of limited availability of bridge design documentation.
A decision analytical approach for reclassifying existing bridges by consistently quantifying the risks and benefits of changing the allowable traffic load level is developed, thereby facilitating structural safety, economic efficiency and sustainability. The approach utilises pre-posterior and posterior modelling of in-situ and laboratory test measurements in updating bridge ultimate load carrying capacity and enables quantifying the information value and optimal capacity classification.
The introduced and developed approaches enable identifying optimal strategies leading to an expected utility gain in the management and classification of structural systems. Further, it is possible to identify optimal proof load levels, decision rules for proof loading and reclassification and comparing the efficiency of various Structural Health Information approaches.