Abstract

The average age of strategic constructions in the Western world is becoming higher and higher. Many of these structures need inspection, maintenance or replacement, resulting in significant costs. The accurate estimate of structural condition can make operators optimize the allocation of resources. Nowadays, the progress of technology and machine learning has made structural health monitoring appealing to the agencies that manage important structures. This has encouraged the research community in the study of new structural health monitoring methods. In spite of this, the use of monitoring data is often disregarded by practitioners, who still prefer to gather more information and then act based on experience. Similarly, unlike the design of civil structures, the design of structural health monitoring systems is carried out based on heuristics rather than on rigorous evaluations of the expected monitoring system effectiveness. In this doctoral thesis, I apply expected utility theory for the development of decision support systems to be used in structural health monitoring and I develop a procedure for the design of structural health monitoring systems that follows the scheme of semi-probabilistic structural design. The use of monitoring data in a decision support system that implements expected utility theory financially optimizes the management of civil structures. The proposed monitoring system design method enables practitioners to design monitoring systems using their experience and guarantees that the installation of a monitoring solution is financially convenient. I present the mathematical formulation for monitoring-based decision support systems and monitoring system design. Then, I propose the numerical algorithms for the development of monitoring-based decision support systems and solutions for monitoring data analysis. Finally, the proposed methods are applied to three case studies, which enabled me to discuss the application in real life and the hypotheses. The applications show also the feasibility of the proposed approaches and test the numerical algorithms.