Seismic Vulnerability Assessment of Historical Masonry Buildings in Aggregate

The seismic vulnerability of unreinforced masonry (URM) buildings in aggregate, typical of historic centers, is an extremely complex and highly debated topic in the field of earthquake engineering. This PhD thesis addresses the seismic vulnerability of historic buildings in aggregate by employing two complementary approaches, i.e.: an innovative machine learning (ML) technique and detailed numerical nonlinear analyses (NLDA). The two approaches have complementary and distinct goals, too. Specifically, the ML technique is applied to two historic centers with the main goal of identifying the vulnerability parameters most strongly correlated with damage. The main outcome expected from this approach is the identification of a subset of factors to be collected in future applications to address the vulnerability characterization of buildings in aggregate also for large scale risk assessment. NLDAs are conducted on three case studies representatives of the built environment under examination with the main goal of providing a replicable procedure to quantify some key feature of their seismic behavior, such as the ”aggregate effect” to the in-plane seismic response. Accordingly, the research is structured into three main chapters. The first chapter presents a comprehensive review of the state of the art regarding the architectural complexity of URM buildings in aggregate, which are characteristic of urban historic centers, with a specific focus on their seismic vulnerability. The study examines the key features and vulnerability factors of this structural typology and reviews the existing methodologies proposed in the literature for assessing the seismic vulnerability of buildings in aggregate. The chapter highlights the strengths and limitations of each methodology, paying particular attention to unresolved issues, such as the aggregate effect, the appropriate modeling of the boundary conditions of the structural units (SUs) and the integration between in-plane (IP) and out-of-plane (OOP) mechanisms for seismic verification. The second chapter introduces the innovative application of ML techniques as support to interpret the seismic vulnerability in historic centers. A detailed study is conducted on the small historic center of Casentino (AQ) to explore the correlations between various vulnerability parameters and the possible activation of IP and OOP damage mechanisms. From these correlations, a critical subset of parameters is identified and validated through the historic center of Visso (MC), demonstrating the potential of ML to enhance the predictive accuracy of seismic vulnerability assessments. The third chapter focuses on the numerical analysis of representative case studies using advanced modeling tools, particularly the equivalent frame approach implemented in the Tremuri software. NLDAs are conducted from which fragility curves are derived. The seismic response of buildings in aggregate is evaluated by analyzing SUs both in isolated and in aggregate configurations, with special attention given to the modeling of boundary conditions and the integration of OOP and IP responses. Through a combination of literature review, ML techniques, and numerical analyses, this thesis provides an in-depth investigation into the seismic vulnerability of historic buildings in aggregate, offering insights for improved seismic assessment methodologies strategies.