PSI - Issue 78

Ebrahim Aminifar et al. / Procedia Structural Integrity 78 (2026) 1466–1473

1467

Central Italy (2016 – 2017), has revealed the high seismic risk these buildings face and underscored the urgency of effective vulnerability assessment strategies (Lagomarsinoand Podestà, 2004a). Among ecclesiastical structures, single-nave churches are particularly common and exhibit various morpho-typological features influenced by historical, regional, and religious developments. Their seismic performance is often compromised by out-of-plane failure mechanisms, weak wall-to-roof connections, and structural discontinuities introduced by non-coeval additions or retrofits (Borri et al., 2019; De Matteis et al., 2016). To further clarify such vulnerability patterns, several post-earthquake reconnaissance campaigns have been conducted, notably following the Central Italy earthquakes of 2016 – 2017. The efforts were supported by the development of the Da.D.O. database (Database of Observed Damage), which catalogues over 4000 churches, with detailed records on architectural features, damage mechanisms, and usability status (Di Meo et al., 2023). The resource, combined with systematic typological analysis, enables large-scale assessments of seismic performance and the identification of recurrent fragility indicators (Valente et al., 2017; Lagomarsino, S., & Podestà, S. 2004a). Within this context, typological classification is increasingly recognized as a critical first step in the development of data-driven fragility models. Parameters such as plan geometry, aspect ratios, vaulting systems, and retrofitting history can be statistically analyzed to correlate specific architectural traits with observed damage levels (Cianchino et al., 2023). The approach lays the groundwork for integrating machine learning techniques that can automate fragility curve estimation based on typological attributes, offering a scalable solution for the assessment of vast historic inventories. The present study aims to establish a rigorous typological and geometric framework to support the seismic vulnerability assessment of historical churches in Italy. Based on data extracted from the Da.D.O. (Database of Observed Damage), the most recurrent and structurally coherent architectural configuration, namely, the single-nave rectangular typology, is identified and isolated. Variability within this subset is examined with respect to construction period, geographical distribution, and geometric features in order to support stratified classification. The resulting framework is intended to inform the selection of representative cases suitable for future numerical simulations. Outputs from these analyses will serve as input for training models that use machine learning to predict typology-specific fragility curves. Although numerical modeling and predictive algorithms are beyond the scope of the present study, the proposed methodology provides the foundational dataset and classification strategy necessary for data-driven seismic risk assessment of heritage masonry churches. 2. Data Description The data utilized in the study were sourced from the Da.D.O. (Database of Observed Damage), an IT platform conceived by the Italian Civil Protection Department (ICPD) and developed by EUCentre (European Centre for Training and Research in Earthquake Engineering). It systematically documents post-seismic damage observed in architectural heritage across Italy, with a particular focus on ecclesiastical buildings, which are among the most vulnerable typologies in seismically active regions. The dataset includes 6,362 churches surveyed primarily in the aftermath of major seismic events that occurred from the 1976 Friuli earthquake through the Central Italy seismic sequence of 2016 – 2017. Regional data coverage includes 3,357 entries from Central Italy, 1,148 from Umbria Marche, 1,028 from L’Aquila, 265 from Emilia, 277 from Salò, 150 from Piemonte, 117 from Molise-Puglia, and 28 from Ischia (Di Meo et al., 2023). In addition to damage reconnaissance, geometrical features and other information are collected for each church; however, for some churches, this data is unavailable. For instance, churches from Emilia earthquake were not considered since non information was present. For the purpose of the present study, thus, only churches with full and consistent data across key parameters such as construction period, architectural layout, number of naves, geometric dimensions (length, width, height, and area), and primary construction material were retained. The filtered subset forms the analytical foundation of the present work, enabling statistically robust typological classification and spatial comparison across regions. The regional breakdown is presented in Fig. 1, with the highest concentrations of churches documented in Central Italy (1,054 ) and the Province of L’Aquila (516). Smaller clusters are noted in Piemonte (114), Salò (58), Umbria Marche (23), and Ischia (16).