Influence of record-to-record variability on the assessment of unreinforced masonry buildings

A major source of uncertainty in the seismic assessment of existing masonry buildings lies in the definition of the seismic action. While nonlinear quasi-static or dynamic analyses typically address epistemic uncertainties, record-to-record variability can significantly influence the statistical distribution of structural response. This study investigates the impact of record-to-record variability on the seismic assessment of unreinforced masonry (URM) buildings representative of pre-code construction in seismic-prone urban regions. The building stock is modelled in OpenSees using a three-dimensional macroelement approach capable of reproducing both in-plane and out-of-plane wall behaviour, and simplified nonlinear material laws for interface elements. The seismic action is defined based on pre-defined seismological scenarios for multiple soil types. The combination of building archetypes (varying in plan configuration and number of storeys) and seismic intensity levels resulted in approximately 45,000 nonlinear response-history analyses executed via cloud computing. From these simulations, capacity curves and empirical distributions of engineering demand parameters (EDPs) and intensity measures (IMs) are derived across multiple damage states. Fragility functions are developed, and a bootstrap-based framework is implemented to quantify ( i ) the number of records required for stabilisation of median and dispersion estimates under different convergence thresholds, and ( ii ) the minimum number of analyses needed to capture a target seismic demand with prescribed probability levels. The results provide quantitative guidance on ground motion suite size for nonlinear dynamic assessment of existing URM buildings, supporting a rational balance between statistical reliability and computational efficiency.