An Open-Source OpenSeesPy Framework for Simplified Micro-Modeling and OMA-Based Model Updating of Masonry Infill Walls under Cyclic Loading

<title>Abstract</title> <p>This paper presents a fully open-source, two-tier numerical investigation of the cyclic and dynamic behavior of masonry infill walls, implemented in OpenSeesPy and validated against a half-scale experimental campaign that combined FEMA 461 quasi-static tests and operational modal analysis (OMA). A simplified micro-model is adopted, in which the masonry units are expanded and the mortar joints together with the infill–frame boundary are reduced to cohesive–frictional zero-thickness interfaces; every constitutive parameter is traceable to the calibration of the reference study, so that the benchmark is reproducible without commercial software. In the first tier, a two-dimensional model of the specimen is driven through the displacement protocol and reproduces the measured peak lateral capacity of approximately 96 kN within 1.5%, together with the pinched hysteresis and the diagonal, corner-separation damage pattern observed in the test. In the second tier, a three-dimensional model with stabilized brick elements first resolves an ambiguity in the reported unit geometry—only a panel thickness of 185 mm reproduces the measured out-of-plane frequency, within 2%—and is then updated to the OMA frequencies by bound-constrained least squares. The updating Table 2 compares the updated model frequencies with the measurements. The updated model matches the three measured modes within + 2.4%, − 5.8%, and + 12.1% (mean 6.8%) and identifies an effective in-plane infill–frame interface stiffness roughly five orders of magnitude below its cyclic-range value. This ratio provides a direct, experiment-based measure of the amplitude dependence of frame–infill interaction and implies that natural frequencies obtained from ambient vibration characterize the unlocked-contact state and should not be used to calibrate the stiffness entering seismic capacity models. The study thus delivers a reproducible open-source benchmark and demonstrates that separating the small- and large-amplitude regimes exposes interface behavior that a single-state calibration conceals.</p>