Influence of wall-to-floor connections and pounding on pre- and post-diction simulations of a masonry building aggregate tested on a shaking table

This paper presents numerical simulations within the frame of the project SERA-AIMS (Seismic Testing of Adjacent Interacting Masonry Structures). The study includes blind pre-diction and post-diction stages. The former was developed before performing the shaking table tests at the laboratory facilities of LNEC (Lisbon), while the latter was carried out once the test results were known. For both, three-dimensional finite element models were prepared following a macro-modelling approach. The structure consisted of a half-scaled masonry aggregate composed by two units with different floor levels. Material properties used for the pre-diction model were based on preliminary tests previously provided to the participants. The masonry constitutive model used for the pre-diction study reproduced classical stress-strain envelope, whereas a more refined model was adopted for the post-diction. After eigenvalue analysis, incremental nonlinear time history analysis was performed under a unique sequence based on the given load protocol to account for damage accumulation. In the post-diction, the numerical model was calibrated on the data recorded during the shaking table tests and nonlinear dynamic analysis repeated under the recorded accelerogram sequence. The interaction between the two units was simulated through interface elements. Moreover, the timber floors were accounted following different strategies: not modelling or considering nonlinear wall-to-floor connections. Advantages and disadvantages are then analysed, comparing the pre-diction and post-diction results with the experimental data. Numerical results differ from the experimental outcomes regarding displacements and interface pounding, although a clear improvement is visible in the post-diction model.