PREDICTION OF MODAL FEATURES FOR DIFFERENT DAMAGE STAGES AND RETROFIT METHODS OF A MASONRY BUILDING

Under seismic and gravity loading, existing unreinforced masonry (URM) cultural heritage buildings are vulnerable and prone to damage. With careful calibration and thorough validation, credible results can be obtained from the numerical tools, but this process requires reliable experimental data. Therefore, various blind prediction contests have been held recently. For example, a blind prediction contest of a two-story URM building with a masonry vault above the ground floor and timber floor above the first floor was launched in 2023 under the DETECT-AGING project to test the damage detection capabilities of the (Structural Health Monitoring) SHM system and the reliability of numerical tools. Since the experimental data of the measurements are not yet published, we only present the results of the detailed numerical model. For building the model geometry, we used CAD tools and software Abaqus. A series of sensitivity analyses followed to calibrate key parameters of the numerical model, such as the activation and deactivation of installed steel anchors and horizontal ties, stiffness of the orthotropic timber material and CDP (Concrete Damage Plasticity) material model for masonry, by comparing experimental data included in the published competition documentation. A simple parameterization of the CDP material model was done by developing a visual script that allows the user to generate a series of input files for Abaqus. The results of the numerical model showed differences in modal shapes and frequencies compared to a reference state with a significant change in the damaged state, as expected. Although the blind prediction competition was aimed at estimating the damage accumulation due to ageing, degradation and soil settlement, the results also make a significant contribution to numerical simulations of cost-effective, preventive and non-invasive retrofitting methods of existing URM buildings in case of earthquake loading. Future work includes post-diction benchmarking of the numerical model using the experimental data.