PSI - Issue 78

Alina Elena Eva et al. / Procedia Structural Integrity 78 (2026) 387–394

390

Fig. 2. Geometry of the panel and fac¸ade design

3.1. Masonry modelling

To generate the data required for training and validating the DANN neural network, a finite element model (FEM) of a masonry panel was developed using Abaqus. The objective is to simulate seismic behavior and analyze damage evolution in masonry, particularly through strain measurements collected by smart brick sensors. The damage was assessed by comparing the distribution of deformations under dead-load alone, before and after the seismic event. The panel, built with Gothic bond pattern, measures 1095 × 1235 mm and consists of solid clay bricks (250 × 120 × 55 mm) and cement mortar joints 10 mm thick (see Fig. 2). A continuous micromechanical modeling approach was adopted, in which bricks and mortar are modeled sepa rately, without including explicit interfaces due to the lack of detailed experimental characterization A.M. D’altri et al. (2018). The mortar used is of type M3, and the bricks have a compressive strength of 35 MPa. Based on the results of Drougkas et al., it was considered that the average ratio between Young’s modulus and compressive strength is approximately 700 for mortar and 300 for bricks A.Drougkas et al. (2015). These values were used in the model. The mechanical properties adopted are listed in the Tab. 1. To simulate the nonlinear behavior of masonry, the Concrete Damage Plasticity (CDP) model was used, which allows distinguishing between tensile and compressive behavior through dimensionless damage parameters (dc, dt), representing an isotropic reduction of the elastic modulus from 0 (no damage) to 1 (zero sti ff ness). The tensile damage of the material was defined assuming a loss of sti ff ness of 90% at the maximum crack displacement ut = 1mm. This corresponds to a fracture energy G c f = ( σ t 0 · u t 0 ) / 2 ABAQUS, ® Theory Manual (1999). Compression damage was defined assuming a 90% loss of sti ff ness at inelastic deformation, corresponding to a 90% decay of stress. In defin ing the compression behavior, being a theoretical model without experimental validation, the compression behavior proposed by A.Drougkas et al. (2015) was considered, modeling a parabolic compression curve based on the fracture energy in compression, assuming that the damage index d = ( G c f / f c ) is equal to 1 mm. The FE model was discretized with C3D8R elements.

3.2. Damage scenarios

In this case study, several damage scenarios induced by seismic action on the panel were considered. The seis mic response was simulated through an in-plane cyclic loading test under displacement control. Displacements were