PSI - Issue 78

Ciro Canditone et al. / Procedia Structural Integrity 78 (2026) 1855–1862

1859

uniaxial behavior. Shear behavior is, on the other hand, assumed linear up to the attainment of peak Mohr-Coulomb shear stress. The adopted failure criterion is Mohr-Coulomb with cut-offs both in tension and compression. 5. Discussion of numerical analyses As previously stated, two different numerical models were developed. The approximately 0.75 m thick loadbearing walls were modelled in ABAQUS via eight-node hexahedral elements (C3D8R), hence disregarding masonry bond pattern. In the AEM software package, on the other hand, masonry texture was simulated via 0.40 × 0.20 × 0.75m 3 (width × height × thickness) units jointed in a running bond pattern, hence disregarding the potential for transverse splitting. Good quality wall-to-wall interlocking was assumed and modelled via staggered blocks distributed along wall-to-wall intersections. A high-fidelity modelling of the ceiling system, comprising barrel vaults and deformable timber diaphragms, was pursued in both environments. Within ABAQUS, 2D linear elastic beam elements were adopted for the timber beams, with planking being modelled implicitly via distributed loads. In the AEM model, on the other hand, both timber beams and planking were modelled via 3D solid elements; quasi-brittle behaviour was assumed, based on (Khorsandnia & Crews, 2015), and nailed connections considered via a simplified approach outlined in works such as (Calò, et al., 2021). This resulted in a total of 646,458 geometric DOFs for the FEM model and 72,214 DOFs for the AEM model, respectively. Axonometric views of both numerical models can be appreciated in Figure 2; the main façade is hidden in Figure 2b to highlight the modeling of vaults and floors. Perfectly fixed restraint conditions were, preliminarily, assumed in both models, with regards to live loads and self-weight analysis, combined according to Eurocode provisions with regards to exceptional loading conditions (EC1, 2002). In the subsequent analysis stage, different settlements were considered by imposed downward displacements according to a Gaussian-shaped soil settlement profile based on the procedure detailed in (Peck, 1969). A Gaussian distribution is in fact, as discussed in (Prosperi, 2025), able to capture both either tunnelling-induced or subsidence induced settlements, thus fostering a further generalization of numerical results. Maximum vertical settlement, δ z max , was taken equal to 15 mm; this value was applied at an approximately 15 m distance from the main façade’s left edge. A cylindrical distribution was considered, i.e., the subsidence curve was assumed to lay parallel to the main façade and to maintain its magnitude and distribution along the transverse axis (see Fig. 3c). Despite several studies (Giardina, et al., 2013; Prosperi, et al., 2023a; Prosperi, 2025) highlighting the important role played by soil-structure-interaction effects, a choice was here made to disregard it and consider greenfield displacements applied directly below pier bases. Settlements were applied incrementally via static analyses, considering a 0.1 mm and 0.15 mm step sizes with regards to the FEM and AEM model, respectively. The AEM simulations were run using a 16 GB RAM, AMD Ryzen 7 4700U (2 GHz) processor-equipped machine and took approximately 4 minutes. Analyses on the FEM model were, on the other hand, run using a 128 GB RAM AMD Ryzen 5 7600X 6-Core Processor-equipped machine and took approximately 5 hours each. As can be appreciated, despite the large number of considered DOFs, the AEM model proved more computationally efficient than its FEM counterpart.

(a)

(b)

(c)

Figure 3. Axonometric views of the FEM model (a) and AEM model (b). Position (c) of the building archetype on the normalized soil settlement distribution.