PSI - Issue 78

Livio Pedone et al. / Procedia Structural Integrity 78 (2026) 1609–1616

1614

deformation. Numerical results show that moderate damage a ff ects the external beam–column joints of the first story, while slight damage is observed for base columns and remaining joint panels of the first 2 stories. Di ff erently, no damage is observed for members on the third story.

4.3. Pre- and post-earthquake nonlinear static (pushover) analyses

The seismic performance of the structure in its as-built (intact) and damaged configurations is investigated through nonlinear static analyses. The simplified analytical / mechanical SLaMA method is employed, considering its recent extension for applications in post-earthquake scenarios (Matteoni et al., 2023). Fig. 5a shows the results in terms of the force-displacement capacity curve for the as-built structure; moreover, a comparison between the SLaMA based pushover curve and the result obtained through a more refined numerical (software-based) simulation is also reported. As expected, due to the lack of capacity-design principles, a low-ductile global behavior is obtained. The simplified procedure allows for a relatively good estimation of the global capacity curve, both in terms of strength and displacement / ductility capacity. It is worth noting that, the modified SLaMA method proposed by Matteoni et al. (2023) is herein adopted to obtain a better estimation of the initial sti ff ness.

(a)

(b)

Fig. 5: Results of the nonlinear static analyses for the structure in its (a) as-built and (b) damaged configuration.

For the damage configuration, information on the earthquake-induced damage to structural components in the sim ulated post-earthquake scenario is used. It is assumed that images of earthquake damage to the structural members are collected and processed using the developed CNN model. To account for potential errors in the classification of damage severity, the confusion matrices shown in Fig. 2b are used to evaluate the probability of correct and incor rect labelling. These probabilities are then used to define alternative realizations of the structure in its damaged state through a Monte Carlo simulation. Specifically, 2,000 realizations are derived, and for each, a random damage level classification is sampled for every beam–column–joint subassembly (including undamaged ones). For each struc tural component, all possible damage levels are associated with capacity reduction factors. In this application, the λ -factors provided in the FEMA 306 report are adopted. This way, 2,000 alternative configurations of the damaged structure are generated. Each configuration is then analyzed using the SLaMA method. The resulting SLaMA-based force–displacement capacity curves are shown in Fig. 5b. The proposed approach allows for the evaluation of a range of possible capacity curves for the structure in its damaged configuration, explicitly accounting for uncertainty in the CNN-based damage classification process. Fig. 5b also compares the “SLaMA-Damaged” method with numerical re sults, highlighting a good agreement in terms of global strength; however, the SLaMA-Damage method overestimated the ultimate displacement.

4.4. Pre- and post-earthquake safety evaluation

Seismic safety evaluation is finally performed according to the spectrum-based approach reported in the Italian “Seismic-bonus” guidelines (Cosenza et al., 2018). This document defines the Italian seismic risk classification of existing buildings (from A + to G), based on the evaluation of two indexes: the “Safety Index” IS-V (equivalent to the %New Building Standard, %NBS, NZSEE (2017)) and the expected annual losses (EAL / PAM) index. In this