PSI - Issue 84

Marianna Crognale et al. / Procedia Structural Integrity 84 (2026) 898–905

904

Fig. 5. Global pushover response of the viaduct (base shear versus control displacement) for a representative degraded material state (C2–S0).

4. Seismic Assessment with ML-Informed Interpretation of Global Response Fig. 6 presents the global pushover response from the conventional fiber-based finite-element model, along with an ML-informed interpretation of section-level degradation effects along the same loading path. The global equilibrium response, expressed in terms of base shear versus control displacement, is computed exclusively through finite-element analysis and remains unchanged by the ML framework. The ML surrogates are employed in post-processing to quantify the evolution of effective stiffness and strength degradation at the critical pier section, expressed through the degradation indicators and . While the pushover curve itself exhibits the expected elastic–inelastic transition and post-yield behavior, the ML-inferred indicators reveal that stiffness degradation develops rapidly with increasing deformation demand, whereas strength degradation evolves more gradually and stabilizes at larger displacements. This ML-informed interpretation enables a deterioration-aware assessment of global seismic performance without altering the nonlinear solution process, highlighting how section-level material degradation manifests along the global response trajectory.

Fig. 6. Comparison between a nonlinear finite-element model with distributed plasticity, together with an ML-informed interpretation of global behavior. While the global capacity curve is obtained exclusively from the nonlinear FEM analysis, the ML surrogates provide section-level stiffness and strength degradation indicators ( , ) evaluated along the same loading path, without altering the nonlinear solution process.