Neural-network-based assessment of slenderness effects for dual system buildings in Mexico City

The Mexican regulations establish precise criteria for the design of buildings, especially about seismic design. To achieve acceptable levels of safety the seismic design spectra must be adjusted considering variables such as the building scale, ductility, hyperstability, specific irregularities, among others. This research focuses on examining risk increment due to the slenderness irregularity in dual systems with frame-wall systems according to the Mexico City building regulations 2017 (RCCDMX17) and its Complementary Technical Standards (NTC17). A large number of simulated buildings realizations with 11, 15 and 20 levels were extensively analyzed using neural networks (multi-layer feed forward, backpropagation) to investigate reliability trends associated with slenderness effects in dual structural systems designed according to the Mexican building regulations (RCCDMX17). These large-scale parametric evaluations of reliability levels in the multi-degree-of-freedom structural systems were performed under a set of seismic actions that would occur in the most unfavorable soil conditions. Based on selected variables that globally represent the analyzed responses (soil conditions, structural parameters, and earthquake motions), the neural model output corresponds to structural reliability expressed through the Cornell β index. Using the obtained neural-β values, the study identifies reliability trends associated with increasing slenderness in dual-system buildings subjected to seismic demands representative of Mexico City soft-soil conditions. The numerical findings suggest that the interaction between geometry, dynamic properties, and material characteristics plays an important role in the reliability behavior of slender structures under severe seismic conditions. Within the analyzed parameter domain, the analyzed archetypes suggest a tendency toward lower reliability for configurations approaching λ ≈ 3 However, this observation should be interpreted as an exploratory reliability-based trend rather than as a generalized or normative slenderness limit.