An analytical database for reinforced concrete structural walls based on an experimentally validated ML-integrated modeling framework

Abstract

Experimental datasets of reinforced concrete (RC) structural walls remain limited in both size and parameter diversity, leaving important regions of the design space underrepresented and constraining the development of reliable predictive models. To address these limitations, this study develops an analytical database through nonlinear cyclic simulations using a previously experimentally validated machine learning-integrated modeling framework implemented in OpenSeesPy. The database systematically expands underrepresented combinations of axial load ratio, concrete compressive strength, and reinforcement detailing. A hybrid deterministic–LHS parametric sampling strategy combines engineering-informed reinforcement configurations with Latin Hypercube Sampling (LHS) of concrete compressive strength and axial load ratio, while maintaining the experimentally validated reference specimen geometry, shear-span ratio, fixed-base boundary condition, and cyclic loading protocol. Concrete compressive strength and axial load ratio vary from 20 to 60 MPa and from 5% to 45%, respectively, while the investigated reinforcement ratios span 0.54–3.50% for boundary longitudinal reinforcement, 0.40–3.20% for boundary transverse reinforcement, 0.15–1.20% for vertical web reinforcement, and 0.05–0.60% for horizontal web reinforcement. A total of 620 nonlinear cyclic simulations are performed based on the experimentally validated wall specimen. The resulting database includes complete hysteretic responses together with key engineering response parameters, including maximum lateral strength, yield displacement, ultimate displacement capacity, ductility ratio, and FEMA P-2208-based failure mode classification. The results show that axial load ratio exhibits the strongest association with displacement ductility within the investigated parameter domain, with increasing axial compression leading to significant reductions in deformation capacity and response variability. In contrast, reinforcement detailing primarily affects ductility within low- and moderate-axial-load regimes, while its influence diminishes under high axial compression demands. The proposed analytical database provides a physically consistent and experimentally grounded resource for performance-based seismic assessment, interpretable machine learning, and future data-driven modeling of RC structural wall behavior within the investigated applicability domain.