This study presents a comprehensive framework for machine learning-enhanced risk assessment and multi-hazard optimization of aerodynamic core-tube structural systems in supertall buildings (>300 m) within seismic- and wind-sensitive regions of China’s megacities. As urban skylines continue to evolve vertically, the structural complexity and vulnerability to combined wind and seismic hazards demand advanced predictive and decision-support tools. The proposed methodology integrates computational fluid dynamics (CFD), nonlinear time-history seismic analysis, and finite element modeling (FEM) with supervised machine learning algorithms—including Gradient Boosted Trees, Support Vector Regression, and Deep Neural Networks—to predict critical structural responses such as interstory drift, peak acceleration, base shear, and torsional amplification under combined load scenarios.
Using an extensive dataset derived from parametric simulations and physical wind tunnel and shake table test data of recent Chinese supertall buildings, the ML models are trained to capture complex interdependencies between geometric parameters, aerodynamic shaping strategies, structural stiffness distributions, and hazard intensities. The framework also incorporates probabilistic risk assessment through Monte Carlo-based uncertainty propagation and Bayesian model updating for seismic fragility estimation.
Results demonstrate that the machine learning-enhanced approach improves risk prediction accuracy by over 25% compared to traditional deterministic models, while also enabling rapid design iterations and adaptive performance-based optimization. In addition, the optimized core-tube configurations showed up to 20% reductions in seismic base shear and 15% improvements in wind-induced comfort criteria. This study highlights the transformative potential of data-driven intelligence in supporting the resilient, efficient, and hazard-adaptive design of China's next generation of supertall structures.