Artificial intelligence (AI)-assisted simulation-driven earthquake-resistant design framework: Taking a strong back system as an example

Abstract

Traditional earthquake-resistant structural design considers only a limited number of factors, mainly elastic structural properties, to determine key design parameters. However, these parameters are often not optimal because they do not take into account the extensive plasticity expected in building structures during earthquakes. To address this issue, an artificial intelligence (AI)-assisted simulation-driven framework has been developed in this study. This framework can automatically output optimal design parameters while considering nonlinear structural response under strong earthquakes and a large number of input factors. The primary innovation of the proposed framework lies in the fusion and integration of nonlinear numerical simulation and AI tools for earthquake-resistant design of building structures, marking a promising trend in this field. The framework consists of two steps. In the first step, a database that consists of optimal design parameters and covers a wide range of design inputs will be created through numerical nonlinear response history analyses (NRHAs). In the second step, AI models will be created and trained based on the database to automatically output the optimal design parameters. To illustrate the basic components underlying the proposed framework, the determination of the height-wise distribution (denoted by Ψ) of the total design lateral force for a strong back system is taken as an example. A database of 1200 samples was created through NRHAs, and an artificial neural network (ANN) model was created, optimised, and trained. The developed ANN model yielded optimal Ψ with the majority of absolute errors within 1%, demonstrating the feasibility of the proposed AI-assisted simulation-driven earthquake-resistant design framework.