第18届世界地震工程会议报告视频：从基于模拟的城市减灾到基于生成式AI的抗震设计

2024年7月3日，18th World Conference on Earthquake Engineering (WCEE2024)上的学术报告《From "Simulation-driven" urban disaster mitigation to "Generative AI-powered" seismic design》。

Abstract: "Resilience" has emerged as a prominent focal point for building and urban disaster prevention. Consequently, thoroughly examining the mechanisms underlying urban disasters such as earthquakes and the evolution of resilience from both building and city scales is of utmost importance. Traditional methods are mainly limited to historical disaster data-driven empirical models. In contrast, this presentation emphasizes the physics-based earthquake simulation by introducing the nonlinear time history analysis method at both building and city scales. Moreover, the presentation will highlight a physics-based multi-hazard simulation framework. The framework, rooted in the city information model (CIM), is designed to evaluate building and community resilience in the face of multiple hazards, including earthquakes, fires, windstorms, and even the COVID-19 pandemic. The employed physics-based models have the advantage of being independent of historical disaster data, enabling their application across various regions. Furthermore, using the CIM-powered database standardizes the required data format for simulations across different hazard types and scales, thereby enhancing simulation efficiency.

Furthermore, the potential of generative artificial intelligence (AI) in structural designs with integrated seismic design knowledge is explored. Generative AI technologies such as generative adversarial networks (GANs), graph neural networks (GNNs), and diffusion models exhibit outstanding capabilities in learning high-dimensional data features and absorbing domain knowledge from the seismic design field. The technology facilitates the intelligent design of component layouts and sectional dimensions for both vertical and horizontal structural elements, taking into account seismic design criteria. It also enables intelligent prediction of material usage and seismic responses for building structures. A generative AI design software for building structures has been developed and implemented in engineering practice. Application studies indicate that the structural designs generated by AI are on par with those created by human experts, fulfilling the seismic design requirements for building structures while significantly enhancing design efficiency.