Generative-AI and LLM powered seismic design and optimization скачать в хорошем качестве

Generative-AI and LLM powered seismic design and optimization

Presentation at 1st International Conference on Safety and Resilience of Infrastructure The existing seismic design methods of building structures are inefficient and rely heavily on engineers' experience. Therefore, this study proposes a seismic design and optimization method driven by generative AI and large language models. Based on generative AI, the intelligent design algorithm has acquired the ability to generate seismic design schemes for building structures by learning from many existing design drawings and comprehensively considering the constraints of architectural layout, design conditions, mechanical principles, and empirical rules. Through deep learning-enabled computational models and evaluation functions, the intelligent optimization algorithm can rapidly optimize and continuously improve seismic performance. The design algorithm, integrating intelligent generation and optimization, can efficiently enhance the safety and economy of the generated schemes and overcome the bottlenecks of limited and low-quality training data. Additionally, by utilizing large language models as controllers, a conversational, interactive design method is constructed to achieve an integrated design process for concepts, architectures, and structures. Case studies indicate that the structural schemes generated by intelligent design are comparable to those of human experts and generally meet the seismic requirements at the design stage of building structures, significantly improving design efficiency. The developed intelligent design platform has been used by over a hundred design and research institutions for thousands of engineering designs.