Application and Research of the DeepSeek‐Python/FISH Modeling Method in Coupled Multi‐Physics Simulation in Geotechnical Engineering

Le résumé fourni par la source

ABSTRACT In the rapidly evolving field of artificial intelligence, the advancement of large language models (LLMs) has fundamentally transformed the processing, generation, and application of knowledge across various disciplines. Addressing persistent challenges in intelligent coupled multi‐physics simulation in geotechnical engineering, this study introduces the innovative DeepSeek‐Python/FISH method through synergistic integration of emerging DeepSeek generative AI with the code‐driven ITASCA (Itasca Consulting Group, Inc.) platform (FLAC3D/PFC), establishing a unified paradigm that bridges cutting‐edge artificial intelligence and established computational mechanics methodologies. Through a comprehensive validation process, the effectiveness and versatility of the DeepSeek‐Python/FISH method were demonstrated across three representative case studies: fluid–solid coupled rock hydraulic fracturing, thermo–solid coupled rock thermal damage, and twin shield‐driven tunnel excavation in saturated strata. The method's ability to seamlessly translate engineering semantics into professional numerical simulations was systematically evaluated, emphasizing its strengths in cross‐modal transformation and adaptive optimization. The results indicate significant achievements, including: (1) the successful capture of dynamic feedback mechanisms between pore pressure evolution and rock deformation, (2) the effective representation of heat transfer and thermal stress development within the rock mass, and (3) the accurate modeling of the dynamic evolution of the pore water pressure field during tunnel excavation. The DeepSeek‐Python/FISH method's iterative optimization and adaptive learning mechanisms overcome cognitive constraints and normative challenges in initial code generation through cross‐modal engineering‐to‐simulation transformation, bridging domain expertise with computational mechanics while enhancing workflow efficiency, reducing operational complexity, and advancing coupled multi‐physics simulation in geotechnical engineering. Nevertheless, the method currently relies on the ITASCA software ecosystem and may be affected by potential biases in the large language model's training data, which require further optimization and expansion in future research.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.