Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Automating CFD Simulations with Local Large Language Models

Yang, Ge LU (2026) MVKM05 20261
Department of Energy Sciences
Abstract
This thesis develops a locally deployed LLM-assisted system for automating CFD simulations from natural-language physical modelling tasks, including OpenFOAM case setup generation, solver execution, error diagnosis, and iterative correction. Unlike many recent LLM-driven CFD agents that mainly rely on frontier cloud models, this work investigates whether open-weight local LLMs with stronger data privacy, internet independence, and economic attractiveness can support practical CFD automation. The proposed framework combines a knowledge retrieval layer constructed from OpenFOAM tutorials and documentation with an agentic workflow consisting of five functional roles. To improve reliability under the relatively weaker reasoning, limited... (More)
This thesis develops a locally deployed LLM-assisted system for automating CFD simulations from natural-language physical modelling tasks, including OpenFOAM case setup generation, solver execution, error diagnosis, and iterative correction. Unlike many recent LLM-driven CFD agents that mainly rely on frontier cloud models, this work investigates whether open-weight local LLMs with stronger data privacy, internet independence, and economic attractiveness can support practical CFD automation. The proposed framework combines a knowledge retrieval layer constructed from OpenFOAM tutorials and documentation with an agentic workflow consisting of five functional roles. To improve reliability under the relatively weaker reasoning, limited knowledge capacity, and less stable instruction-following ability of local LLMs, a series of robustness-oriented strategies such as hybrid retrieval, constrained prompts, cross-file dependency support, and self-correction loops are incorporated, shifting part of the reasoning burden from the model to the system design. The system is evaluated on representative and industrially motivated physical modelling tasks covering laminar incompressible flow, buoyancy-driven heat transfer, wall-bounded internal turbulence, and external aerodynamics over a Volvo EX90 geometry with multiple STL surfaces. The results show that locally deployed LLMs can generate executable and physically meaningful OpenFOAM cases with system-level support and iterative refinement. The study demonstrates the practical feasibility of local LLM-assisted CFD automation, while also indicating future directions in quantitative validation and CFD accuracy control for production-grade applications. (Less)
Popular Abstract
Numerical modelling of physical phenomena involving fluid flow and heat transfer has a wide range of industrial applications. However, successful simulations using Computational Fluid Dynamics (CFD) depend on substantial domain expertise. This thesis develops a locally deployed large language model (LLM)-assisted system for automating CFD simulations from natural-language descriptions. Open-weight local LLMs are attractive in industrial settings for stronger data privacy and cost efficiency, but their comparatively weaker capabilities pose challenges. The proposed framework combines a knowledge retrieval layer built on OpenFOAM documentation with a role-specialized agentic workflow that performs case setup generation, solver execution,... (More)
Numerical modelling of physical phenomena involving fluid flow and heat transfer has a wide range of industrial applications. However, successful simulations using Computational Fluid Dynamics (CFD) depend on substantial domain expertise. This thesis develops a locally deployed large language model (LLM)-assisted system for automating CFD simulations from natural-language descriptions. Open-weight local LLMs are attractive in industrial settings for stronger data privacy and cost efficiency, but their comparatively weaker capabilities pose challenges. The proposed framework combines a knowledge retrieval layer built on OpenFOAM documentation with a role-specialized agentic workflow that performs case setup generation, solver execution, error diagnosis, and iterative correction. Robustness-oriented strategies such as hybrid retrieval, constrained prompts, and cross-file dependency support are introduced to improve reliability. The developed system is evaluated on representative and industrially motivated physical modelling tasks covering laminar incompressible flow, buoyancy-driven heat transfer, internal turbulence, and external aerodynamics over a Volvo EX90 geometry with multiple STL surfaces. The results show that locally deployed LLMs can generate executable and physically meaningful CFD cases with system-level support and iterative refinement, demonstrating the practical feasibility of local LLM-assisted CFD automation, while also indicating future directions in quantitative validation and accuracy control. (Less)
Please use this url to cite or link to this publication:
author
Yang, Ge LU
supervisor
organization
course
MVKM05 20261
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Computational Fluid Dynamics, OpenFOAM, Local Large Language Models, Retrieval-Augmented Generation, CFD Automation
report number
ISRN: LUTMDN/TMHP-26/5697-SE
ISSN
0282-1990
language
English
id
9238654
date added to LUP
2026-06-16 13:35:13
date last changed
2026-06-16 13:35:13
@misc{9238654,
  abstract     = {{This thesis develops a locally deployed LLM-assisted system for automating CFD simulations from natural-language physical modelling tasks, including OpenFOAM case setup generation, solver execution, error diagnosis, and iterative correction. Unlike many recent LLM-driven CFD agents that mainly rely on frontier cloud models, this work investigates whether open-weight local LLMs with stronger data privacy, internet independence, and economic attractiveness can support practical CFD automation. The proposed framework combines a knowledge retrieval layer constructed from OpenFOAM tutorials and documentation with an agentic workflow consisting of five functional roles. To improve reliability under the relatively weaker reasoning, limited knowledge capacity, and less stable instruction-following ability of local LLMs, a series of robustness-oriented strategies such as hybrid retrieval, constrained prompts, cross-file dependency support, and self-correction loops are incorporated, shifting part of the reasoning burden from the model to the system design. The system is evaluated on representative and industrially motivated physical modelling tasks covering laminar incompressible flow, buoyancy-driven heat transfer, wall-bounded internal turbulence, and external aerodynamics over a Volvo EX90 geometry with multiple STL surfaces. The results show that locally deployed LLMs can generate executable and physically meaningful OpenFOAM cases with system-level support and iterative refinement. The study demonstrates the practical feasibility of local LLM-assisted CFD automation, while also indicating future directions in quantitative validation and CFD accuracy control for production-grade applications.}},
  author       = {{Yang, Ge}},
  issn         = {{0282-1990}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Automating CFD Simulations with Local Large Language Models}},
  year         = {{2026}},
}