Towards LLM-enabled autonomous combustion research: A literature-aware agent for self-corrective modeling workflows

TL;DR

The paper tackles the challenge of integrating domain literature with robust CFD execution in combustion research by introducing FlamePilot, an LLM-powered, self-corrective agent. Its single-agent architecture uses atomic tools and a literature-synthesis module to orchestrate OpenFOAM/DeepFlame workflows with transparency and verifiability, guided by physics-informed literature. Empirical results show perfect executability on FoamBench-Advanced ($M_{ ext{exec}}=1.0$) and superior performance over prior agents, along with a MILD combustion case that demonstrates literature-driven configuration, autonomous refinements, and multi-parameter convergence within hours under supervision. This work establishes a framework for AI-enabled combustion science that can accelerate discovery, improve reproducibility, and extend to additional CFD platforms through a modular, human-in-the-loop paradigm.

Abstract

The rapid evolution of large language models (LLMs) is transforming artificial intelligence into autonomous research partners, yet a critical gap persists in complex scientific domains such as combustion modeling. Here, practical AI assistance requires the seamless integration of domain literature knowledge with robust execution capabilities for expertise-intensive tools such as computational fluid dynamics (CFD) codes. To bridge this gap, we introduce FlamePilot, an LLM agent designed to empower combustion modeling research through automated and self-corrective CFD workflows. FlamePilot differentiates itself through an architecture that leverages atomic tools to ensure the robust setup and execution of complex simulations in both OpenFOAM and extended frameworks such as DeepFlame. The system is also capable of learning from scientific articles, extracting key information to guide the simulation from initial setup to optimized results. Validation on a public benchmark shows FlamePilot achieved a perfect 1.0 executability score and a 0.438 success rate, surpassing the prior best reported agent scores of 0.625 and 0.250, respectively. Furthermore, a detailed case study on Moderate or Intense Low-oxygen Dilution (MILD) combustion simulation demonstrates its efficacy as a collaborative research copilot, where FlamePilot autonomously translated a research paper into a configured simulation, conducted the simulation, post-processed the results, proposed evidence-based refinements, and managed a multi-step parameter study to convergence under minimal human intervention. By adopting a transparent and interpretable paradigm, FlamePilot establishes a foundational framework for AI-empowered combustion modeling, fostering a collaborative partnership where the agent manages workflow orchestration, freeing the researcher for high-level analysis.

Figures (3)

• Figure 1: Schematic overview of the FlamePilot agent architecture, illustrating the integration of LLM reasoning, tool-based execution, and dynamic domain knowledge comprehension for automated CFD workflows.
• Figure 2: Agent performance on FoamBench-Advanced somasekharan_cfdllmbench_2025 cases. FlamePilot (Our Work) achieves perfect executability ($M_{\text{exec}}$ = 1.0), demonstrating superior operational robustness. Performance metrics for MetaOpenFOAM chen_metaopenfoam_2025 and Foam-Agent yue_foam-agent_2025 are based on values reported in somasekharan_cfdllmbench_2025.
• Figure 3: Comparison of simulation results for the Adelaide JHC MILD burner: (a) Schematic of the 2D computational domain (not to scale); (b) Scalar profiles showing the improvement from initial simulation to final results after literature-guided optimization, demonstrating much better agreement with experimental data dally_structure_2002. The complete process from initial setup to final optimization required only a few hours, with simulation runtime being the dominant factor. The optimized results achieve accuracy comparable to established simulation results in the literature aminian_numerical_2012.