Deep Learning-Enabled Supercritical Flame Simulation at Detailed Chemistry and Real-Fluid Accuracy Towards Trillion-Cell Scale
Zhuoqiang Guo, Runze Mao, Lijun Liu, Guangming Tan, Weile Jia, Zhi X.Chen

TL;DR
This paper presents a highly optimized deep learning-based supercritical flame simulation software, DeepFlame, capable of simulating trillions of cells with real-fluid accuracy, enabling practical rocket engine combustion modeling at unprecedented scales.
Contribution
The authors significantly enhance DeepFlame's computational efficiency and scalability, achieving simulations of over 600 billion cells on supercomputers, surpassing previous capabilities by three orders of magnitude.
Findings
Simulated supercritical LOX/CH4 combustion with over 600 billion cells.
Achieved 439/1186 PFlop/s performance on Sunway and Fugaku supercomputers.
Enabled practical high-fidelity rocket engine combustion simulations.
Abstract
For decades, supercritical flame simulations incorporating detailed chemistry and real-fluid transport have been limited to millions of cells, constraining the resolved spatial and temporal scales of the physical system. We optimize the supercritical flame simulation software DeepFlame -- which incorporates deep neural networks while retaining the real-fluid mechanical and chemical accuracy -- from three perspectives: parallel computing, computational efficiency, and I/O performance. Our highly optimized DeepFlame achieves supercritical liquid oxygen/methane (LOX/\ce{CH4}) turbulent combustion simulation of up to 618 and 154 billion cells with unprecedented time-to-solution, attaining 439/1186 and 187/316 PFlop/s (32.3\%/21.8\% and 37.4\%/31.8\% of the peak) in FP32/mixed-FP16 precision on Sunway (98,304 nodes) and Fugaku (73,728 nodes) supercomputers, respectively. This computational…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
