Answers are grounded in this paper and cite their sources.
The detonation flow field exhibits high complexity under varying induced wall conditions, posing significant challenges for rapid reconstruction using intelligent methods due to extrapolation errors and numerical oscillations. This study proposes a novel Multi-Fidelity Informed Diffusion Model (MFIDM) framework, in which wall pressure is first predicted and then used as a conditional input to guide the diffusion model. Specifically, owing to its low dimensionality relative to the full flow-field parameters, wall pressure is easier to predict and strongly correlated with key flow features, making it an efficient guidance signal. A multi-fidelity data fusion strategy is adopted to enhance the accuracy of wall pressure prediction with limited high-fidelity viscous data. Coupling this informative guidance with a diffusion-based model mitigates extrapolation errors and suppresses numerical oscillations, improving robustness and generalization. Results show that MFIDM achieves higher accuracy than diffusion models without multi-fidelity fusion, preserves detonation wave structures better than residual networks under extrapolative conditions, and effectively suppresses high-frequency fluctuations. Overall, this study establishes an effective technical framework for combustor optimization, and experimental flow field reconstruction, with practical significance.
No takes yet. Share an insight, caveat, or question.
Bao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: