Frequency-structured cross-fidelity errors in multiscale flow reconstruction
- 1 September 2026
- journal article
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (9)
Abstract
Coarse-grid, filtered, and downsampled flow fields often retain the dominant large-scale organization of a high-fidelity (HF) solution while losing localized gradients, interfaces, shear layers, corner vortices, and curved-wall details. We, therefore, examine cross-fidelity errors from a fluid-mechanics perspective and formulate HF recovery as a frequency-decoupled, multi-fidelity, physics-constrained reconstruction task. Fourier-enhanced multi-fidelity neural networks (FE-MFNN) combine a shared low-frequency backbone, a Fourier-enhanced cross-fidelity residual branch, adaptive Fourier-basis screening, and progressive physics-constrained training based on low-fidelity (LF) data, boundary information, and residuals of the governing partial differential equation (PDE). Direct LF–HF spectral diagnostics show that the Burgers and Allen–Cahn residuals contain substantially larger high-band energy fractions than their corresponding HF fields under the same 75% cumulative-energy cutoff. Computational fluid dynamics diagnostics reveal weak co-location between cavity velocity-residual energy and HF vorticity but a stronger association between cylinder pressure-residual energy and HF pressure-gradient structures. Across scalar PDE benchmarks, lid-driven cavity flow, and subsonic compressible flow around a cylinder, FE-MFNN improves high-frequency reconstruction in shock-like, interface-dominated, and shear-layer-dominated cases and remains competitive for curved-boundary compressible flow. These results support frequency-decoupled reconstruction as a physically interpretable strategy in the tested multiscale-flow settings. The reported frequency structure is an empirical diagnostic of the archived coarse-resolution and separately exported LF datasets, not a universal physical law.Keywords
Funding Information
- the Nation Key Laboratory of Aerospace Physics in Fluids (KT-APF-2025-018)
This publication has 39 references indexed in Scilit:
- Research and progress on thermochemical nonequilibrium in hypersonic vehicleActa Astronautica, 2026
- Multi-Objective Loss Balancing for Physics-Informed Deep LearningComputer Methods in Applied Mechanics and Engineering, 2025
- Finite basis physics-informed neural networks (FBPINNs): a scalable domain decomposition approach for solving differential equationsAdvances in Computational Mathematics, 2023
- On the eigenvector bias of Fourier feature networks: From regression to solving multi-scale PDEs with physics-informed neural networksComputer Methods in Applied Mechanics and Engineering, 2021
- PPINN: Parareal physics-informed neural network for time-dependent PDEsComputer Methods in Applied Mechanics and Engineering, 2020
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equationsJournal of Computational Physics, 2019
- Compact finite difference schemes with spectral-like resolutionJournal of Computational Physics, 1992
- Spectral and finite difference solutions of the Burgers equationComputers & Fluids, 1986
- High-Re solutions for incompressible flow using the Navier-Stokes equations and a multigrid methodJournal of Computational Physics, 1982
- A microscopic theory for antiphase boundary motion and its application to antiphase domain coarseningActa Metallurgica, 1979