A Neural Implicit Flow Super-Resolution Approach for Reacting Flows
Abstract
We present a novel framework for reconstructing fine-resolution scalar fields from coarse-resolution data in 2D unsteady flames using an extension of neural implicit flow (NIF). The framework leverages the novel architecture of NIF to develop an efficient representation of scalar fields in spatio-temporal reacting flows. Different strategies are adopted within the proposed framework to establish super-resolution for multiscalar fields in combustion. They include 1) a reconfiguration of the input and output fields, 2) the use of a latent composition space based on nonlinear principal component analysis (NLPCA) to reconstruct the fine-resolution fields, and the conversion from nonlinear principal components (NLPCs) to species and temperature contours, and 3) the adoption of multiple ParameterNets in the NIF architecture, each accommodating one of the NLPCs. The framework was validated on a 2D unsteady Bunsen-stabilized lean premixed methane-air flame with forcing at the inlet. The results, using a coarsening ratio of 8 in a given direction, show that the adoption of multiple ParameterNets yields more accurate super-resolution than a single ParameterNet without any additional cost in the hyperparameters.
Related
- NeRF-Inspired Neural Fields for 3D Super-resolution of Reacting-Flows
- A conservative coupling method of sharp-interface and multi-species model for compressible reacting gas-liquid flows with phase change
- Generalised Mixing-Plane Method for Compressible Reacting-Mixture Flows in Steady Multiphysics Turbomachinery Simulations
- Acoustic coherence-aware oriented 3D photoacoustic microscopy super-resolution with noise-suppression
- Towards isotropic super-resolution with dual reflection-transmission tomographic diffractive microscopy
BibTeX
@misc{amarathunga2026neural,
title = {{A Neural Implicit Flow Super-Resolution Approach for Reacting Flows}},
author = {Dakshaka Vishwajith Amarathunga and Lanre M. Oshinowo and Ji-Woong Park and Tarek Echekki},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.27215},
eprint = {2610.27215},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.27215},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - A Neural Implicit Flow Super-Resolution Approach for Reacting Flows
AU - Dakshaka Vishwajith Amarathunga
AU - Lanre M. Oshinowo
AU - Ji-Woong Park
AU - Tarek Echekki
PY - 2026
DA - 2026/10/19
PB - arXiv
DO - 10.48550/arxiv.2610.27215
UR - https://arxiv.org/abs/2610.27215
AB - We present a novel framework for reconstructing fine-resolution scalar fields from coarse-resolution data in 2D unsteady flames using an extension of neural implicit flow (NIF). The framework leverages the novel architecture of NIF to develop an efficient representation of scalar fields in spatio-temporal reacting flows. Different strategies are adopted within the proposed framework to establish super-resolution for multiscalar fields in combustion. They include 1) a reconfiguration of the input and output fields, 2) the use of a latent composition space based on nonlinear principal component analysis (NLPCA) to reconstruct the fine-resolution fields, and the conversion from nonlinear principal components (NLPCs) to species and temperature contours, and 3) the adoption of multiple ParameterNets in the NIF architecture, each accommodating one of the NLPCs. The framework was validated on a 2D unsteady Bunsen-stabilized lean premixed methane-air flame with forcing at the inlet. The results, using a coarsening ratio of 8 in a given direction, show that the adoption of multiple ParameterNets yields more accurate super-resolution than a single ParameterNet without any additional cost in the hyperparameters.
ER -