Shift and flip invariant CNNs for predicting laminar flow properties
Open Access
- 1 March 2026
- journal article
- Published by AIP Publishing in APL Machine Learning
- Vol. 4 (1)
Abstract
The integration of machine learning into fluid dynamics has accelerated in recent years, driven by the proliferation of high-fidelity data and enhanced computational resources. Acting as efficient surrogate models for computationally intensive simulations, these data-driven approaches provide substantial benefits, particularly during the preliminary stages of design and optimization. Previous investigations have employed convolutional neural networks (CNNs) to predict thermo-fluid flow properties for a variety of channel geometries. These studies have largely relied on data augmentation techniques to handle geometric transformations. However, such augmentation strategies are often inefficient in capturing the inherent flip and shift invariances of flow channel data. In this study, we demonstrate that embedding these invariances directly into the model architecture not only enhances robustness but leads to superior performance while significantly reducing the number of parameters compared to their invariant-unaware counterparts. In particular, we introduce two novel architectures designed to alleviate the sensitivity of CNNs to periodic signal shifts and vertical flips. This approach allows the model to structurally address the geometric symmetries of the flow channel data, offering a more robust alternative to standard data augmentation.Keywords
Funding Information
- Deutsche Forschungsgemeinschaft (SPP 2331)
- Deutsche Forschungsgemeinschaft (FR 4072/3)
- Ministerium für Wissenschaft, Forschung und Kunst Baden–Württemberg (bwHPC)
- Ministerium für Wissenschaft, Forschung und Kunst Baden–Württemberg (HoreKA project AiMat)
- Federal Ministry of Education and Research (Grant No. 01DM21002A (FLAIM))
- Helmholtz Association (Core-Informatics of the Helmholtz Association)
This publication has 24 references indexed in Scilit:
- Weak baselines and reporting biases lead to overoptimism in machine learning for fluid-related partial differential equationsNature Machine Intelligence, 2024
- Data-driven nonlinear turbulent flow scaling with Buckingham Pi variablesJournal of Fluid Mechanics, 2024
- RIC-CNN: Rotation-Invariant Coordinate Convolutional Neural NetworkPattern Recognition, 2023
- Delving Deeper into Anti-Aliasing in ConvNetsInternational Journal of Computer Vision, 2022
- Enhancing computational fluid dynamics with machine learningNature Computational Science, 2022
- Image classification in frequency domain with 2SReLU: A second harmonics superposition activation functionApplied Soft Computing, 2021
- Convolutional neural networks for fluid flow analysis: toward effective metamodeling and low dimensionalizationTheoretical and Computational Fluid Dynamics, 2021
- A supervised neural network for drag prediction of arbitrary 2D shapes in laminar flows at low Reynolds numberComputers & Fluids, 2020
- Spectral-based convolutional neural network without multiple spatial-frequency domain switchingsNeurocomputing, 2019
- Super-resolution reconstruction of turbulent flows with machine learningJournal of Fluid Mechanics, 2019