CoFINN: Conservation Flux Informed Neural Networks for Physics Problems Governed by Conservation Laws

2026-07-05 · Source: Computer Vision and Pattern Recognition · Field: Science & Research — Engineering & Applied Sciences, Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

CoFINN (Conservation Flux Informed Neural Networks) is a physics-informed deep learning framework designed for predicting compressible flow fields governed by conservation laws. Unlike traditional data-driven convolutional neural networks (CNNs) that focus on pixel-wise similarity, CoFINN directly embeds finite-volume conservation physics into its training process. It differs from classical physics-informed methods by adopting a finite-volume perspective, consistent with modern Computational Fluid Dynamics (CFD), enforcing conservation consistency via numerical flux calculations on CNN output fields interpreted as structured computational grids. Evaluated on transonic flow prediction around airfoils at M=0.7 and Re=6 * 10^6, CoFINN significantly improves aerodynamic force prediction accuracy. It reduced drag prediction error by up to 34% at extreme angles of attack and approximately 15% on average, especially in limited-data scenarios, demonstrating the conservation-based loss as an effective physical regularizer.

Key takeaway

For research scientists developing physics-informed neural networks for fluid dynamics, CoFINN offers a robust approach to enhance model accuracy and physical consistency. You should consider integrating finite-volume conservation physics directly into your CNN training, especially when dealing with complex compressible flows or limited data. This method significantly reduces prediction errors, such as drag, and maintains computational efficiency, providing a more reliable surrogate model for engineering applications.

Key insights

CoFINN integrates finite-volume conservation physics into CNN training for accurate, physically consistent compressible flow predictions.

Principles

• Embed physics directly into training.
• Finite-volume perspective enhances CFD consistency.
• Conservation loss acts as a physical regularizer.

Method

CoFINN interprets CNN output as finite-volume cells, enforcing conservation consistency through numerical flux calculations during training, unlike differential-equation residual enforcement.

In practice

• Predict transonic flow around airfoils.
• Improve drag prediction accuracy.
• Apply in limited-data regimes.

Topics

• Physics-Informed Neural Networks
• Computational Fluid Dynamics
• Conservation Laws
• Compressible Flow
• Aerodynamic Force Prediction
• Neural Network Surrogates

Best for: AI Scientist, Research Scientist

Related on AIssential

• Differential Equations V — Physics Informed Neural Networks (PINNs) — Physics Informed Neural Networks embed physical laws into neural networks for fast, accurate, and physically consistent simulations.
• Learning Interface Breakup: A Geometry-Conditioned Latent Surrogate for Spray Formation — AMR refinement structure can serve as a compact, learnable representation for geometry-conditioned surrogate modeling of transient two-phase flows.
• Physics-informed convolutional neural networks for fluid flow through porous media — Physics-informed CNNs accurately predict fluid flow in porous media, significantly accelerating traditional LBM simulations.
• Co-Learning Port-Hamiltonian Systems and Optimal Energy-Shaping Control — Co-learning pH system models and optimal EB-PBC policies via alternating optimization ensures stability and robustness in control.
• ShipNet: A Geometric Deep Learning Surrogate for Real-Time Ship Hydrodynamics — ShipNet uses geometric deep learning to rapidly predict ship hydrodynamics from hull geometry, achieving 550x speedup over traditional methods.
• FLAME 2023: Diving into the Future of Fluid Dynamics and Machine Learning — AI techniques are transforming simulation by integrating physics-based modeling with data-driven machine learning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.