Abstract:Recent advances in the application of physics-informed learning into the field of fluid mechanics have been predominantly grounded in the Newtonian framework, primarly leveraging Navier-Stokes Equation or one of its various derivative to train a neural network. Here, we propose an alternative approach based on variational methods. The proposed approach uses the principle of minimum pressure gradient combined with the continuity constraint to train a neural network and predict the flow field in incompressible fluids. We describe the underlying principles of the proposed approach, then use a demonstrative example to illustrate its implementation and show that it reduces the computational time per training epoch when compared to the conventional approach.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve some key challenges faced when applying Physics - Informed Learning in computational fluid dynamics. Specifically, it attempts to improve the traditional Navier - Stokes - equation - (NSE) - based methods by introducing a new variational - method - based framework. The following are the main objectives and innovations of this paper: 1. **Reduce computational cost**: - Traditional methods such as PINN (Physics - Informed Neural Networks) need to consider both the velocity field and the pressure field when dealing with incompressible fluid problems. This increases computational complexity and training time. - This paper proposes a new framework - PMPG - PINN (Physics - Informed Neural Network based on the Principle of Minimum Pressure Gradient), which reduces the number of unknown variables and the number of derivative calculations by removing the pressure gradient term, thereby reducing the time per training round. 2. **Introduce a new variational principle**: - The author introduces the **Principle of Minimum Pressure Gradient (PMPG)**, which is a new method different from the traditional Newtonian framework. PMPG regards the pressure gradient as a constraint force to ensure that the continuity condition is satisfied. - This new principle enables the use of more efficient optimization methods, such as the Augmented Lagrangian Method (ALM), to balance the objective function and constraints. 3. **Improve computational efficiency**: - By eliminating the direct calculation of the pressure gradient and reducing the number of output variables, PMPG - PINN significantly improves computational efficiency. Experimental results show that compared with the traditional PINN method, the training time per round of PMPG - PINN is reduced by about 12%. 4. **Verify the effectiveness of the new method**: - The paper verifies the effectiveness of PMPG - PINN through the classic "lid - driven cavity" problem. The results show that PMPG - PINN not only has a significant improvement in computational efficiency but also can obtain similar results to the traditional PINN method. ### Formula summary - **Navier - Stokes equations**: \[ \frac{\partial u}{\partial t}+(u\cdot\nabla)u = -\nabla p+\nu\nabla^{2}u \] \[ \nabla\cdot u = 0 \] - **Objective function of the Principle of Minimum Pressure Gradient (PMPG)**: \[ S=\frac{1}{2}\int_{\Omega}\rho\left(\frac{\partial u}{\partial t}+(u\cdot\nabla)u-\nu\Delta u\right)^{2}dx \] - **Loss function of PMPG - PINN**: \[ L(\theta)=S+\frac{\mu_{c}}{2}L_{c} \] where, \[ S = \frac{A}{NB}\sum_{i}J_{i} \] \[ J=\frac{1}{2}\rho\left(\frac{\partial u}{\partial t}+(u\cdot\nabla)u-\nu\Delta u\right)^{2} \] - **Loss function with Lagrange multipliers**: \[ L(\theta)=S+\frac{\mu_{c}}{2}\sum_{i}g_{i}(x)^{2}-\frac{1}{NB}\sum_{i}\lambda_{i}^{k}

The Principle of Minimum Pressure Gradient: An Alternative Basis for Physics-Informed Learning of Incompressible Fluid Mechanics

The principle of minimum pressure gradient: An alternative basis for physics-informed learning of incompressible fluid mechanics

Minimizing Nature's Cost: Exploring Data-Free Physics-Informed Neural Network Solvers for Fluid Mechanics Applications

A Variational Computational-based Framework for Unsteady Incompressible Flows

What Does Nature Minimize In Every Incompressible Flow?

Learning Incompressible Fluid Dynamics from Scratch -- Towards Fast, Differentiable Fluid Models that Generalize

Physics-Informed Neural Networks for Parametric Compressible Euler Equations

Inferring Parameters and Reconstruction of Two-dimensional Turbulent Flows with Physics-informed Neural Networks

FEM-based Neural Networks for Solving Incompressible Fluid Flows and Related Inverse Problems

Inferring incompressible two-phase flow fields from the interface motion using physics-informed neural networks

Enhancing physics informed neural networks for solving Navier–Stokes equations

Physics-informed Autoencoders for Lyapunov-stable Fluid Flow Prediction

Learning the solution operator of two-dimensional incompressible Navier-Stokes equations using physics-aware convolutional neural networks

Physics-driven Learning of the Steady Navier-Stokes Equations using Deep Convolutional Neural Networks

Validation and parameterization of a novel physics-constrained neural dynamics model applied to turbulent fluid flow

Learning solutions of parametric Navier-Stokes with physics-informed neural networks

A fully-differentiable compressible high-order computational fluid dynamics solver

Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations

On Physics-Informed Deep Learning for Solving Navier-Stokes Equations