Abstract:Data-driven methods demonstrate considerable potential for accelerating the inherently expensive computational fluid dynamics (CFD) solvers. Nevertheless, pure machine-learning surrogate models face challenges in ensuring physical consistency and scaling up to address real-world problems. This study presents a versatile and scalable hybrid methodology, combining CFD and machine learning, to accelerate long-term incompressible fluid flow simulations without compromising accuracy. A neural network was trained offline using simulated data of various two-dimensional transient buoyant plume flows. The objective was to leverage local features to predict the temporal changes in the pressure field in comparable scenarios. Due to cell-level predictions, the methodology was successfully applied to diverse geometries without additional training. Pressure estimates were employed as initial values to accelerate the pressure-velocity coupling procedure. The results demonstrated an average improvement of 94% in the initial guess for solving the Poisson equation. The first pressure corrector acceleration reached a mean factor of 3, depending on the iterative solver employed. Our work reveals that machine learning estimates at the cell level can enhance the efficiency of CFD iterative linear solvers while maintaining accuracy. Although the scalability of the methodology to more complex cases has yet to be demonstrated, this study underscores the prospective value of domain-specific hybrid solvers for CFD.

What problem does this paper attempt to address?

### Problems Addressed by the Paper The paper aims to address the issue of long computation times in simulating transient buoyant plumes in Computational Fluid Dynamics (CFD). Specifically, the researchers propose a hybrid method that combines machine learning with traditional CFD techniques to accelerate long-term incompressible fluid flow simulations without sacrificing accuracy. #### Main Objectives - **Accelerate Simulations**: Speed up the process of solving the Poisson equation by using machine learning to predict the local pressure field. - **Maintain Accuracy**: Ensure that the initial guesses provided by machine learning predictions in the iterative linear solver improve convergence speed while maintaining the physical consistency of CFD simulations. - **Scalability**: Develop a general and scalable method that can be applied to different geometries and initial conditions without the need to retrain the model. #### Method Overview - **Data-Driven Predictions**: Train a neural network offline using simulation data from various 2D transient buoyant plume flows to predict the temporal evolution of local features. - **Local Predictions**: Make predictions at the cell level, allowing the method to be applied to different geometries without additional training. - **Physical Consistency**: Use hand-crafted features based on physical quantities to learn pressure corrections rather than absolute pressure values, thereby improving the model's generalization ability and data efficiency. Through this approach, the researchers demonstrate that applying machine learning estimates at the cell level can significantly enhance the efficiency of CFD iterative linear solvers while maintaining high accuracy. Although the scalability of this method in complex scenarios remains to be validated, this work highlights the potential value of domain-specific hybrid solvers for CFD.

Coupling Machine Learning Local Predictions with a Computational Fluid Dynamics Solver to Accelerate Transient Buoyant Plume Simulations

Machine learning accelerated computational fluid dynamics

Performance and accuracy assessments of an incompressible fluid solver coupled with a deep Convolutional Neural Network

Improving CFD simulations by local machine-learned correction

Stacked Generative Machine Learning Models for Fast Approximations of Steady-State Navier-Stokes Equations

A Multifidelity Machine Learning Based Semi-Lagrangian Finite Volume Scheme for Linear Transport Equations and the Nonlinear Vlasov–Poisson System

Coarse-Grid Computational Fluid Dynamic (CG-CFD) Error Prediction using Machine Learning

Surrogate Modeling of Aerodynamic Simulations for Multiple Operating Conditions Using Machine Learning

Hybrid quantum physics-informed neural networks for simulating computational fluid dynamics in complex shapes

Residual-based physics-informed transfer learning: A hybrid method for accelerating long-term CFD simulations via deep learning

Graph and convolutional neural network coupling with a high-performance large-eddy simulation solver

Predictions of steady and unsteady flows using machine-learned surrogate models

Machine-learning-assisted Blending of Data-Driven Turbulence Models

Machine-Learning Enhanced Predictors for Accelerated Convergence of Partitioned Fluid-Structure Interaction Simulations

In Situ Framework for Coupling Simulation and Machine Learning with Application to CFD

Enhancing computational fluid dynamics with machine learning

A multi-fidelity machine learning based semi-Lagrangian finite volume scheme for linear transport equations and the nonlinear Vlasov-Poisson system

High-resolution fluid–particle interactions: a machine learning approach

Combining Differentiable PDE Solvers and Graph Neural Networks for Fluid Flow Prediction

Machine learning for full spatiotemporal acceleration of gas-particle flow simulations