Abstract:Modern high-order discretizations bear considerable potential for the exascale era due to their high fidelity and the high, local computational load that allows for computational efficiency in massively parallel simulations. To this end, the discontinuous Galerkin (DG) framework FLEXI was selected to demonstrate exascale readiness within the Center of Excellence for Exascale CFD (CEEC) by simulating shock buffet on a three-dimensional wing segment at transsonic flight conditions. This paper summarizes the recent progress made to enable the simulation of this challenging exascale problem. For this, it is first demonstrated that FLEXI scales excellently to over 500 000 CPU cores on HAWK at the HLRS. To tackle the considerable resolution requirements near the wall, a novel wall model is proposed that takes compressibility effects into account and yields decent results for the simulation of a NACA 64A-110 airfoil. To address the shocks in the domain, a finite-volume-based shock capturing method was implemented in FLEXI, which is validated here using the simulation of a linear compressor cascade at supersonic flow conditions, where the method is demonstrated to yield efficient, robust and accurate results. Lastly, we present the TensorFlow-Fortran-Binding (TFFB) as an easy-to-use library to deploy trained machine learning models in Fortran solvers such as FLEXI.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges faced in implementing large - scale computational fluid dynamics (CFD) simulations in modern high - performance computing (HPC) systems. Specifically, the paper focuses on demonstrating the ability to perform efficient CFD simulations in the upcoming exascale computing era by using the discontinuous Galerkin (DG) solver FLEXI. The paper mainly addresses the following key issues: 1. **High parallel efficiency**: The paper first shows that FLEXI can scale to more than 500,000 CPU cores on the HAWK supercomputer and performs well in both strong - scaling and weak - scaling tests, indicating that FLEXI has the ability to operate efficiently in a large - scale parallel environment. 2. **Near - wall resolution requirements**: To meet the high - resolution requirements in the near - wall region in high - Reynolds - number flows, the paper proposes a new wall model that takes into account compression effects and verifies its effectiveness through flow simulations of the NACA 64A - 110 airfoil. 3. **Shock - capturing**: For shock problems in the flow field, the paper implements a shock - capturing method based on the finite - volume method in FLEXI and verifies the effectiveness of this method through simulations of a linear compressor blade row under supersonic conditions, proving that this method can provide efficient, robust, and accurate results. 4. **Machine - learning model integration**: With the development of scientific machine learning (ML), how to efficiently integrate promising ML models into high - performance computing simulations has become an open research question. To this end, the paper introduces TensorFlow - Fortran - Binding (TFFB), an easy - to - use library for deploying trained machine - learning models in Fortran solvers such as FLEXI, and demonstrates its performance and scalability in a large - scale computing environment. In summary, this paper aims to promote the application of FLEXI in the exascale computing environment, especially its efficient simulation ability for complex fluid - dynamics problems, by developing and implementing the above four key technology modules. These technology modules not only improve the computational performance and accuracy of FLEXI but also lay the foundation for future larger - scale CFD simulations.

Towards Exascale CFD Simulations Using the Discontinuous Galerkin Solver FLEXI

GALÆXI: Solving complex compressible flows with high-order discontinuous Galerkin methods on accelerator-based systems

A framework for high-fidelity particle tracking on massively parallel systems

Exascale Computational Fluid Dynamics in Heterogeneous Systems

CFD Based Simulation of Elastic Aircraft in Maneuvering Flight

Towards a Scalable Hierarchical High-order CFD Solver

Efficient and affordable high order, high fidelity Large Eddy Simulations for industrial level applications

A Three-Dimensional Fictitious Domain Method for the Simulation of Fluid-Structure Interactions

An Efficient Partitioned Framework to Couple Arbitrary Lagrangian-Eulerian and Meshless Vector Form Intrinsic Finite Element Methods for Fluid-Structure Interaction Problems with Deformable Structures

A Fluid-Structure Interaction Solver for Compressible Flows with Applications in Blast Loading on Thin Elastic Structures

LES/DNS Fluid-Structure Interaction Simulation of Non-Linear Slender Structures in Nektar++ Framework

Neko: A modern, portable, and scalable framework for high-fidelity computational fluid dynamics

FLEKS: A flexible particle-in-cell code for multi-scale plasma simulations

FAST HIGH ORDER LARGE EDDY SIMULATIONS ON MANY CORE COMPUTING SYSTEMS FOR TURBOMACHINERY

Two-level dynamic load-balanced p-adaptive discontinuous Galerkin methods for compressible CFD simulations

nekCRF: A next generation high-order reactive low Mach flow solver for direct numerical simulations

Efficient Finite Element Analysis/Computational Fluid Dynamics Thermal Coupling for Engineering Applications

Developing a Parallel Density-Based Implicit Solver with Mesh Deformation in OpenFOAM.

XFLUIDS: A SYCL-based unified cross-architecture heterogeneous simulation solver for compressible reacting flows

JAX-Fluids 2.0: Towards HPC for Differentiable CFD of Compressible Two-phase Flows