Start

01/05/2024

End

01/05/2025

Status

In progress

PNRR | ASIG - Aero Simulations by Implicit GPGPUs Solver

Start

01/05/2024

End

01/05/2025

Status

In progress

PNRR | ASIG - Aero Simulations by Implicit GPGPUs Solver

Description of the research and impact

ASIG is an Innovation Grant under the PNRR HPC initiative (Spoke 6), focused on developing all-speed CFD solvers designed to run on Graphical Processing Units (GPUs). The project aims to establish a generalized framework for building heterogeneous GPGPU implicit CFD solvers that integrate both density-based and pressure-based methods. Topology-based algorithms will be utilized for matrix offloading, enabling the efficient distribution of the linear matrices associated with the flow transport problem across multiple GPUs. This implementation will leverage HPC architectures to simulate industrial use cases relevant to the aerospace sector (Leonardo S.p.A.), encompassing a range of flow regimes from subsonic to hypersonic.

Main contributions from DAER

A unified framework, adhering to the quality standards of the OpenFOAM Computational Fluid Dynamics (CFD) solver, for the implicit solution of flows at all speeds. This framework is designed to run seamlessly on both CPUs and GPUs. This requires:

  • Implementation of a time-accurate, implicit block-coupled solver for the Navier-Stokes equations, capable of handling flows at all speeds.
  • A novel matrix assembly method that allows the coupling of multiple equations (e.g., turbulence, species transport).
  • Generalized matrix representation enabling the integration of third-party algebraic solvers, with geometric agglomeration performed on blocks.
  • On-the-fly CPU topology-based matrix conversion (supports moving grids) using LDU ↔ CSR formats to minimize the number of instances.
  • Development of the amgxWrapper to seamlessly connect OpenFOAM with GPU-based algebraic solvers.

Hypersonic Validation Test

Validation is obtained by simulating high-enthalpy flows in hypersonic conditions on industrial geometries, demonstrating its robustness and applicability to complex aerospace scenarios.
Additional Submodeling for hypersonic simulations:

  • Implementation of models for the physico-chemical properties of partially ionized gases, including scenarios with varying degrees of thermal and chemical non-equilibrium.
  • Incorporation of the two-temperature model to account for thermal non-equilibrium effects between translational-rotational and vibrational-electronic temperatures.

Funding

Bando a cascata del PNRR HPC (Spoke 6 - MULTISCALE MODELLING & ENGINEERING APPLICATIONS)

Ente finanziatore
Agenzia Spaziale Italiana (ASI), Ministero Università e Ricerca (MUR)​

Other partners
Leonardo SPA, Sapienza Univ. Roma.

SDGs coperti:
#9 – Imprese, Innovazione e Infrastrutture

Publications

F. Piscaglia, F. Ghioldi, I. Spisso, J. Guerrero. "An Implicit Density-Based GPU Coupled Solver for Hypersonic Flow Computations in OpenFOAM". AIAA SCITECH 2025 Forum, Orlando (USA).

Doi: https://doi.org/10.2514/6.2025-1866