Periodic Reporting for period 2 - AEROSIMULAT (High-performance aerodynamics and aeroacoustics simulations of the new generation of high-speed gas turbines via high-order Galerkin methods)
Okres sprawozdawczy: 2022-09-01 do 2023-08-31
Jet noise has remained the dominant source of aircraft noise since jet airliners were introduced over half a century ago. Further reductions in jet noise are therefore needed to meet the ambitious noise reduction goals set forward by the EU in its “FlightPath 2050” report.
This project addressed the above challenges by performing high-fidelity, massively parallel, aeroacoustic simulations of isolated and installed jets at realistic Reynolds and Mach numbers using the open-source spectral/hp element framework Nektar++. In particular, a discontinuous Galerkin method was used and the highest predicted frequency corresponded to a non-dimensional Strouhal number of 8.
A Reynolds number of 500,000 and Mach number of 0.6 were considered along with different nozzle configurations. The simulation results were validated against experimental data obtained at the University of Southampton. In general, the near-field flow statistics obtained from the simulation were found to agree well with those obtained from experiments. In terms of far-field acoustics, the simulation correctly predicted the shape of the spectra and over-predicted noise levels by 1-2dB. This could be due to a small under prediction of the turbulence levels that was found at the nozzle exit.
This project addressed the above challenges by performing high-fidelity, massively parallel, aeroacoustic simulations of isolated and installed jets at realistic Reynolds and Mach numbers using the open-source spectral/hp element framework Nektar++. In particular, a discontinuous Galerkin method was used and the highest predicted frequency corresponded to a non-dimensional Strouhal number of 8.
A Reynolds number of 500,000 and Mach number of 0.6 were considered along with different nozzle configurations. The simulation results were validated against experimental data obtained at the University of Southampton. In general, the near-field flow statistics obtained from the simulation were found to agree well with those obtained from experiments. In terms of far-field acoustics, the simulation correctly predicted the shape of the spectra and over-predicted noise levels by 1-2dB. This could be due to a small under prediction of the turbulence levels that was found at the nozzle exit.
The work carried out focused on high-performance aerodynamic and aeroacoustic simulation of compressible jets via high-order Galerkin methods. In particular, a high-order discontinuous Galerkin method available within the compressible flow solver in Nektar++ open-source software platform was used.
In March 2021 the fellow became official observer of the H2020 project DJINN (Decrease Jet INstallation Noise) (Grant agreement ID: 861438) which funds her supervisor, Prof. Sherwin, and colleagues, Dr. Lindblad and Dr. Cantwell at Imperial College London. The AEROSIMULAT project has benefited enormously from its partnership with DJINN. It provided collaboration and input from leading industrial and academic partners, as well as access to representative cases of study and experimental data to validate the simulation results. The fellow participated in two DJINN consortium meetings and presented results in one of them.
Given the very high computational needs of the AEROSIMULAT project, a lot of work was carried out preparing proposals to obtain high-performance computing (HPC) allocations. In order to prepare the proposals, the fellow requested two PRACE Preparatory Access Type A allocations. She obtained access to three different PRACE Tier-0 systems: MareNostrum4 from February to July 2021 under PRACE Preparatory Access 2010PA5654, SuperMUC-NG from October 2021 to April 2022 and JUWELS Cluster from October to December 2021 under PRACE Preparatory Access A 2010PA6122. During these preparatory allocations she could study and demonstrate the scalability and performance of Nektar++ on the target architectures up to a maximum of 36,864 cores.
A total of five proposals were prepared and submitted to highly competitive calls for HPC allocations: three RES (Red Española de Supercomputación) calls and two PRACE Project Access calls. The fellow was the principal investigator of all the proposals. She succeeded for one of the RES calls (ID: IM-2021-3-0004) (3.5 million core hours, Nov 2021 – Feb 2022, 4 months) and for the PRACE Project Access Call 24 (proposal No. 2021250043) (31 million core hours, April 2022 – March 2023, 1 year). The computing time that was awarded through these calls was crucial for the fulfillment of the objectives of the project.
The fellow and her colleagues at Imperial College London improved the compressible flow solver in Nektar++ and implemented the coupling between Nektar++ and the ANTARES library. They enabled Nektar++ to perform high- fidelity large-eddy simulations of compressible flows at high Reynolds and Mach numbers and to predict jet noise. The improvements done were made available as open-source in Nektar++’s repository in GitLab. The fellow also participated in the completion of Nektar++’s user and developer guides, available through Nektar++’s webpage and GitLab, respectively.
The fellow gave two seminars to the Computational Fluid Dynamics (CFD) group at BCAM, one in July 2022 and the other in January 2023, in which she presented the progress on the performed jet noise simulations using Nektar++. In June 2022 the fellow was also included in the MACROPISTAS project (Manufacturing of Curved Objects via Path-design of custom-shaped tools) (Id. PID2019- 104488RB-I00) funded by the Spanish Government and led at BCAM by Dr. Michael Barton. She worked for one of the work packages of the project, on the simulation of the flow through a 2D profile of a turbine blade extruded to 3D, using Nektar++.
In December 2021 the fellow attended the workshop “Industrially Oriented Jet Noise Reduction Technologies” organized by DJINN. In December 2021 she made a short visit at Imperial College London (from the 7th to the 17th December) and attended the “Nektar++ Workshop 2021”. In November 2022 she participated to the ENODISE/DJINN/INVENTOR joint workshop: “Advanced porous treatment and metamaterials for engine and airframe noise mitigation”. In May 2022 she participated to the “42nd VI-HPS Tuning Workshop” on performance analysis tools which was held online by the POP CoE project and available to her through the awarded PRACE Project Access allocation.
In March 2021 the fellow became official observer of the H2020 project DJINN (Decrease Jet INstallation Noise) (Grant agreement ID: 861438) which funds her supervisor, Prof. Sherwin, and colleagues, Dr. Lindblad and Dr. Cantwell at Imperial College London. The AEROSIMULAT project has benefited enormously from its partnership with DJINN. It provided collaboration and input from leading industrial and academic partners, as well as access to representative cases of study and experimental data to validate the simulation results. The fellow participated in two DJINN consortium meetings and presented results in one of them.
Given the very high computational needs of the AEROSIMULAT project, a lot of work was carried out preparing proposals to obtain high-performance computing (HPC) allocations. In order to prepare the proposals, the fellow requested two PRACE Preparatory Access Type A allocations. She obtained access to three different PRACE Tier-0 systems: MareNostrum4 from February to July 2021 under PRACE Preparatory Access 2010PA5654, SuperMUC-NG from October 2021 to April 2022 and JUWELS Cluster from October to December 2021 under PRACE Preparatory Access A 2010PA6122. During these preparatory allocations she could study and demonstrate the scalability and performance of Nektar++ on the target architectures up to a maximum of 36,864 cores.
A total of five proposals were prepared and submitted to highly competitive calls for HPC allocations: three RES (Red Española de Supercomputación) calls and two PRACE Project Access calls. The fellow was the principal investigator of all the proposals. She succeeded for one of the RES calls (ID: IM-2021-3-0004) (3.5 million core hours, Nov 2021 – Feb 2022, 4 months) and for the PRACE Project Access Call 24 (proposal No. 2021250043) (31 million core hours, April 2022 – March 2023, 1 year). The computing time that was awarded through these calls was crucial for the fulfillment of the objectives of the project.
The fellow and her colleagues at Imperial College London improved the compressible flow solver in Nektar++ and implemented the coupling between Nektar++ and the ANTARES library. They enabled Nektar++ to perform high- fidelity large-eddy simulations of compressible flows at high Reynolds and Mach numbers and to predict jet noise. The improvements done were made available as open-source in Nektar++’s repository in GitLab. The fellow also participated in the completion of Nektar++’s user and developer guides, available through Nektar++’s webpage and GitLab, respectively.
The fellow gave two seminars to the Computational Fluid Dynamics (CFD) group at BCAM, one in July 2022 and the other in January 2023, in which she presented the progress on the performed jet noise simulations using Nektar++. In June 2022 the fellow was also included in the MACROPISTAS project (Manufacturing of Curved Objects via Path-design of custom-shaped tools) (Id. PID2019- 104488RB-I00) funded by the Spanish Government and led at BCAM by Dr. Michael Barton. She worked for one of the work packages of the project, on the simulation of the flow through a 2D profile of a turbine blade extruded to 3D, using Nektar++.
In December 2021 the fellow attended the workshop “Industrially Oriented Jet Noise Reduction Technologies” organized by DJINN. In December 2021 she made a short visit at Imperial College London (from the 7th to the 17th December) and attended the “Nektar++ Workshop 2021”. In November 2022 she participated to the ENODISE/DJINN/INVENTOR joint workshop: “Advanced porous treatment and metamaterials for engine and airframe noise mitigation”. In May 2022 she participated to the “42nd VI-HPS Tuning Workshop” on performance analysis tools which was held online by the POP CoE project and available to her through the awarded PRACE Project Access allocation.
The highest frequencies predicted in most jet noise computations to date correspond to a non-dimensional Strouhal number of St=3-6. This project expanded the current state of the art by simulating jet noise up to St=8. Increasing the resolved frequency spectrum is an important goal within the aeroacoustics community since some noise reduction technologies, such as chevrons, may also increase noise at higher frequencies.
In addition to advancing the current state of the art in jet aeroacoustics, this project demonstrated that the high order discontinuous Galerkin method is suitable for predicting jet noise at high Reynolds and Mach numbers and brought high-order methods closer to industrial needs. It is important to remark that historically jet noise simulations have been solved using finite volume and finite difference schemes, which present some restrictions regarding complex geometries and scalability. New best-practice guidelines for jet aeroacoustics, tailored to high-order Galerkin methods in general and to the discontinuous Galerkin method in particular, were established in this project.
Throughout this project, Nektar++ received new functionalities and improvements, transforming it into an open-source software capable of conducting high-fidelity aeroacoustic simulations of turbulent jets via high-order Galerkin methods, which made it suitable to be used for the scientific community but also in a commercial relationship with certain industrial companies, in particular, those that develop noise reduction technologies.
In addition to advancing the current state of the art in jet aeroacoustics, this project demonstrated that the high order discontinuous Galerkin method is suitable for predicting jet noise at high Reynolds and Mach numbers and brought high-order methods closer to industrial needs. It is important to remark that historically jet noise simulations have been solved using finite volume and finite difference schemes, which present some restrictions regarding complex geometries and scalability. New best-practice guidelines for jet aeroacoustics, tailored to high-order Galerkin methods in general and to the discontinuous Galerkin method in particular, were established in this project.
Throughout this project, Nektar++ received new functionalities and improvements, transforming it into an open-source software capable of conducting high-fidelity aeroacoustic simulations of turbulent jets via high-order Galerkin methods, which made it suitable to be used for the scientific community but also in a commercial relationship with certain industrial companies, in particular, those that develop noise reduction technologies.