Performance of high-order implicit large eddy simulations

Konstantinos Ritos; Ioannis W. Kokkinakis; Dimitris Drikakis

doi:10.1016/j.compfluid.2018.01.030

Performance of high-order implicit large eddy simulations

Konstantinos Ritos, Ioannis W. Kokkinakis, Dimitris Drikakis

Department of Engineering

University of Strathclyde

Research output: Contribution to journal › Article › peer-review

Abstract

The performance of parallel implicit Large Eddy Simulations (iLES) is investigated in conjunction with high-order weighted essentially non-oscillatory schemes up to 11th-order of accuracy. Simulations were performed for the Taylor Green Vortex and supersonic turbulent boundary layer flows on High Performance Computing (HPC) facilities. The present iLES are highly scalable achieving performance of approximately 93% and 68% on 1536 and 6144 cores, respectively, for simulations on a mesh of approximately 1.07 billion cells. The study also shows that high-order iLES attain accuracy similar to strict Direct Numerical Simulation (DNS) but at a reduced computational cost.

Original language	English
Pages (from-to)	307-312
Number of pages	6
Journal	Computers and Fluids
Volume	173
DOIs	https://doi.org/10.1016/j.compfluid.2018.01.030
Publication status	Published - 15 Sept 2018

Keywords

High-Order methods
iLES
Parallel computing
Turbulent flows

Access to Document

10.1016/j.compfluid.2018.01.030

Cite this

@article{d84afc0a6c564646b1344bb6f7369444,

title = "Performance of high-order implicit large eddy simulations",

abstract = "The performance of parallel implicit Large Eddy Simulations (iLES) is investigated in conjunction with high-order weighted essentially non-oscillatory schemes up to 11th-order of accuracy. Simulations were performed for the Taylor Green Vortex and supersonic turbulent boundary layer flows on High Performance Computing (HPC) facilities. The present iLES are highly scalable achieving performance of approximately 93% and 68% on 1536 and 6144 cores, respectively, for simulations on a mesh of approximately 1.07 billion cells. The study also shows that high-order iLES attain accuracy similar to strict Direct Numerical Simulation (DNS) but at a reduced computational cost.",

keywords = "High-Order methods, iLES, Parallel computing, Turbulent flows",

author = "Konstantinos Ritos and Kokkinakis, {Ioannis W.} and Dimitris Drikakis",

year = "2018",

month = sep,

day = "15",

doi = "10.1016/j.compfluid.2018.01.030",

language = "English",

volume = "173",

pages = "307--312",

journal = "Computers and Fluids",

issn = "0045-7930",

publisher = "Elsevier Ltd",

}

TY - JOUR

T1 - Performance of high-order implicit large eddy simulations

AU - Ritos, Konstantinos

AU - Kokkinakis, Ioannis W.

AU - Drikakis, Dimitris

PY - 2018/9/15

Y1 - 2018/9/15

N2 - The performance of parallel implicit Large Eddy Simulations (iLES) is investigated in conjunction with high-order weighted essentially non-oscillatory schemes up to 11th-order of accuracy. Simulations were performed for the Taylor Green Vortex and supersonic turbulent boundary layer flows on High Performance Computing (HPC) facilities. The present iLES are highly scalable achieving performance of approximately 93% and 68% on 1536 and 6144 cores, respectively, for simulations on a mesh of approximately 1.07 billion cells. The study also shows that high-order iLES attain accuracy similar to strict Direct Numerical Simulation (DNS) but at a reduced computational cost.

AB - The performance of parallel implicit Large Eddy Simulations (iLES) is investigated in conjunction with high-order weighted essentially non-oscillatory schemes up to 11th-order of accuracy. Simulations were performed for the Taylor Green Vortex and supersonic turbulent boundary layer flows on High Performance Computing (HPC) facilities. The present iLES are highly scalable achieving performance of approximately 93% and 68% on 1536 and 6144 cores, respectively, for simulations on a mesh of approximately 1.07 billion cells. The study also shows that high-order iLES attain accuracy similar to strict Direct Numerical Simulation (DNS) but at a reduced computational cost.

KW - High-Order methods

KW - iLES

KW - Parallel computing

KW - Turbulent flows

UR - http://www.scopus.com/inward/record.url?scp=85041558850&partnerID=8YFLogxK

U2 - 10.1016/j.compfluid.2018.01.030

DO - 10.1016/j.compfluid.2018.01.030

M3 - Article

AN - SCOPUS:85041558850

SN - 0045-7930

VL - 173

SP - 307

EP - 312

JO - Computers and Fluids

JF - Computers and Fluids

ER -

Performance of high-order implicit large eddy simulations

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this