SNIC
SUPR
SNIC SUPR
LES and uncertainty quantification for turbulent boundary layers
Dnr:

SNIC 2017/1-446

Type:

SNAC Medium

Principal Investigator:

Mattias Liefvendahl

Affiliation:

Uppsala universitet

Start Date:

2017-11-01

End Date:

2018-11-01

Primary Classification:

20301: Applied Mechanics

Secondary Classification:

20306: Fluid Mechanics and Acoustics

Tertiary Classification:

10105: Computational Mathematics

Allocation

Abstract

This investigation forms part of a VR-funded research project (for which I am the PI) concerned with near-wall models for LES (NWM-LES). The research project was started during the spring 2013 and is planned to end in the fall 2018, with the PhD defence of the two students recruited for the project. This investigation concerns the simulation of canonical flow problems dominated by turbulent boundary layers, and their separation. This project builds on two previous Medium Allocations (SNIC 2013/1-241 & SNIC 2014/1-281) at Uppmax, and two previous (SNIC 2015/1-310 & 2016/1-440) at PDC. The project will be concluded after the completion of the medium allocation applied for here. During the first three allocations, the overall work-cycle was evaluated for the simulations of interest. In addition to this, significant simulation campaigns were carried out on transient flows, in particular inflow conditions were investigated. During the fourth (ongoing) allocation, definite simulations were carried out to produce data and results for publications. Three journal articles (one accepted already), and a number of conference contributions, have been produced, all with SNIC-acknowledgements. The current proposal builds directly on the previous ones, and it is absolutely critical for the successful completion of the VR-project (and the two PhD:s) that it is approved. The bulk of the computations will consist of the simulation of turbulent boundary layers with/without separation, using NWR/NWM-LES. A typical simulation can use 1000 cores and run for 24hrs, but there will be large differences in the number of cores and the simulation time required. Recently, the group has started with uncertainty quantification with respect to simulation parameters. This entails a larger number of simulations, in order to map the parameter space. Post-processing has a significant requirement on internal memory and graphics since it involves the manipulation of data in three space dimensions represented on a computational grid with in the order of 5-100 million cells. A suitable workcycle for the hardware at PDC was developed during SNIC 2015/1-310.