SLIDE 1
1
Use of Nektar++ and OpenFOAM for the simulation of bluff-body flows Hongyi Jiang
School of Engineering The University of Western Australia
SLIDE 2 D
2
Introduction
y (transverse) , ,
y y x x z z
u u u u u u u y z z x x y Reynolds number Vorticity (a measure of the rotation of the velocity field) / Re UD U
SLIDE 3
3
Re = 220, Mode A (3D) Re = 100, 2D
A nominally 2D cylinder → A 3D flow
Introduction
SLIDE 4
4
Re = 220, Mode A Re = 100, 2D
Introduction
SLIDE 5
5
(a) Re = 45 (2D steady) (b) Re = 185 (2D unsteady) (c) Re = 194 (Mode A*) (d) Re = 270 (Mode B) (e) Re = 400 (Mode B) (f) Re = 1000 (Mode B)
Introduction
SLIDE 6
6
Advantage 1
A weak but persistent Mode A flow (OpenFOAM)
SLIDE 7
7
Advantage 1
(Williamson, 1996. Three-dimensional wake transition. J. Fluid Mech. 328, 345–407.)
Experimental evidence: End of story?
SLIDE 8
8
Advantage 1
Mesh skewness → Three dimensionality (→ Stable mode A)
(a) Re = 100, uz = ±2×10-5 (b) Re = 100, uz = ±2×10-6
SLIDE 9 9
Advantage 1
100 125 150 175 200 10
10
10
10
10
10
Ez Re Standard 3D mesh Modified 3D mesh Beyond Recr
Mesh skewness↓ → Three dimensionality↓
SLIDE 10
10
Advantage 1
Nektar++ results: Transient growth and then decay to 2D.
Nektar++ (Fourier) Transient initial disturbance OpenFOAM (Replication) Persistent disturbance (uncontrolled) Experiment Real-life flow with ambient disturbance
(Jiang et al., 2016. Phys. Fluids 28, 104103.)
SLIDE 11
11
Advantage 2
Vorticity field at Re = 300 calculated with 2D DNS. Primary Two-layered Secondary Location for the onset of the secondary vortex street (Re = 200): Vorobieff et al. (2002) x/D = 60 Kumar and Mittal (2012) x/D = 100 Thompson et al. (2014) Not within x/D = 280 But there has been no mesh independence check on the far-wake mesh.
SLIDE 12
12
Advantage 2
200 400 600 800 1000 40 80 120 160 200 240 400 38 39 40 NektarNp=6 The second transition The first transition
x/D Re
OpenFOAM1 OpenFOAM2 NektarNp=5 NektarNp=3 NektarNp=2
~ 2000 elements for x/D < 2 and 44,000 elements for x/D = 2 – 400
For extremely mesh-sensitive cases, Nektar++ shows better convergence.
(Jiang et al., 2019. J. Fluid Mech. 867, 691–722.)