The turbulence appeared patchy in the Q-criterion contour plot

[ChengKunxiang]

Hi Everyone!

Whilst performing a shock train simulation, I incorporated a synthetic turbulence plugin and noticed that the turbulence appeared patchy in the Q-criterion contour plot like this:

This is my 3D mesh, which mesh nodes is 616000:

Here is its slice, which seems normal:

My configure is as follows:

[backend]
precision = double
rank-allocator = linear

[backend-cuda]
device-id = local-rank
mpi-type = standard
[backend-openmp]
cc = gcc

[constants]
gamma = 1.4
mu = 1.45766e-5
;nu = 3.53e-5
Pr = 0.72
rhom = 0.41351
;Ubulk = 598.68
Tref = 273.15 ; [K]
Ts = 110.4 ; [K]
Tw = 300.0 ; [K]
U_ref = 598.68 ; [m/s]
cp = 1004.703 ; gamma * rgas / (gamma - 1.0)
cpTref =230401
cpTs = 187359.12

[solver]
system = navier-stokes
order = 2
;order = 4
viscosity-correction = sutherland
shock-capturing = entropy-filter
anti-alias = none ;flux

[solver-entropy-filter]
d-min = 1e-6
p-min = 1e-6
e-tol = 1e-4
niters = 12
formulation = linearised

[soln-filter]
nsteps = 10
alpha = 36.0
order = 8
cutoff = 1

[solver-time-integrator]
formulation = std
scheme = tvd-rk3
tstart = 0.000
tend = 2.0e-04
dt = 1e-9
controller = none

[solver-interfaces]
riemann-solver = rusanov
ldg-beta = 0.5
ldg-tau = 0.1

[solver-interfaces-line]
flux-pts = gauss-legendre
quad-deg = 6
quad-pts = gauss-legendre

[solver-interfaces-quad]
flux-pts = gauss-legendre
quad-deg = 6
quad-pts = gauss-legendre

[solver-interfaces-tri]
flux-pts = gauss-legendre
quad-deg = 6
quad-pts = gauss-legendre

[solver-elements-hex]
soln-pts = gauss-legendre
quad-deg = 6
quad-pts = gauss-legendre ;alpha-opt

[soln-bcs-sym]
type = slp-adia-wall

[soln-bcs-wall]
type = no-slp-adia-wall

[soln-bcs-wall_front]
type = slp-adia-wall

[soln-bcs-wall_cowl]
type = no-slp-adia-wall

[soln-bcs-inlet]
type = sup-in-fa
rho = 0.41351
u = 598.68
v = 0.0
w = 0.0
p = 26500.0

[soln-bcs-outlet]
type = sup-out-fn

[soln-ics]
rho = 0.41351
u = 598.68
v = 0.0
w = 0.0
p = 26500.0

[solver-plugin-turbulence]
turbulence-intensity = 3
turbulence-length-scale = 0.0003
sigma = 0.8
avg-mach = 2
avg-u = 598.68
avg-rho = 0.41351
y-dim = 0.005
z-dim = 0.012
centre = [-0.065, 0.0025, 0.006]
rot-axis = [0.0, 0.0, 1.0]
rot-angle = 0.0

I also tried another test for a tube, and the main flow velocity is 1m/s.It looks good:

Maybe synthetic turbulence plugin is only suitable for low-velocity flow？I’ll appreciated it if someone can answer my question.

Regards!

[fdw]

Have you collected any statistics downstream of where the turbulence is being injected (for example the TI so you can compare against expectations)?

Regards, Freddie.

[ChengKunxiang]

Actually I didi it. Here is TI vs time plot when I set TI = 10:

The yellow point is that probe point.

I tried a lot of variable TI and turbulence-length-scale, but I got the same thing: *flaky * turbulence. If you want I can give you more details of it:

Normally, we should see vortices such as horseshoe vortices, rather than this kind of chaotic/disordered vortices.

[fdw]

The TI looks about right. How do the Reynolds stresses look? Turbulence is fundamentally a statistical process and needs to be analysed as such; as in the generated turbulence is appropriate so long as it has the correct statistical properties.

Regards, Freddie.

[ChengKunxiang]

Thanks for your suggestion. But I don’t know how to get the Reynolds stresses. I noticed [soln-plugin-fluidforce-name].Is this plugin designed for the Reynolds stresses or not?And how about its configuration?

Regards, Cheng.

[fdw]

You can compute the Reynolds stresses yourself and see if they look sensible.

Regards, Freddie.

[ChengKunxiang]

I understand, thanks! Here is Reynolds stresses:

After calculation, I got the average of Reynolds stresses:
R_xx = 2.78, R_yy = 4.25, R_zz = 4.42, R_xy = 0.66, R_xz = -0.05. R_yz = -0.56.
Is this something you want?

[fdw]

This is a question only you can answer. Presumably you are looking to reproduce the results of an experimental campaign with a turbulent inflow. As such, you need to see if the statistical properties of the turbulence align with the experiment or not. If everything matches you should be fine, otherwise you’ll need to switch to a different approach.

Regards, Freddie.

[ChengKunxiang]

Thanks for your reply. I will do more research on my project. If I get new progress, I will tell you immediately.

Regards, Cheng.

[ChengKunxiang]

Now I know shat wrong it is. I found something interesting in Pyfr’s document:

Upon opening an exported file in ParaView one should always run the Clean to Grid filter. This will eliminate duplicate vertices along the faces and edges between elements that arise as a consequence of the discontinuous nature of the FR approach. For best results it is recommended to set the Point Data Weighting Strategy to Average by Number. Running this filter will not only result in cleaner visuals but will also improve the performance of ParaView.

So I followed it and did some configures. It works!

Before

Now

Hope my post will help someone, and thanks for @fdw reply!