open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

When selecting a horizontal and vertical turbulence model are the calculated coefficients also automatically adopted for tracers?

For constant eddy viscosity models it appears that the diffusion coefficients can be specified separately for velocities and tracers. However, I'm not sure if constant diffusion coefficients could be be applied to the tracers while adopting a different model for the velocities. I am worried that doing so may sum both the constant specified value alongside the value calculacted from the turbulence model.

Is anybody aware of what is going on here? it's not too clear to me reading the manual

Thanks,
Toby

Hello Toby,

If you use constant velocity for HORIZONTAL or VERTICAL TURBULENCE MODEL (= 1), you schematically have:
visc_vit = visc_mol_vit
visc_tra = visc_mol_tra
for each direction, with common value for x and y, and possible different one for z. visc_mol_vit and visc_mol_tra are viscosities given by the keywords COEFFICIENT FOR HORIZONTAL/VERTICAL DIFFUSION OF VELOCITIES and COEFFICIENT FOR HORIZONTAL/VERTICAL DIFFUSION OF TRACERS. If no turbulence model is chosen, the coefficients should include molecular and turbulent viscosities.

If you use e.g. k-epsilon model for HORIZONTAL or VERTICAL TURBULENCE MODEL (= 3), you schematically have:
visc_vit = C_mu*k^2/eps + visc_mol_vit
visc_tra = C_mu*k^2/eps/Pr + visc_mol_tra
for each direction, with common value for x and y, and possible different one for z. Pr is Prandtl number. visc_mol_vit is molecular viscosity given by the keywords COEFFICIENT FOR HORIZONTAL/VERTICAL DIFFUSION OF VELOCITIES and should be kept to default value (= 10^-6 m^2/s). visc_mol_tra are given by COEFFICIENT FOR HORIZONTAL/VERTICAL DIFFUSION OF TRACERS.
Thus, turbulence model is included for tracers if using k-epsilon model.

The philosphy is the same for mixing length as for k-epsilon. So is it for k-omega and Smagorinski, but with Prandtl number = 1.
Yet, it is not the case for Spalart-Allmaras where nothing is implemented for tracers (that should be wrong).

Hope this helps,

Chi-Tuan

Hi Chi-Tuan.

Thanks for the reply.

From what you've mentioned it seems that the only way to control the tracer diffusion coefficient when using an eddy viscosity model (apart from constant) is through the Prandtl number.

On this basis, am I right that the only parameter that could be used for calibration of the tracer diffusion is the Pradtl number except in the case that both constant eddy viscosity (velocity) and turbulent dispersion/diffusivity coefficients (tracers) are adopted?

Regards,
Toby

Hello Toby,

Indeed, with "variable" turbulence model, I would say Prandtl number is the single parameter to tune with tracers.

Anyway, you can try to change COEFFICIENT FOR HORIZONTAL DIFFUSION OF TRACERS and/or COEFFICIENT FOR VERTICAL DIFFUSION OF TRACERS if it improves your simulation. You can use an array as one value can be given per tracer.
It is not the classic way with "variable" turbulence model.

Do not forget you have variants of the k-epsilon model (see CSTKEP subroutine) if you use this turbulence model. Depending on what you try to model, using Yap correction and/or linear option for production may improve your computation.

Which turbulence model do you intend to use and for what kind of simulation?

Hope this helps,

Chi-Tuan

FYI, I have just pushed the possibility to use Prandtl number for k-omega and Smagorinski turbulence models on the main.

Chi-Tuan

FYI, I have just pushed the computation of tracer viscosity when using Spalart-Allmaras turbulence model on the main as nothing was done before.

Chi-Tuan

Hi Chi Tuan,

I'm performing a relatively straight forward dredge plume dispersion model. As the works are yet to be carried out there is no data available to really 'tune' the disspersion coefficients of the model - was more interested in carrying out a sensitivity anslysis.

At the moment, I'm using the Smagorisnky model for horizontal eddy viscosity. Will likely use a mixing length model for the vertical. No particuarly reason as to why though.

Regards,

Hello Toby,

With mixing length model for the vertical, you can play with Prandtl number, with the choice of mixing length model and damping function if needed.
As you have read before, with Smagorinski up to v8p3, Prandtl number is not taken into account.

You can try the k-epsilon turbulence model also for both horizontal and vertical directions

Chi-Tuan

Hello!

Interesting topic. I am also performing similar dredge plume dispersion modelling. We always use Smagorinsky in XY and Tsanis mixing length in Z. Tsanis is what gave the best results compared with calibration data w.r.t. surface current velocities. For stratification, k-epsilon was slightly better though, compared with above set-up with damping function.

It would be great if you could share some results from your sensitivity analysis regarding the sediment dispersion, as it is indeed a lot of guessing involved...

Kind regards
PL

Hi,

It's possible to get visc_vit and visc_tra as output from Telemac?
I'm interested in the diffusion of the tracers.
I try to get NAX,NAY and NAZ as an output from the model but I didn't get anything. I don't know if the variables are not set to be printed or is a bug. The output from the model indicated that I ask for these variables but there are not in the output file.

Thanks