[agustinvo]

Hello

my question is referred to the triggering of turbulence in LES and DNS in natural convection.

It is widely studied how to trigger it in forced convection (like sinthetic turbulence, providing momemtum sources, buoyancy terms...), but I would like to know if those methods are valid too for natural convection, or if there are other methods to do that.

In my case, I have a natural boundary layer. Turbulence is expected to develop by itself, but it seems it take so much time. I am imposing a fixed heat flux in the wall.

Thanks for your attention!

Agustin

[FMDenaro]

Originally Posted by agustinvo

Hello

my question is referred to the triggering of turbulence in LES and DNS in natural convection.

It is widely studied how to trigger it in forced convection (like sinthetic turbulence, providing momemtum sources, buoyancy terms...), but I would like to know if those methods are valid too for natural convection, or if there are other methods to do that.

In my case, I have a natural boundary layer. Turbulence is expected to develop by itself, but it seems it take so much time. I am imposing a fixed heat flux in the wall.

Thanks for your attention!

Agustin

have you tried to superimpose a noise to the initial condition?

[agustinvo]

Originally Posted by FMDenaro

have you tried to superimpose a noise to the initial condition?

I

I tried before, but since my incoming flow has a constant temperature, it dissapear quickly. However, I can try to play a bit with it.

I was thinking to use a fluctuating heat flux q_w(t) in the wall. At the end of one period, the given heat should be the same, and I start provoking fluctuations close to the wall. Up to know, in 2D and fixed value of q_w it appears a fluctuation, but when I go for a 3D, a beautiful laminar profile appear, so I don't know how to make it turbulent.

Thanks for your interest!

[FMDenaro]

Originally Posted by agustinvo

I

I tried before, but since my incoming flow has a constant temperature, it dissapear quickly. However, I can try to play a bit with it.

I was thinking to use a fluctuating heat flux q_w(t) in the wall. At the end of one period, the given heat should be the same, and I start provoking fluctuations close to the wall. Up to know, in 2D and fixed value of q_w it appears a fluctuation, but when I go for a 3D, a beautiful laminar profile appear, so I don't know how to make it turbulent.

Thanks for your interest!

you should superimpose a disturbance to the initial velocity field and let the BC.s unchanged... if the flow solution dumps the disturbance, it is likely you have too much dissipation (numerical or due to the SGS modelling)

[agustinvo]

Originally Posted by FMDenaro

you should superimpose a disturbance to the initial velocity field and let the BC.s unchanged... if the flow solution dumps the disturbance, it is likely you have too much dissipation (numerical or due to the SGS modelling)

Thanks! Should I add a disturbance in my temperature field too (since it is a buoyant driven flow)?

I am using a second order upwind for my divergence schemes, but I can move to a centered scheme, if my Peclet is not so big. This is the cause of a big dissipation, isn't it?

[davidwilcox]

Just be careful. Adding random signals would mean not satisfying the Navier stokes

[FMDenaro]

Originally Posted by davidwilcox

Just be careful. Adding random signals would mean not satisfying the Navier stokes

yes, this issue is true but:
1) the noise has a very small amplitude
2) at t=0 you compute the pressure equation with a source term that take into account the perturbed field in order to get pressure gradients that ensure the divergence-free constraint.

[davidwilcox]

How about putting point vortices at the inlet? By incorporating the Biot-Savart law, there will be an induction of velocities and you will get your fluctuations

[FMDenaro]

Originally Posted by davidwilcox

How about putting point vortices at the inlet? By incorporating the Biot-Savart law, there will be an induction of velocities and you will get your fluctuations

This would cause problems... The perturbation has just the goal to force the onset of the turbulence, but it is a numerical perturbation (has not correlations like in real turbulence) therefore the flow solution must lose memory of the initial perturbation. Usually, before to get a physical developed turbulence the code must run for some time in order to eliminate any trace of the initial condition.