Calculating the local time scale or reading it from available expressions

honcho62 · January 13, 2016, 04:18

I am in the process of modeling vortex generators as source terms. To do this I will apply a certain momentum in some cells. This momentum is calculated with the use of the local time scale factor, defined by the paper('A tuning free body force vortex generator model by F. Wallin') as

$\tau = \frac{\delta{}x}{|u| + c}$

I'm having trouble finding a way to implement $\delta{}x$ , which is the local grid size, there does not seem to be an available CEL expression like there is for the others, magnitude of velocity or speed of sound. Does anyone have any tips?

Furthermore, I will need to integrate this constant together with the local velocity. That means I would need to a transient simulation, right? There is no point in integrating the local velocity in a steady state simulation I would assume.

ghorrocks · January 13, 2016, 06:35

Speed of sound and Magnitude of Velocity: Variables exist for these, look in the CFX documentation in the reference guide for a list of available variables.

I don't think loal mesh size is available. You could use the cube root of the volume variable (which is the control volume's volume). But I am very wary of your function as it goes to zero as the mesh gets smaller - it is inherently mesh sensitive. This does not sound physically realistic to me.

You need to integrate your function with the local velocity - what do you mean? What type of integral? Please give the equation you wish to model.

honcho62 · January 13, 2016, 06:46

Quote:

Originally Posted by ghorrocks

Speed of sound and Magnitude of Velocity: Variables exist for these, look in the CFX documentation in the reference guide for a list of available variables.

I don't think loal mesh size is available. You could use the cube root of the volume variable (which is the control volume's volume). But I am very wary of your function as it goes to zero as the mesh gets smaller - it is inherently mesh sensitive. This does not sound physically realistic to me.

You need to integrate your function with the local velocity - what do you mean? What type of integral? Please give the equation you wish to model.

Thank you very much for your reply.

I have found those two expressions, it's the local mesh size that I haven't found yet. Cube root might work, but then your elements would need to resemble a cube for it to be accurate.

The paper gives the derivative of body force

$\frac{\partial{}f}{\partial{}t} = - \lambda{}(\rho{}u_in_i)$

with $\lambda$ being equal to the $\tau^{-2}$ from the previous equation, and the others density, local velocity and the VG normal vector.

So in order to implement that the body force you need to integrate that equation. Which would mean you would have to integrate the local velocity. Something to tune of

$f = \int{} - \lambda{}(\rho{}u_in_i) dt$

ghorrocks · January 13, 2016, 07:22

Quote:

then your elements would need to resemble a cube for it to be accurate.

This is another problem with the approach - how does it handle irregular element shapes? This is another problem with functions like this.

And integrating it over time just adds to the problem. Can you explain to me how a body force can be a function of flow conditions in the past?

Regardless - to answer your question: Have a look at the units of momentum source terms. That will tell you the units of the function you are looking for and I suspect will help you. Don't forget you will probably need a momentum source term coefficient as well - these are described in the documentation.

honcho62 · January 13, 2016, 07:50

Quote:

Originally Posted by ghorrocks

This is another problem with the approach - how does it handle irregular element shapes? This is another problem with functions like this.

And integrating it over time just adds to the problem. Can you explain to me how a body force can be a function of flow conditions in the past?

Regardless - to answer your question: Have a look at the units of momentum source terms. That will tell you the units of the function you are looking for and I suspect will help you. Don't forget you will probably need a momentum source term coefficient as well - these are described in the documentation.

This paper does seem to raise more questions than answer them.

That doesn't seem physical, but I suppose the author wants to mimic a certain affect. He states the expression comes from equating the flow tangency condition to the rate of change in the normal force. But doesn't explain what the implications of it are, as you have just shown.

The units do match though, I did check those. And thanks for the tip. I have found other models, but this had one promising feature. Though I should take a more critical look at my choice now. Thanks for the help

honcho62 · January 14, 2016, 04:11

Somewhat related to this issue. Is it at all possible to obtain more geometrical information about a mesh volume (in CFX pre that is, not in the MESH environment)?

I want to check whether a line (or a point on that line) is crosses/inhabits a mesh volume. So I would need data about the grid edges but this seems unavailable, is that correct?

I have managed to make a method that checks whether grid nodes are in a volume, but the latter would be preferred.

I have added an image to illustrate my point a bit better.

ghorrocks · January 14, 2016, 05:56

You would need to do this using user fortran. But I am no expert on user fortran so cannot help you there.

January 13, 2016, 04:18	Calculating the local time scale or reading it from available expressions	#1
honcho62 Member Join Date: Jan 2016 Posts: 34 Rep Power: 10	I am in the process of modeling vortex generators as source terms. To do this I will apply a certain momentum in some cells. This momentum is calculated with the use of the local time scale factor, defined by the paper('A tuning free body force vortex generator model by F. Wallin') as $\tau = \frac{\delta{}x}{\|u\| + c}$ I'm having trouble finding a way to implement $\delta{}x$ , which is the local grid size, there does not seem to be an available CEL expression like there is for the others, magnitude of velocity or speed of sound. Does anyone have any tips? Furthermore, I will need to integrate this constant together with the local velocity. That means I would need to a transient simulation, right? There is no point in integrating the local velocity in a steady state simulation I would assume.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
AMI speed performance	danny123	OpenFOAM	21	October 24, 2020 05:13
High Courant Number @ icoFoam	Artex85	OpenFOAM Running, Solving & CFD	11	February 16, 2017 14:40
Extrusion with OpenFoam problem No. Iterations 0	Lord Kelvin	OpenFOAM Running, Solving & CFD	8	March 28, 2016 12:08
Floating point exception error	lpz_michele	OpenFOAM Running, Solving & CFD	53	October 19, 2015 03:50
Upgraded from Karmic Koala 9.10 to Lucid Lynx10.04.3	bookie56	OpenFOAM Installation	8	August 13, 2011 05:03

January 13, 2016, 06:35		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Speed of sound and Magnitude of Velocity: Variables exist for these, look in the CFX documentation in the reference guide for a list of available variables. I don't think loal mesh size is available. You could use the cube root of the volume variable (which is the control volume's volume). But I am very wary of your function as it goes to zero as the mesh gets smaller - it is inherently mesh sensitive. This does not sound physically realistic to me. You need to integrate your function with the local velocity - what do you mean? What type of integral? Please give the equation you wish to model.

January 14, 2016, 04:11		#6
honcho62 Member Join Date: Jan 2016 Posts: 34 Rep Power: 10	Somewhat related to this issue. Is it at all possible to obtain more geometrical information about a mesh volume (in CFX pre that is, not in the MESH environment)? I want to check whether a line (or a point on that line) is crosses/inhabits a mesh volume. So I would need data about the grid edges but this seems unavailable, is that correct? I have managed to make a method that checks whether grid nodes are in a volume, but the latter would be preferred. I have added an image to illustrate my point a bit better.

January 14, 2016, 05:56		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	You would need to do this using user fortran. But I am no expert on user fortran so cannot help you there.