Dear all,

I have a set of transport equations for scalars with some extra terms are involved apart from convection and diffusion. One of these extra terms is something as a second order derivative for the scalars in a virual space (not (xyz)space).

Is it possible to add a step such as above by solving for the n-th timestep this second order derivative and then add it to the source term (with usrsrc), so that it will be accounted for in the (n+1)time step?

Has anyone of you experience with this?

Looking forward to lots of replies,

Bram

PS I use CFX 4.3

1) How does your virtual space depend on xyz? Not at all? Then, what is it?

2) I think it is possible. I have to check whether the second derivative is directly available.

3) I have no experience with virtual spaces, but we can always try to get you on the road.

4) Why don't you use CFX4.4??

Veel succes, Astrid

Thanks a lot for your answer!

It might be a bit abstract but I'll try to explain the problem as good as possible. It's a combustion problem and the 'virtual space' is mixture fraction space [0..1]. That space itself does not depend on xyz. The second order derivative of the scalar can be considered as a mixing/diffusion parameter of the scalar in that space.

The mixture fraction space is discretised into a number of points. This gives intervals, and for each interval a scalar transport equation is being solved throughout the flow domain. However this transport equation contains this second order derivative in mixture fraction space as well. And that's where the bottleneck is located at the moment....

I found something in the manual which allows the user to modify the coefficients of the differencing scheme and it could be that that opens a way to solve this extra term. (The actual values of the scalar for each interval are known and the discretisation of mixture fraction space is known as well, so somehow it should be possible to calculate coeffients which could be added to the matrix..)

So indeed it might be possible. What do you think?

Unfortunately CFX 4.4 is not available to me. Would it make a huge difference?

Bram

PS Beetje veel tekst en waarschijnlijk niet echt lekker duidelijk uitgelegd, maar ik ben benieuwd!

1) You don't need CFX 4.4

2) It sound like you have implemented the mixture fraction space succesfully. Correct? How did you implement it? As a UserScalar?

3) Indeed you can modify your discretisation scheme using USRFDS. Perhaps you need the second derivative of your mixture fraction. Therefore you might use the utility routine GRADS(.....). Then you have to use it twice to get a second order derivatie.

Not sure if it will work. Therefore I need to know more of your (typisch universiteits) project.

Astrid

Indeed, mixture fraction space is implemented succesfully (vingers crossed). I stored it as an array in permanent workspace. The second derivate I just calculate with a finite difference calculation. The outcome of that I add to the solution by putting it into the USRSRC for each specific scalar.

I skipped the idea of modifying the discretisation scheme, as the matrix being used there is one in physical space.

Het is inderdaad een typisch universiteits project. Het doel is om gedeeltelijk voorgemengde methaanvlammen te kunnen modelleren. (reactants prediction, NOx prediction, temperatures etc.).

Thanks for your suggestions again!

Bram

PS Ik heb respect voor je 'posting' tijden. Of zit je aan de andere kant van onze planeet?