Second derivative with cfx 5.7.1
 

Hi,i want to create a source term which depends oin a second derivative of velocity. How can I get second derivative within CFX? I can get all componenets of first derivative of velocity but I don't know how to get second derivatives of these components from first derivatives? Should I compile a second fortran subroutine with User Get var? Can anyone help me?

Re: Second derivative with cfx 5.7.1
 

I have to calculate the second derivative on velocity within CFX 5.7.1. Can anyone help me? Thank you. Alessandro

Re: Second derivative with cfx 5.7.1
 

Dear Alessandro,

Would you mind sharing the origin of your source term? If it comes from a div (coef * grad Phi) form. It is better to assemble it as a summation of fluxes around a control volume than as a source term.. In that case you do not need second derivatives at all..

Good luck, Opaque

Re: Second derivative with cfx 5.7.1
 

Hi Opaque, thank you for your answer. I'm modeling an entrained bed gasifier and I want to insert in the momentum equation a term which is neglected in CFX model. This term is the solid stress tensor

Re: Second derivative with cfx 5.7.1
 

Dear Alessando,

Have you checked ANSYS CFX 10.0? I understand that the solid pressure term (you called solid stress tensor) is already included..

In any case, the solid stress tensor has the form I mentioned:

div( coef1 * (grad U + grad^T U) + coef2 * div U)

so, there is no need for second derivatives..

Good luck, Opaque

Re: Second derivative with cfx 5.7.1
 

How can I take this term into account without second derivative evaluation? Thank you

Re: Second derivative with cfx 5.7.1
 

Dear Alessandro,

ANSYS CFX uses finite volume approach; therefore, you are taking the volume integral of the term

div( coef1 * (grad U + grad^T U) + coef2 * div U)

Using Gauss theorem

volInt (div Phi) dV = areaInt (Phi dot dA)

Then, you must sum the fluxes around a control volume.. Those fluxes are

sum around CV [coef1 * (grad U + grad^T U) + coef2 * div U) dot dA ]

Of course, you can only do this via User Fortran.. It is not trivial, but that is the "correct way" to do it..

You should contact the CFX help desk for more information on how to do such work..However, I rather upgrade (besides many other benefits that 10.0 has)..

Good luck, Opaque

Re: Second derivative with cfx 5.7.1
 

Thank you opaque. I can get solid velocity gradient through user fortran subroutines. If I refer to a summation of fluxes around a control volume then , as you already said, I don't need second derivatives at all. But a question comes, how can I insert the term in the momentum equation for the solid dispersed phase?

Re: Second derivative with cfx 5.7.1
 

Opaque, have you any suggestion how to add the terms to the momentum equation?

Re: Second derivative with cfx 5.7.1
 

Dear Alessandro,

My best advice here is to contact the CFX help desk.. This is not a trivial operation.. Here is why:

- You must create a subdomain

- Add a momemtum source term, and use User Fortran functions to specify the values

Now the User Fortran

- The source term will be called for a particular group of elements. Be careful with the ENTITY (please read theory doc + USER_CALC_INFO doc).

- Within the group of elements you must gather the Normal Area vector on integration points, velocity gradient on integration points, molecular coefficients at i.p's and multiply them as needed. For example, coef*gradU dot A_ip

- Add the fluxes to the corresponding entities as required.. For this one, you must understand how ANSYS CFX discretizes the equations.

I am still convinced that you should upgrade to ANSYS CFX 10.0.

Good luck, Opaque

Re: Second derivative with cfx 5.7.1
 

Thank you opaque. I also thik i should upgrade to cfx 10. Unfortunately we have cfx 5.7.1 and i think we cannot upgrade.. About what you saild we have already create a subdomain, a user function and routine related to a fortran subroutine which calculates the velocity gradient.. the subdomain requires a source term to be inserted specifying the different components, x,y,z components. So, we have some chances..don't you thik so?