Learning Forum

[John Smith]

Hello,

I am modelling a turbulence phase change problem in CFX, and I am applying some modifications to the 2-equation k-omega model via source terms. I am trying to compare the magnitude of the source/sink terms that I'm implementing in the k-equation against all other terms (e.g., advection of k, production and dissipation, etc.).

To do so, I have expanded the k-equation terms and written them as CEL expressions, using the built-in gradient and time derivative operators. To check that I have set up those expressions correctly, I summed up the k-equation terms to form a "derived" rate of change of k and compared it with the rate of change calculated by CFX (in CEL, Turbulence Kinetic Energy.Time Derivative). 

The two values do not match very well, especially in regions of high velocity gradients. I suspect this might've resulted from the numerical errors related to spatial discretization, which leads me to my questions:

1. How are the .Gradient and .Time Derivative variable operators implemented? In CFX-Pre, one can specify the spatial and temporal discretization schemes in solver control. Are these schemes used by the CEL operators as well?
2. Is there any suggestions on how to evaluate the turbulence equation terms in a less cumbersome way?

Thank you!

[rfblumen]

Variable gradients are calculated internally using tri-linear shape functions as outlined in "11.1.1 Discretization of the Governing Equations" in the CFX Theory Guide.  The interpolation is essentially second-order accurate.  Time derivatives are calculated from values from the previous time step using essentially a backward difference approach which you can verify by creating a monitor point of a given variable in a transient simulation, calculating the time derivative from this and comparing to the .Time Derivative variable.

It's not possible to provide suggestions on specific issues like this without further information.  This would be best handled through submitting a case in Ansys Support.