[mahmoudtmt]

This is Mahmoud Ali, a petroleum engineering PhD. student at Texas A&M University. I am working on Ansys Fluent to simulate a process which includes chemical species. I have two question actually, the first is related to the spatial discretization in the case of chemical species.


The gradient is for the diffusion term.
The pressure is typically for the pressure.
The momentum is for the convective term.
Now, What the user scalar 0 (chemical species) spatial discretization is used for?

Secondly, in my simulation, when I change user scalar 0 spatial discretization from First order upwind (FOU) to SOU or Quick, I had a difference in the result of more than 30%. Is that normal? Also, I get better match with experimental data for the case with FOU for the user scalar 0 spatial discretization. How could that be the case? As far as I know, SOU is more accurate than FOU.

Thanks for your time,

[LuckyTran]

Originally Posted by mahmoudtmt

The gradient is for the diffusion term.

The gradient is quite literal, it's how the gradients of all variables are calculated.

The diffusion terms are central differenced. The discretization is another step from calculating the gradient. The gradients are needed inside the discretization of terms in the governing equations. E.g. for 2nd order upwind, you need a gradient, and that gradient is calculate via the gradient scheme. It might not be obvious why the gradient calculation is not considered a "discretization," but it's because in the FVM sense, what you are discretizing is fluxes and not values.

Originally Posted by mahmoudtmt

Now, What the user scalar 0 (chemical species) spatial discretization is used for?

It's for the advective term in the species transport equation, just like momentum.

Solution change of 30% is possible.

Originally Posted by mahmoudtmt

As far as I know, SOU is more accurate than FOU.

SOU and QUICK are second order accurate but they are unstable and can have spurious oscillations. When unabated, it's possible to have non-physical results using higher order schemes. Fluent has limiters to prevent non-physical oscillations, but that doesn't mean that there aren't oscillations. It's very easy to run into these issues when you have a bad mesh.

SOU is second order accurate, that does not mean the same thing as "SOU is always more accurate."

Newton Raphson method also has a 2nd order rate of convergence, but that doesn't mean that it always converges.

SOU is second order accurate, when it is accurate. One of the requirements is you need a specific type of flow, an advection dominated one. The other requirement is that you actually calculate the correct gradient. If your gradient is biased/skewed because of a skewed mesh or bad solution, then your solution will be skewed even though the underlying technique should be more accurate.

[mahmoudtmt]

Thanks for your elaboration.
I am using a rectangular cube domain and I QC'd the mesh. So, I do not expect bad mesh problem. What could be the other reasons? And what do you mean by bad solution?

Thanks