Y+ values in ANSYS/FLOTRAN
 

My question : ( hope that someone could help )

ANSYS/FLOTRAN Manual says that "Optimum values of Y+ are between approximately 30 and 1000. If the pressure is decreasing in the direction of flow, values up to 5000 are acceptable".

I've already seen SIMPLE and CAST Programs using 30<Y+<400 values. For STAR-CD, the recommended y+ range (from the vendor) is from 30. We also used a FLUENT Model and 150~200 were OK.

So, the recommended range for ANSYS/FLOTRAN doesn't seem too high ? I've tried to see the impact of Y+ in the results of some models and this range was confirmed (i.e, results like Heat Transfer Coefficient and Loss Coeficient and other values were only really impacted above these limits).

When I show the Flotran results using Y+ (average) ~ 700 there is always someone to look at me with distrust (they are used to have lower values ...).

Could be the difference in terms of the method of the solution one of the reason to this difference ? (FEM vs. Finite Volume )

I hope to listen to a really truthful and wise advice on this. Thanks for help !

Re: Y+ values in ANSYS/FLOTRAN
 

Hey my friend !

Forget Flotran ! The criteria for Y+ of this joke is a joke !

Actually I don't believe so much in CFD results ...

Zé

Re: Y+ values in ANSYS/FLOTRAN
 

Flotran is not able to handle prisms. This means the mesh consists of pure tetras. From my point of view there must be at least one layer with same center distance for a correct y+ and boundary layer calculation. If you have a lock at an 2D Tetra mesh in the boundary layers region you can see that you always have cells with two nodes and one node connected to a wall. The calculations for the 2 node cells is clear but the approach for the 1 node cell is not reliable. This is the reason for the huge bandwidth of y+ values.

I have tried Flotran by myself and the results for pressure loss calculations are the worst I have every seen. I would recommend to use a "high quality" CFD tools like Fluent,Star-CD,CFX,Numeca.....

good luck

James

Re: Y+ values in ANSYS/FLOTRAN
 

ok. Enough Flotran bashing. I think the issue is the wording in the documentation. They are giving a range of "reasonable" values to use in determining y+ values, obviously this range is problem dependent.

Ze doesn't believe in CFD? hmm why are you reading a CFD newsgroup and worse, making a post?

James, ANSYS can certainly handle hex/tet meshes with wedges and pyramids. This is in addition to obviously handling a mapped hex mesh. So could the "worst pressure loss calculations" be due to being misinformed/trained?

Even the "highest quality" of tools will give garbage to the inexperienced!

cheers,

Derrek

Re: Y+ values in ANSYS/FLOTRAN
 

Dear Derrek,

for no cases y+ values of e.g. 1000 are acceptable. Have a look at the numerical approach for boundary layers.

And James is absolutely right. Flotran is a low quality CFD Tool in comparrison to Fluent or Star-CD.

b.g. numerical user

Re: Y+ values in ANSYS/FLOTRAN
 

Define acceptable. Can you safely say that there is no instance where you would ever have a value of say 3000 for a y+? Have you ever seen a u+ vs y+ plot, what is the range of data? I think in general problems typically range within 30-1000, but there is always an exception.

My point in all of this is wording and the ability to be "general" in your statements can be misleading.

Take your general statement that Flotran is low quality. Come on defend your statements. Low quality in what respect? Fluent and Star are stand alone CFD tools exclusively. So they will have additional specialized capabilities that a general purpose CFD code would not. Just as Flotran has substantially more specific capabilities that Star and Fluent do not.

cheers,

derrek

Re: Y+ values in ANSYS/FLOTRAN
 

First of all I would like to say that I appreciate this discussion. Thanks to all for sharing opinions. For me, the range for Y+ in Flotran (according to the Ansys manual) is really an issue and (I believe) should be more carefully treated in a Flotran new version. Actually I have made some sensitivity analysis for some simple models (2D), and (depeding the model characteritics) found out that (roughly) there is not much difference in the 30 < Y+ < 1000. For other cases the range 30 < Y+ < 400 has shown more adequated (I guess basically it is a Reynolds number function, or Nusselt number in Heat transfer problems). So, my conclusion is, the range 30< Y+ < 1000 is not valid for everything. The problem is : How (quick and easily) make this analysis for a 3D complex model (1 Million elements)? It is a tough work ! I think Flotran is ok. I do not know others codes to judge if it is better or worse. But this point should be clarified.

Thanks to all,

Franklin