CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Any experience of PowerFlow from Exa (

Jen April 1, 2003 03:17

Any experience of PowerFlow from Exa
Based on my research, Lattice Boltzmann method does have more nature approach near wall region and also has more potential solving complex geometries. However, it does not mean PowerFlow is the best candidate. Any opinion?

Jonas Larsson April 1, 2003 04:25

Re: Any experience of PowerFlow from Exa
I'm not sure what you base your conlcusions about Lattice Boltzmann methods on (references?). In any case, 99% of all CFD work is done with other types of methods. As far as I know Exa is the only commercial Lattice Boltzmann CFD code, and it has been so for a long time. I have no idea if PowerFlow is any good. I think that it is mainly used for external car-aero problems.

andy April 1, 2003 04:29

Re: Any experience of PowerFlow from Exa
>>Lattice Boltzmann method does have more nature approach near wall region

Why? Do the Navier-Stokes equations become invalid for some reason.

>> and also has more potential solving complex geometries

Why? and how does this square(?) with requiring a hexagonal lattice to avoid bias.

Concerning PowerFlow as the best candidate, this link was posted when PowerFlow last came up: "Evaluation of PowerFLOW for aerodynamic applications"

I believe there is a, not wholly convincing, response to this report on the exa web site.

My opinion is that deriving something useful from the Boltzman equation in one or two difficult areas of CFD may well be possible. However, the little I have seen so far is not very convincing for engineering purposes. Perhaps I am wrong and I would certainly be interested any concrete experience others may have.

Jen April 1, 2003 11:27

Re: Any experience of PowerFlow from Exa
What my basis is from the benchmark tests results from IBM research report. Navier-Stokes is a subset of conservation equations which makes particle physics easier. Meshing is totally different from conventional meshing. Their cubic elment can be cut through by boundary surface into solid and fluid parts with different shape. Noted only fluid particles can move. I am not in office. I do not have detailed in hand. Thanks for the links. I beleve This is good, (not too many people know that) and that is why you can not hear too much about this.

Jen April 1, 2003 11:38

Wall functions are awkwards
You still don't know if your wall functions are working after you changed complicate geometries. All I can say it probably work. How about give your geometry a big body force? Everything becomes unsure.

Jen April 2, 2003 02:58

LBM does not have to solve differential equ.
God creates the turbulent flows without using wall functions, He uses the Laws of conservations, law of .... He doesn't know the so-called Navier Stokes equs. Numerical errors from diff. equ. mostly are from the limits of human or computers. However, PowerFlow may not be mature anyway. Thank you Andy and Jonas!

Free_and_SIMPLE April 3, 2003 05:02

Re: LBM does not have to solve differential equ.
Dear Jen,

"LBM does not have to solve differential equ."

This is definitly not true.

Using the LBM you are solving the so called Lattice Boltzmann Equation (LBE), which is a discretized form of the discrete Boltzmann equation (the collision operator is usually modelled by a BGK model for small deviations from the Maxwellian).

Something special on LBM is that you are solving the differential equation explicit in time and due to the orthgonal, equidistant lattice you end up with a kind of Lagrangian method (contrary to Eulerian ones). But this is a very special case, which has its origin in the special discretization. You are still solving a differential equation for "the transport of some particle distribution function/probability density function".

Recent developments allow implicit time discretization, unstructured grids, etc within the LBM framework.

Regarding wall functions: If you don't have a grid which is fine enough near the wall, you should use it even in LBM. In principle I see no difference in the problems regarding near wall treatment in LBM and NS-solvers.

Regards, F & S

Jen April 3, 2003 13:57

LBM does not have to solve diff. CFD way
Good points. My mistakes. Thanks!

Jen April 4, 2003 18:13

PowerFlow's improvement on version 3.3
I got their response. However, under my investigations a month ago, Powerflow 3.2 got ~30% underestimate of eddy viscosity for turbulent flow over a 2-D backstep benchmark test, you know the voxtex shedding will be stronger as in the previous paper.

All times are GMT -4. The time now is 14:40.