CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   what's the type of conjugate heat transfer method in cfx? loose or direct coumpling? (http://www.cfd-online.com/Forums/cfx/109790-whats-type-conjugate-heat-transfer-method-cfx-loose-direct-coumpling.html)

skinga281 November 26, 2012 23:15

what's the type of conjugate heat transfer method in cfx? loose or direct coumpling?
 
hello everyone!
here is my question. In a paper,it says that
"In a direct method, an identical solver is employed for both the fluid and solid blocks and they are coupled through a common wall temperature. The common wall temperature is obtained by balancing the local fluid and solid heat fluxes at the block interface. This method has been used by several researchers such as Bohn et al.(2001,2003,2005), Davis et al.(2010).
The loose coupling is performed by coupling the flow solver with a Finite Element or Boundary Element solver to obtain the temperature distribution in the solid domain and on its surface, Talya(2002), Heidmann(2003), Kassab et al(1994,2003,2009), Groce(2006), Verstraete(2007)."

So what's the type in CFX when solving conjuate heat transfer? I can't find any paper talking about that. Thank you !!! :):)

ghorrocks November 27, 2012 06:16

CFX solves each domain seperately, with a coupled solver in each domain. The exact coupling used by CFX is proprietry and therefore not fully described in the documentation. It is described in the documentation as strictly conservative and fully implicit, but I have no idea of the detail beyond that.

cdegroot November 27, 2012 11:23

My feeling is that it is direct coupling because I doubt inside of CFX there is also a finite element or boundary element method build in. It just makes sense to use the same solver. There is a bit of detail in the Theory Guide, section 15.1.1. Of note:

-At a GGI interface, the CFX Solver calculates both fluid-side and solid-side temperatures based on heat flux conservation. These values are representative of the temperature within the half-control volumes around the vertices on the interface.

Seems to me to indicate that the same solver is used and that heat flux conservation is the constraint of the coupling scheme.

skinga281 November 27, 2012 22:20

thank you,cdegroot. I have the same feeling. In fact only the energy equation exsits in the solid domain referred to the Guide. But when solving the CHT problem, I think the same solver is used for the whole filed including the flow and solid domain in each iteration. Some details about that will be very helpful!

ghorrocks November 28, 2012 07:08

CFX definitely does not have a FE or BE solver for coupling. Yes, CFX uses the same solver, but it applies it separately to each domain - you can see that in the output file because it reports the residuals of each domain separately. The question is how it couples those two domains together. There is not enough detail in the manual to answer this - heat conservation is of course applied, it would not be coupled if it wasn't :)

cdegroot November 28, 2012 10:25

You are right we don't know the details based on the documentation but based on the definitions in the first post I think we can conclude that it is "direct" not "loose".

ghorrocks November 28, 2012 17:22

Yes, definitely. The coupling is certainly tight.

brunoc November 28, 2012 17:42

Well, it is not "loose" in the sense that it uses a distinct FE or BE model to retrieve a wall temperature, but I don't think it is coupled through a "common wall temperature" either. My feeling is that it uses lagged heat fluxes to calculate next-to-the-wall temperatures at either side of the interface.

I think that could be checked by running a simple experiment with distinct two solid domains connected through a GGI interface. At the opposite side of these two domain wall temperatures would be set. This problem should converge in 1 iteration if a big enough timescale is used, unless the interface is lagged.

ghorrocks November 29, 2012 17:07

I do not think your 1 iteration experiment will be conclusive either way.

The "tightness" of coupling is hard thing to define. For example, SIMPLE has loose coupling between the p/mass equation and the momentum equations as they are solved seperately and mass conservation is done through a correction step. A coupled solver is tightly coupled because the p/mass to momentum coupling is all contained in the same equation matrix and is solved in one go by the linear solver. This is the sort of thing most people think of when they think of tight and loose coupling - not whether an intermediate solver is in the way (but that certainly would result in loose coupling :) ).


All times are GMT -4. The time now is 05:19.