Conjugate heat transfer
 

When modeling conjugate heat transfer problems with periodic boundary conditions, what is the best available CFD software?

Kanghua Chen

Re: Conjugate heat transfer
 

(1). There is no logic in this question. You need to re-phrase it. It is hard to understand. (2). "when modeling" means when "someone" is modeling something. You need to identify this "someone" first. Becasue this "someone" can be anyone, and the answer can be very different. (3)"conjugate heat transfer problems with periodic boundary conditions" means that the issue is on the "periodic boundary conditions". Actually, periodic boundary condition is one way to save the memory and computing resources. It is not necessary to use it. The periodic boundary condition is very simple, it simply says that identical values must appear at the identical locations (through translation or rotation). (4). "the best available CFD software?" is focused on the "best" issue. I guess, it also need to be defined first. Because it could mean "the most expensive", "the least expensive", "the fastest", or "the most accurate one to specify the periodic boundary condition",.... (5). I can't find the question, can you?

Re: Conjugate heat transfer
 

I believe most of general-purpose CFD packages should be able to handle conjugate heat transfer with periodic boundaries, as long as the flow and heat transfer has a periodicity in it. Could you be a little bit more specific about the situation ?

Re: Conjugate heat transfer
 

A lot of CFD software packages can successfully solve fluid flow and heat transfer problems with periodic boundary conditions applied. However, I found the unrealistic results when including the solid heat transfer region in the model. By the way, I use Fluent 5.1.

Re: Conjugate heat transfer
 

Like a heatsink with many fins. Because I don't want to model the whole heatsink, I pick a unit with just two half-fins. Symmetric boundary conditions were applied on two side surfaces and periodic boundary conditions were applied on inlet and outlet surfaces. In this model, both fluid and solid regions are included. I couldn't get a right temperature distribution by using Fluent 5.1 as a solver.

Re: Conjugate heat transfer
 

(1). I can not comment on the solution accuracy related to any software, since I don't have access to the exact software you are using. (2). But, I can make some suggestions, if it is related to the proper modeling and specification of boundary conditions. (3). For the heat sink problem, can you try to describe the computational domain more clearly? Is it a 2-D or 3-D problem?

Re: Conjugate heat transfer
 

Which software you use is not the only issue. It is also important how you use it - How did your grid look? Which turbulence model did you use? What boundary conditions? Which convergence criteria? Which solver? Which physical models? Did you have correct inlet turbulence level and length scale?

Heat transfer is one of the most difficult things to get correct and all these small details are very important and can ruin your results if you don't make the right choice. Please give more details about your case and how you tried to solve it. Fluent is not bad at these kinds of problems.

Re: Conjugate heat transfer
 

Surely you don't impose a periodic temperature field?

I'm surprised you're not using Icepak which seems particulary suited to academic electronic thermal applications.

Robin.

Re: Conjugate heat transfer
 

When you said "I couldn't get a right temperature distribution", do you mean that the result doesn't make sense at all or just that the result is not quite satisfactory ?

If the result doesn't make sense at all or seems to you completely counter-intuitive, please take some moments in a quite place and revisit the BCs, material properties, etc. to see if they really represent the physical situation in question.

If the result just fall shy of satisfying you (e.g., temperature or heat flux peaks are overpredicted or underpredicted by 30 %, the location of peak temperature or heat flux is too much shifted toward the fin, etc.), there are many possible reasons for that, including mesh resolution, inappropriate choice of discretization scheme, turbulence model, convergence, among many many others.

I have no doubt you can get help from many seasoned CFD practitioners on this web by posting questions. However, the best way to resolve this problem in a timely manner is talk to your support engineer. If you have any problem getting help from the support engineer, you can contact me.

Re: Conjugate heat transfer
 

Could you explain more about not imposing a periodic temperature field? Is that not a physical boundary condition?

Re: Conjugate heat transfer
 

I think the result doesn't make sense. I specified the Upstream Bulk Temperature as 300K in periodicity condition panel. However, the result showed that the temperature in some fluid regions was far below 300K.

Re: Conjugate heat transfer
 

It's a 3D problem like a fluid flowing through a bunch of tube banks.

Re: Conjugate heat transfer
 

I believe it does not matter which solver is used. But I really have problem to solve the conjugate heat transfer problem with the periodic boundary conditions applied.

Re: Conjugate heat transfer
 

Sung-Eun is correct the best way to solve your problem is to contact someone at Fluent who can answer your questions and guide your analysis

Re: Conjugate heat transfer
 

(1). Without the detailed configuration, I can only make some general suggestions. (2). Try to model it in 2-D if possible, check the result to see whether you still have the problem. (3). Try to isolate the conduction part by not solving the solid region, in this way, you can solve only the fluid region with specified wall temperature conditions. Make sure that the fluid problem is well defined. (4). Try different boundary conditions, such as specifying the inlet, exit conditions, to see whether you are getting any solutions at all. This is very important, because a weak boundary condition sometimes is hard to give the right answer. (5). By the way, "periodic boundary condition" simply says that the values are unknown and must be part of the solution,and should appear at the periodic boundary locations. Symmetry condition means the variable itself is unknown but the normal gradient is zero. So, somewhere along the line, you need some boundary conditions which will be strong enough to give you a flow field and a temperature field. (6). Only the vendor knows whether the code is programmed properly. Perhaps, you can ask for an example case which is similar to your problem and has a reasonable temperature distribution. (I spend most of the time to figure out which part of the solution is really useful to me as a user.)

Re: Conjugate heat transfer
 

Very good points!! Thanks!!!