Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
FEA vs CFD for heat transfer analysis
- Thread starter Tunalover
- Start date
- Thread starter
- #3
- Thread starter
- #5
All-
I admit that I was "fishing" for a particular answer. Is it not true that with FEA codes one must provide the heat transfer coefficients? Doesn't the user usually have to calculate these coefficients from handcalculations?
It seems that an accurate mesh isn't worth much if the boundary conditions are "a leap of faith" from the get-go.
Isn't it true that Finite Difference Method (traditional CFD) codes simply have no need for convection coefficients because they get deeper into the physics of the problem where such simplifications are unecessary?
The reason I ask these questions is that my boss uses FEA to solve thermal problems and I've been arguing that we go with a FDM code that is designed specifically for heat transfer problems.
Are there any heat xfer experts out there who can confirm these beliefs with a theoretical backbone?
Tunalover
I admit that I was "fishing" for a particular answer. Is it not true that with FEA codes one must provide the heat transfer coefficients? Doesn't the user usually have to calculate these coefficients from handcalculations?
It seems that an accurate mesh isn't worth much if the boundary conditions are "a leap of faith" from the get-go.
Isn't it true that Finite Difference Method (traditional CFD) codes simply have no need for convection coefficients because they get deeper into the physics of the problem where such simplifications are unecessary?
The reason I ask these questions is that my boss uses FEA to solve thermal problems and I've been arguing that we go with a FDM code that is designed specifically for heat transfer problems.
Are there any heat xfer experts out there who can confirm these beliefs with a theoretical backbone?
Tunalover
CFD can be solved with FD or FE, Heat Transfer can be solved with FD or FE, Structural dynamics can be solved with FD or FE...I don't see what you are getting at. The most important thing is what equations you are discretizing and how well they can model the actual physics. If you want to use a CFD code to compute heat than go ahead..but if you want to then apply the heat load to a structure to see what will happen then you will need to model the structure as well. So then you will need two codes anyway unless the CFD code also has structural mechanics capabilites. In a perfect world you would have a code which does full coupling between the fluid and structure..some codes like ANSYS can do this.
- Thread starter
- #7
The heat transfer coefficient equations are rather robust and have stood the test of time for quite a while. So long as your "hand-calcs" are appropriate for your case, the I don't see why you would have to model the fluid as well as the solid for a HT analysis.
OTOH, if your geometry or flow does not follow the restrictions on the hand-calc equations, then CFD _may_ be the way to go.
I've done things both ways, and positively demonstrated both of these principles.
OTOH, if your geometry or flow does not follow the restrictions on the hand-calc equations, then CFD _may_ be the way to go.
I've done things both ways, and positively demonstrated both of these principles.
So you are asking me whether one discretization method is better than another for problems in heat transfer? If they both are used to discretize the same set of equations they they both should converge the same answer as you refine the respective meshes....ok I know what you meant was not that but that is what you asked...
The thing I am trying to hammer home is that you keep saying FEA code..there are FEA codes for fluid dynamics as well as structural mechanics. FEA is just a discretization method. What you should be saying are structural mechanics code vs fluid dynamics code.
The thing I am trying to hammer home is that you keep saying FEA code..there are FEA codes for fluid dynamics as well as structural mechanics. FEA is just a discretization method. What you should be saying are structural mechanics code vs fluid dynamics code.
- Thread starter
- #10
Is it not a fundamental limitation that FEA codes who do heat transfer require the user to provide handbook-calculated heat transfer coefficients? It seems like a waste of an accurate method when you are forced to assign coarse boundary conditions.
Tunalover
Tunalover
I had always assumed that heat transfer coefficients are derived from empirical data and that CFD codes could only work out the Reynold's number for which you need the heat load on the boundaries, for which you needed to know the Reynold's number etc. etc. Is this wrong or can CFD codes coupled with structural models derive these heat transfer relationships?
corus
corus
- Thread starter
- #12
Corus-
The Reynolds Number, film coefficients, etc. are all conveniences created to express the behavior of fluids and heat on a piece of paper. Based on my limited exposure to them, true CFD codes utilizing FDM (or FVM) have no need for these terms; instead they discretize and solve select PDEs from the Navier-Stokes equations. These codes use more realistic boundary conditions (temperatures, heat loads, adiabatic walls, etc.). You won't find inputs in CFD codes for convection coefficients or Reynold's Numbers! Of course real problems sometimes mean even THESE boundary conditions are rough at best!
Tunalover
The Reynolds Number, film coefficients, etc. are all conveniences created to express the behavior of fluids and heat on a piece of paper. Based on my limited exposure to them, true CFD codes utilizing FDM (or FVM) have no need for these terms; instead they discretize and solve select PDEs from the Navier-Stokes equations. These codes use more realistic boundary conditions (temperatures, heat loads, adiabatic walls, etc.). You won't find inputs in CFD codes for convection coefficients or Reynold's Numbers! Of course real problems sometimes mean even THESE boundary conditions are rough at best!
Tunalover
"Is it not a fundamental limitation that FEA codes who do heat transfer require the user to provide handbook-calculated heat transfer coefficients? It seems like a waste of an accurate method when you are forced to assign coarse boundary conditions."
It may be a limitation for finite element analysis (FEA) codes which solve problems in solid mechanics but not for finite element analysis (FEA) codes which solve problems in fluid mechanics...I'm going to keep bugging you until you stop associating the word FEA with solid mechanics..grrrrrrrr!!!
It may be a limitation for finite element analysis (FEA) codes which solve problems in solid mechanics but not for finite element analysis (FEA) codes which solve problems in fluid mechanics...I'm going to keep bugging you until you stop associating the word FEA with solid mechanics..grrrrrrrr!!!
tunalover,
On the rare occasions I have used CFD codes the only boundary conditions I was permitted to use were either fixed temperatures or a fixed heat flux. These were thus assumptions made about heat reansfer into the solid body. As you say they are a rough estimate of reality.
Likewise the heat transfer coefficients derived empirically can also be an estimate but I've always found good agreement with measurements where taken to validate models, even though I think measurements are equally erroneous and are at best an estimate too.
corus
On the rare occasions I have used CFD codes the only boundary conditions I was permitted to use were either fixed temperatures or a fixed heat flux. These were thus assumptions made about heat reansfer into the solid body. As you say they are a rough estimate of reality.
Likewise the heat transfer coefficients derived empirically can also be an estimate but I've always found good agreement with measurements where taken to validate models, even though I think measurements are equally erroneous and are at best an estimate too.
corus
I agree with pja - the fundamental question is really - do you need to model the fluid to accurately capture the heat transfer, or is a model of only the solid with assumed heat transfer coefficients sufficient? The numerical methodology for performing the caluclations - FEA (Galerkin discretization), FDM, FVM, etc is irrelevant.
My answer is that if your geometry is close to the conditions under which the heat transfer coefficients are applicable, then you only need to model the solid. Other, you should model both the solid and the fluid together.
My answer is that if your geometry is close to the conditions under which the heat transfer coefficients are applicable, then you only need to model the solid. Other, you should model both the solid and the fluid together.
- Thread starter
- #16
In my last post I quoted what I was referring to..specifically your reference to needing heat transfer coefficients if FEA is used..but then you say you don't need this for CFD..that doesn't make any sense unless you are talking specifically about FEA for solid mechanics...follow now? If I use a Galerkin finite element discretization of the Navier-Stokes equations I won't need heat transfer coefficients correct? All the physics is contained in the equations I am solving and the assumptions I make. Is this not FEA?
btrueblood
Mechanical
On a more generic note,
The near-wall modeling of flow fields requires certain assumptions be made regarding turbulence. The methods (e.g. kappa-epsilon models) typically provide an averaged or equivalent viscosity, and are "tuned" to provide the best results for flow field simulation, not for heat transfer. Further, the scaling of grids in near-wall zones to acheive the best resolution of flow parameters is not the same scaling as should be used to obtain the best resolution of thermal transfer. If you really want to hurt your brain on this topic, suggest you look at NASA and AIAA archives using "combustion" and "aero heating" in your search terms.
Most heat transfer models that are anywhere near succesful at prediction use some method of empirical "anchoring"; I would trust a CFD model heat transfer number much less than its flow numbers, and that is not very much without some test data and/or good old-fashioned engineering judgement and hand-calc's. to back it up.
The near-wall modeling of flow fields requires certain assumptions be made regarding turbulence. The methods (e.g. kappa-epsilon models) typically provide an averaged or equivalent viscosity, and are "tuned" to provide the best results for flow field simulation, not for heat transfer. Further, the scaling of grids in near-wall zones to acheive the best resolution of flow parameters is not the same scaling as should be used to obtain the best resolution of thermal transfer. If you really want to hurt your brain on this topic, suggest you look at NASA and AIAA archives using "combustion" and "aero heating" in your search terms.
Most heat transfer models that are anywhere near succesful at prediction use some method of empirical "anchoring"; I would trust a CFD model heat transfer number much less than its flow numbers, and that is not very much without some test data and/or good old-fashioned engineering judgement and hand-calc's. to back it up.
- Status
Similar threads
- Question