manifold to many small channel design

mullenc525 · March 24, 2011, 15:46

I'm designing a manifold for a device similar to a heat exchanger consisting of many small channels, about 300-3000 depending on the scale of the device and whether I decide upon a symmetrical design or not.

Currently I am attempting to mesh the channels with sweep then use an unstructured mesh on the manifolds. This is OK but I cannot figure out how to make the mesh connections 1:1. I am importing a solidworks assembly into CFX-mesh so I have separate 3D regions to mesh. I am unsure of which feature will force the mesh to be 1:1 and uncertain whether an assembly is the most appropriate method to import this geometry.

Also, I have a gut feel the design of the inlet side manifold is much more important than the outlet manifold. I'd like to do some work where I model only the manifold, and each channel is a 2-D region where the pressure on that face is a function of the flux through that region. How can I easily set this up for 300-3000 regions?!

mullenc525 · March 24, 2011, 16:26

It just occured to me that the gaps between the channels probably don't matter much since there are so many channels, on a big scale that surface might behave quite similar to a porous sheet. Does anybody have any opinion on modelling the surface where the channels begin as a porous sheet with an appropriate pressure drop per flux?

ghorrocks · March 24, 2011, 21:36

Exactly.

So get the resistance of the structure - either experimentally or with a CFD simulation of a single channel, then apply that to the big model as a resistance. You might consider an anisotropic resistance in this case as it looks different in the X and Y directions.

mullenc525 · March 31, 2011, 17:54

I have this working fairly well, though I have to use the non physical expert parameter 1 for porous cs discretization to make it converge - I'm not sure what impact that will have on my solution.

Now I'm wondering how to evaluate the flow distribution through the porous interface quantitatively. Is there anything like a variance function built in? How could I enter this manually in an expression?

ghorrocks · April 1, 2011, 06:06

By variance do you mean statistical variance? No, that is not built in.

mullenc525 · April 1, 2011, 13:55

Is there a better way to quantitatively evaluate flow distribution?

How can I enter something like statistical variance?

I'm not familiar with the fortran features if those are required, I know I need something like an expression:

sum over all locations of (Velocity w(some location)-areaAve(Velocity w )@Domain Interface 1 Side 1)^2

ghorrocks · April 3, 2011, 07:56

You should be able to write this as a CEL expression - and it looks like you have just about done it yourself. Are you able to get it working from there? What is your question?

mullenc525 · April 4, 2011, 16:12

I'm not sure how to sum over all locations in CEL or how to specify the arbitrary location to evaluate velocity w for any given part of that sum.

ghorrocks · April 4, 2011, 20:15

What do you mean by sum? Do you mean integrate - there are functions to do that. Do you mean sum - again, functions to do that. Or simply add a few cases together - well that is simply addition.

Read the CFX CEL reference guide for available commands and functions. I think everything you want will be in there.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
micro channel heat exchanger design	serpil kaya	Main CFD Forum	6	March 25, 2018 08:34
exhaust manifold cfd design	irfan	FLUENT	2	September 5, 2013 17:22
CFD package for small propeller turbine design	Drona Upadhyay	Main CFD Forum	13	November 17, 1999 00:20
Info: Short Course On Thermal Design of Electronic Equipment	Arnold Free	Main CFD Forum	0	August 10, 1999 11:18
CFD - Trends and Perspectives	Jonas Larsson	Main CFD Forum	16	August 7, 1998 17:27

March 24, 2011, 16:26		#2
mullenc525 New Member Join Date: Dec 2010 Posts: 26 Rep Power: 16	It just occured to me that the gaps between the channels probably don't matter much since there are so many channels, on a big scale that surface might behave quite similar to a porous sheet. Does anybody have any opinion on modelling the surface where the channels begin as a porous sheet with an appropriate pressure drop per flux?

March 24, 2011, 21:36		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Exactly. So get the resistance of the structure - either experimentally or with a CFD simulation of a single channel, then apply that to the big model as a resistance. You might consider an anisotropic resistance in this case as it looks different in the X and Y directions.

March 31, 2011, 17:54		#4
mullenc525 New Member Join Date: Dec 2010 Posts: 26 Rep Power: 16	I have this working fairly well, though I have to use the non physical expert parameter 1 for porous cs discretization to make it converge - I'm not sure what impact that will have on my solution. Now I'm wondering how to evaluate the flow distribution through the porous interface quantitatively. Is there anything like a variance function built in? How could I enter this manually in an expression?

April 1, 2011, 06:06		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	By variance do you mean statistical variance? No, that is not built in.

April 1, 2011, 13:55		#6
mullenc525 New Member Join Date: Dec 2010 Posts: 26 Rep Power: 16	Is there a better way to quantitatively evaluate flow distribution? How can I enter something like statistical variance? I'm not familiar with the fortran features if those are required, I know I need something like an expression: sum over all locations of (Velocity w(some location)-areaAve(Velocity w )@Domain Interface 1 Side 1)^2

April 3, 2011, 07:56		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	You should be able to write this as a CEL expression - and it looks like you have just about done it yourself. Are you able to get it working from there? What is your question?

April 4, 2011, 16:12		#8
mullenc525 New Member Join Date: Dec 2010 Posts: 26 Rep Power: 16	I'm not sure how to sum over all locations in CEL or how to specify the arbitrary location to evaluate velocity w for any given part of that sum.

April 4, 2011, 20:15		#9
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	What do you mean by sum? Do you mean integrate - there are functions to do that. Do you mean sum - again, functions to do that. Or simply add a few cases together - well that is simply addition. Read the CFX CEL reference guide for available commands and functions. I think everything you want will be in there.