50 inlets / /50 outlets?

Steffen595 · May 16, 2016, 07:51

Hi,

I want to model a heat exchanger, 50 tubes. Is it better to have 50 inlets and outlets or connect the tubes via a manifold? I made the tubes as sweep, and the manifiolds as sweep, thats 52 solids. The tubes are in then another fluid. So now I have 2 domains and lots of interfaces (tubes to manifold both ends, domain 2nd fuid to tubes). Thats lots of trouble defining interfaces. If I was to get rid of the manifolds, I would have to define inlets/outlets each tube.

Any ideas?

cheers,

Steffen

ghorrocks · May 16, 2016, 08:19

It depends what you are trying to get out of the model. If you want to know what non-uniformity exists because of the manifold design then obviously you have to model it.

Either way this model is going to be difficult to do as you have so many paths to model. This sounds like it could be a good target for a reduced order model. If you really have to do this by 3D CFD then you will need some powerful computers and be careful with convergence tolerances (as if the convergence is poor on a few tubes is bad and all the others it is good - how will you pick that up and ensure you are converged?)

What are you trying to learn by doing this model?

Steffen595 · May 16, 2016, 08:40

I want to check heat exchange in crossflow or parallel flow. If I give each pipe 1/50 of the flow, I can skip the manifold, as I am not interested in that.This would simplify the model. I tried with auto generated, but if its too coarse, the results are not good. If the mesh is too fine, there is not enough RAM. So splitting the solids and do it in sweeps makes a better mesh, but lots of interfaces. The mesh was 60MB but the solver allocated 1.1 GB RAM and gave up. Must be the interfaces. I managed to run 150MB meshes, no problem. Did not even go into heat exchange.
Tried ICEM , it does not want to merge the meshes.
image attached

ghorrocks · May 17, 2016, 00:33

Quoting my previous post: "This sounds like it could be a good target for a reduced order model."

Why do you need to do this model with CFD? A far simpler reduced order model will tell you the effect of the difference in flow direction.

Steffen595 · May 17, 2016, 02:09

if you mean heat transfer calculations, there are lots of factors and assumptions, and at the end the result will be doubled. Also I could not find anything useful for tube bundles in parallel flow, only cross flow. The closest was flow across flat plate and then work your way up from there, so lots of assumptions there as well. So why not model up, whats been built vs some idea?
Got some ideas for merging meshes in ICEM.
The 50 inputs I could make a named selection, then set it at constant velocity. This will force same mass flow through every tube, which is a fair enough assumption, as the pressure drop through a pipe is higher than in a manifold, hence forcing pretty even flow rate through each pipe.

ghorrocks · May 17, 2016, 02:33

There are lots of factors and assumptions in 3D CFD as well. Don't assume that 3D CFD is more accurate because it makes prettier pictures

Each tube is a pipe, so 1D pipe flow equations should apply. Also the heat transferred might be able to be calculated as well with a simple ODE solver. Just some ideas, I don't really know what you are trying to do so cannot say if it is better than what you propose.

Steffen595 · May 17, 2016, 03:21

essentially, option 1 if the medium outside the tubes flows parallel to them, eventually will be turbulent, and it utilises the entire surface for heat transfer
option 2 crossflow in slalom pattern, then the outside fluid is faster, more turbulent, but uses less of the surface area. Maybe option 1 does the same heattransfer as 2 or close, its way cheaper to build.
I am not so much interested, whats going on inside the pipes. But I want to simplifiy the interdaces, that will make things a lot simpler. If there were 2 domains with no interfaces in them (because they are made from several solids), then I only get 1 interface between the tube walls and the other fluid domain.
The tubes would be sweep and the manifolds perfect cubes, sweeped. So even mesh merge in ICEM would make the mesh more complicated. If only I had more time to do that fun stuff.

edit: tubes on their own work, check for isolated regions - off, and they all the same with veolocity as inlet.

ghorrocks · May 17, 2016, 07:19

I don't see what point you are trying to make with your last post. Regardless of whether the flow is laminar or turbulent, parallel or cross flow, I would be designing this system initially with reduced order 1D models. I would only consider 3D CFD models when I was looking at optimising a system which was already pretty close. It sounds like you are still at the concept design stage and it is better to use very simple models at that stage.

Steffen595 · May 17, 2016, 18:42

how can you reduce a 3 dimensional flow pattern to 1 dimension?

I did some simplified things first with a porous domain symbolising the tube pattern to check flow patterns.

ghorrocks · May 17, 2016, 20:10

Pipe friction factors, pipe flow equations. And in many cases the 3D flow is a chamber which can be replaced with a point value.

A porous model is a reduced order model as well, you replace a complex flow field through the porous material with a simpler 3D flow with a resistance term.

Steffen595 · May 18, 2016, 01:44

the porous domain approximation was just too much simplified for heatexchange, but good enough as a starting point.

May 16, 2016, 07:51	50 inlets / /50 outlets?	#1
Steffen595 Senior Member Join Date: May 2012 Location: Melbourne Posts: 161 Rep Power: 13	Hi, I want to model a heat exchanger, 50 tubes. Is it better to have 50 inlets and outlets or connect the tubes via a manifold? I made the tubes as sweep, and the manifiolds as sweep, thats 52 solids. The tubes are in then another fluid. So now I have 2 domains and lots of interfaces (tubes to manifold both ends, domain 2nd fuid to tubes). Thats lots of trouble defining interfaces. If I was to get rid of the manifolds, I would have to define inlets/outlets each tube. Any ideas? cheers, Steffen

May 17, 2016, 03:21		#7
Steffen595 Senior Member Join Date: May 2012 Location: Melbourne Posts: 161 Rep Power: 13	essentially, option 1 if the medium outside the tubes flows parallel to them, eventually will be turbulent, and it utilises the entire surface for heat transfer option 2 crossflow in slalom pattern, then the outside fluid is faster, more turbulent, but uses less of the surface area. Maybe option 1 does the same heattransfer as 2 or close, its way cheaper to build. I am not so much interested, whats going on inside the pipes. But I want to simplifiy the interdaces, that will make things a lot simpler. If there were 2 domains with no interfaces in them (because they are made from several solids), then I only get 1 interface between the tube walls and the other fluid domain. The tubes would be sweep and the manifolds perfect cubes, sweeped. So even mesh merge in ICEM would make the mesh more complicated. If only I had more time to do that fun stuff. edit: tubes on their own work, check for isolated regions - off, and they all the same with veolocity as inlet. Last edited by Steffen595; May 17, 2016 at 06:49.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[ANSYS Meshing] How Can I choose inlets and outlets for a spherical enclosure ?	bboynido	ANSYS Meshing & Geometry	8	March 15, 2016 04:02
This case has both inlets & outlets	yossir4	FLUENT	0	April 27, 2014 14:21
supersonic inlets filling take with no outlets	pt39	ANSYS	0	January 24, 2013 03:22
About Inlets & Outlets	Floydian	Phoenics	3	January 22, 2005 03:48
more outlets or inlets?	cfxbeginer	CFX	2	November 29, 2001 23:39

May 16, 2016, 08:19		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,711 Rep Power: 143	It depends what you are trying to get out of the model. If you want to know what non-uniformity exists because of the manifold design then obviously you have to model it. Either way this model is going to be difficult to do as you have so many paths to model. This sounds like it could be a good target for a reduced order model. If you really have to do this by 3D CFD then you will need some powerful computers and be careful with convergence tolerances (as if the convergence is poor on a few tubes is bad and all the others it is good - how will you pick that up and ensure you are converged?) What are you trying to learn by doing this model?

May 17, 2016, 00:33		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,711 Rep Power: 143	Quoting my previous post: "This sounds like it could be a good target for a reduced order model." Why do you need to do this model with CFD? A far simpler reduced order model will tell you the effect of the difference in flow direction.

May 17, 2016, 02:09		#5
Steffen595 Senior Member Join Date: May 2012 Location: Melbourne Posts: 161 Rep Power: 13	if you mean heat transfer calculations, there are lots of factors and assumptions, and at the end the result will be doubled. Also I could not find anything useful for tube bundles in parallel flow, only cross flow. The closest was flow across flat plate and then work your way up from there, so lots of assumptions there as well. So why not model up, whats been built vs some idea? Got some ideas for merging meshes in ICEM. The 50 inputs I could make a named selection, then set it at constant velocity. This will force same mass flow through every tube, which is a fair enough assumption, as the pressure drop through a pipe is higher than in a manifold, hence forcing pretty even flow rate through each pipe.

May 17, 2016, 02:33		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,711 Rep Power: 143	There are lots of factors and assumptions in 3D CFD as well. Don't assume that 3D CFD is more accurate because it makes prettier pictures Each tube is a pipe, so 1D pipe flow equations should apply. Also the heat transferred might be able to be calculated as well with a simple ODE solver. Just some ideas, I don't really know what you are trying to do so cannot say if it is better than what you propose.

May 17, 2016, 07:19		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,711 Rep Power: 143	I don't see what point you are trying to make with your last post. Regardless of whether the flow is laminar or turbulent, parallel or cross flow, I would be designing this system initially with reduced order 1D models. I would only consider 3D CFD models when I was looking at optimising a system which was already pretty close. It sounds like you are still at the concept design stage and it is better to use very simple models at that stage.

May 17, 2016, 18:42		#9
Steffen595 Senior Member Join Date: May 2012 Location: Melbourne Posts: 161 Rep Power: 13	how can you reduce a 3 dimensional flow pattern to 1 dimension? I did some simplified things first with a porous domain symbolising the tube pattern to check flow patterns.

May 17, 2016, 20:10		#10
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,711 Rep Power: 143	Pipe friction factors, pipe flow equations. And in many cases the 3D flow is a chamber which can be replaced with a point value. A porous model is a reduced order model as well, you replace a complex flow field through the porous material with a simpler 3D flow with a resistance term.

May 18, 2016, 01:44		#11
Steffen595 Senior Member Join Date: May 2012 Location: Melbourne Posts: 161 Rep Power: 13	the porous domain approximation was just too much simplified for heatexchange, but good enough as a starting point.