heat transfer coefficient parallel to pipe axis

[SteffenH]

Hello,

can I calculate heat transfer parallel to pipe axis with the same formulas as heat transfer along a flat plate? I was reading several articles and there is always just heat transfer perpendicular to pipe axis.
As the pipes are quite long, flow is turbulent, so I think, the HTC will be sufficient for what it does. And pressure drop is way smaller.

Cheers,

Steffen

 

[MikeHalloran]

I'm not quite sure where you're going.

It sounds like you might be better off with a piecewise approximation, i.e. calculate the radial heat transfer for a meter or two, estimate the temperature change of the turbulent fluid along the length,
calculate the radial heat transfer for the next meter or two, estimate the temperature change for that distance, and keep doing it until you run out of pipe.
You can probably set it up nicely in a spreadsheet and beat on the problem until you have a better handle on it.

In fact, there are already a crapload of heat exchange spreadsheets available for the download; maybe you can find one that you can hack into what you think you need.




Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

I want to compare a tube bundle heat exchanger in crossflow with one that has flow parallel to the tubes. There is lots of material for cross flow, forced convection over flat plate, perpendicular to pipe, over a sphere and so on. But nothing parallel to tube

 

[MikeHalloran]

I have a tiny amount of experience with axial flow within a jacketed tube, i.e., one large tube within a slightly larger shell, the whole assembly used as an exhaust pipe for a gas turbine powering a boat, with water pumped through the annulus so that the exterior wouldn't burn persons who fell or leaned against it.

There was some concern about temperature gradients around the periphery of the outer tube, e.g., if the flow were not forced into some semblance of uniform velocity all around, that areas with lower flow would get too hot and burn people. So we laser-cut some spacer rings with a circular array of orifices to be installed near the inlets and outlets of the jacket/shell. The pipes performed as intended, and made the customer happy, and we got paid. There was no CFD performed. We sized the orifices to produce a 1/2 psi drop at each stage at the design coolant flow.

... none of which is of great interest to you;

I just wanted to point out that even with a tube bundle within a long shell, the shell connections can't be axial, so the flow can't be assumed uniform over the transverse area, unless you force it with some sort of pierced baffle/ screen as generally described above.


Even if you do that, you will probably find density gradients and resulting velocity gradients and temperature gradients across the bundle when it exceeds some unknown minimum length, so you may need intermediate baffles to uniformly redistribute the flow, and don't forget that you need baffles or something to support the tubes to keep them from abrading each other.

Given that you need baffles to support the tubes anyway, it makes sense to turn them into partial baffles to induce crossflow, but maybe you can find a way around that.

Also don't forget the difference in thermal expansion between the tubes and the shell, which is likely to cause buckling of one or the other as the exchanger goes from dead cold to operating temperature and back. ... which is one very good argument for using u-shaped tubes with both ends anchored in the same tube sheet. If you use long straight tubes and two tube sheets, how do get the bundle into the shell? Not impossible, but it adds complexity.

There are valid reasons why heat exchangers have evolved to use a limited number of possible topologies. Exploration of some of the possible evolutionary branches happened a long, long time ago. If you look back far enough, you might find that someone has already been where you plan to go. ... but if their branch turned out to be a dead end, they had little incentive to record their adventure. It might be great fun to explore someone's library of engineering logbooks from a hundred years ago. ... if said library still exists, and you could gain access to it, and had the patience to wade through the chaff and misunderstanding.

Have fun with your adventure.




Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

The tubes are supported every 800-1000mm (about 3 feet). Its 4 tube rows vertically and 15 tubes horizontally. So there is not much trouble with gravity. The flow is slightly uneven because thw outflow is at the bottom of one end and the inflow at the other end.
I think there will be some device to redistribute here and there. But the current design is one baffle every 150mm/6 in, this is too much and pressure drop to high.

 

[srfish]

Concerning the heat transfer coefficient for parallel flow, there is chapter 3.3.12 Longitudinal flow in the Heat Exchanger Design Handbook.

 

[MikeHalloran]

So the resistance to flow comprises a bunch of rectangular orifices, 150mm x the tube-to-tube separation distance, arrayed in parallel and in series. ... but there are a bunch of said orifices.

I don't know what your limit is to call 'too much'.

I have had occasion to measure or estimate the pressure drop across the main engine coolant heat exchanger on a bunch of boats. I can't recall any of them coming out >~2psi, except when fouled by barnacles.

ISTR from very long ago that the 'laminar sublayer' was the greatest impediment to heat flow, and that direction changes tended to reduce its thickness. Am I remembering wrong?



Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

something like this in the picture.

 

[SteffenH]

and this one has 2PSi pressure drop in the simulation at 1.5 gal/6 L per s, without even putting tubes in
The tubes have fins and if I use the formulas for flat plate, flow along the tubes is turbulent, so heat transfer should be ok.

 

[MikeHalloran]

Tubes or no, it appears that the geometry of your shell is severely truncated, so I don't know how you're getting credible data from your simulation.

Can you find a real HX, remove the tubes, model that, actually test it, and validate your modeling technique?

I'd sure want to do that...



Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

this is the actual shape of the fluid if you want (for CFD), its a sub partition of a round shell. The upper bit would be a condenser and then the liquid would enter at either end towards the tubesheet and works its way back to the centre, where the outlet is. So this shape greatest height is 108 mm and length is 2m. Between the baffles is 150 mm.
Without the baffles would be easier to build and less pressure drop. So, if the HTC would besufficient in colinear flow, it would save lots of headache.
check this out:

http://www.embaffle.com/technology/case-studies/crude-preheat-exchanger.html

mesh baffle instead of zigzag. Removes dead zones, reduces pressure drop, turns it into longitudinal flow.

 

[MikeHalloran]

So, the bottom ~1/4 of what is otherwise a condenser, is a heat exchanger?
Separated from the condenser space by a floor of sorts?
Intended to cool the condensate?
Producing a sharp gradient in shell temperature at the floor?
Has this ever worked In Real Life, or is the whole thing new?
Why are you bothering to model the flow without the tubes?

Would it work as well without the floor?
I.e., with a weir around the drain,
and with the tubes just sitting in the pooled condensate?




Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

the plan is, liquid flows around the tubes as long as possible.
It works with all these baffles, but its 24 pieces and a pain to assemble. With the baffles the liquid takes 10 s to get through, because there is lots of dead zones, so it accelerates and just slaloms through. Without the baffles its 15 sec.
I did not model the tubes, its 51 off. Would take forever to calculate.

 

[MikeHalloran]

Well, now I'm really lost.

Have you got a mentor nearby?


Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

no. Its not homework.
We built some of these and it works, the liquid gets a bit cooler.
Its just troublesome to build, lots of little baffles and the ceiling. Longitudinal flow would just have 2 tubesupports, not 24 baffles.Pressure drop would be 1/4 of the cross flow design or even better.
I am just curious, if there is a way to calculate the expected performance. What I found so far is tubes in cross flow, formulas with a bunch of factors. If I could calculate heat transfer coefficient in longitutinal flow based on flow across a flat plate.
I dont want to do the CFD with 51 tubes, gigantic mesh. And without tubes its already obvious that the fluid stays in there for longer.
Heat transfer coefficient is based on velocity, so longitudinal has slower velocity and should have lower HTC. But the flow is still turbulent. For crossflow I got insane heat transfer coefficients.
10 or 15 seconds will not make much difference on the heat transfer, because the difference in temperature is fairly small.

 

[MikeHalloran]

Did you follow up on srfish's message of 26 Mar?

Mike Halloran
Pembroke Pines, FL, USA

 

[SteffenH]

there are 3 heat exchanger handbooks
Handbook for Heat Exchangers and Tube Banks design - Donatello Annaratone
Heat Exchanger Design Handbook - E. U. Schlunder
Heat Exchanger Design Handbook 2nd ed - Kuppan Thulukkanam
and this one has something on heat exchangers, too
Fundamentals of Heat and Mass Transfer 7th - Incropera, DeWitt, Bergman, Lavine

still reading

"As reported in Refs. [55,56], RODbaffle exchanger heat transfer rates compare favorably with dou-
ble segmental plate baffle exchangers and are generally higher than that in comparable triple seg-
mental plate baffle and “NTIW” designs. Shellside pressure losses in RODbaffle exchangers are
lower because neither bundle crossflow form drag nor repeated flow reversal effects are present."

 

[srfish]

The Handbook I was referring to has the executive editor G.F. Hewitt . It is part 3 and published by Begell House.

 

[georgeverghese]

The tube metal heat transfer resistance, whether this be for heat flow perpendicular to the flow through the tube or parallel is small in comparison to the sum of the fluid heat transfer resistances.

Hence, would imagine the net result is that you may find the heat transfer behavior is dominated by the fluid resistances ( in cross flow or parallel flow) with only a small contribution from the resistance across or through the metal wall.

 

[SteffenH]

essetamente. But that is governed by flow 90 deg or 0 deg to pipe axis, which gets Reynolds no and so on. Back to square 1.
Seems srfshs book is the only one, that has an answer, any other one I found so far only got xflow. Even that is more a cookbook, lots of factors to choose from, Nusselt numbers from 400 to 800 easily for same initial values....