Forced Convection 1

[GreenRoger]

I have a problem similar to a thread I found here:

thread391-227383

I am designing a generator cooling system. The generator is in a sealed steel tube which is air filled and under water. I intend to weld a steel channel to the side of the tube and blow air down it to power circulation. So I need to work out how big a fan I need.

HeatPower = GeneratorPower x (1- Efficiency)

TempGenerator = TempWater + HeatPower x Area x Coefficient

The previous thread recommended a coefficient of 100W/m2K. That's a good start but it doesn't tell me how big a fan I need. I suppose I'll need to use CFD to get a better value than that but is there a rough formula I can use to relate the fans flow rate to the coefficient?

(Relating to the previous thread, under water vehicles use an electrical umbilical which may carry was much as 2MW. These are wrapped on a winch drum which is water cooled.)

 

[jmiles]

What yuo really want to do is calculate your own convection coefficient, by doing that you will have to start with a fluid velocity, and THAT is how you will begin to size your fan, the real problem is you need to know how fast the air is moving over the pipe. from there you can get a much better picture of your convection coefficient.

I would putt he formula up here, but tis kinda ugly and wont look at all nice ont he form style, so it might not be much help to you.

 

[IRstuff]

The value given in that thread did not come with a detailed calculation.

You need to figure out what you need, and design accordingly.

How deep is this assembly, since you need to account for the pressure at depth for a pump to work?

One also wonders why you don't use the water, since it's already there?

TTFN

FAQ731-376

 

[GreenRoger]

Thanks Jmiles, I know the temperature of the water outside the pipe; that's what I'm calling TempWater.

Thanks IRstuff, The generator is in an environment sealed at atmospheric pressure, so the pressure depends on the temperature.
Pressure = PressureAtmos x Temp / TempAtmos
Water cooling would be a bit risky in case it leaked into the air filled tube. The tube can't be entirely full of water since it needs to float.

 

[IRstuff]

OK, just to clarify, your "tube" is on, or near, the surface of the water?

I wasn't suggesting filling the tube with water, just running a pipe through it to the generator. As for leaks, it's only marginally worse that the sealing of the tube itself, no? In fact, you've got way bigger joint seals for the tube than any liquid cooling would actually require. However, liquid cooling can pull way more heat out than the air can.

TTFN

FAQ731-376

 

[GreenRoger]

Yes, it's near the surface of the water.

To compare liquid cooling to air cooling I still need to answer my original question; how to calculate the convection coefficient.

 

[IRstuff]

Well, you can start here if you really want to try and calculate it yourself:

http://www.efunda.com/formulae/heat_transfer/convection_forced/calc_lamflow_isothermalplate.cfm

There are lots of heat transfer calculators on the web that you can look at. You haven't said how much power you're trying to dissipate and the criticality of the cooling. That will dictate the design approach, not the htc itself.

TTFN

FAQ731-376

 

[jmiles]

Ok, this will be messy, I got my convection coefficient equation from Process Heat Transfer by Kern, and ive since put it into excel, so I will start with just giving you the excel formula, then i will break down what everything is.

E34*(0.3+(0.62*E31*E32)/((1+E33)^0.25)*(1+(B32/282000)^(5/8))^(4/5))

E34 = k/D
E31 = Reynolds number to the 1/2
E32 = Prandtl Number to the 1/3
E33 = (.4/Pr)^2/3
B32 = Reynolds Number

k = Thermal conductivity of the external fluid in W/mK
D = Pipe Diameter in m

So thats what i use to calculate the external forced Convection Coefficient, the formula is different for Natural Convection

Natural Convection (Please note this one is truly messed up as i typically work in metric, but the formula was in Imperial, so all the G units are imperial units, and F is the conversion factor back to metric
=116*((G56^3*G57^2*G58*G59/G60)*(G61/G62))^0.25*F65

G56 = k (BTU/ft/hr/ft^2/°F)
G57 = rho (lb/ft^3)
G58 = cp (Specific Heat of the external Fluid in BTU/lb°F)
G59 = B (Coefficient of Thermal Expansion Length/Length/°F)
G60 = mu (Dynamic Viscosity in cP)
G61 = temp (°F)
G62 = diameter (in)
F65 = 5.678263398

To see all my conversion factors check this website

http://www.unitconversion.org/unit_converter/thermal-conductivity.html

I hope this helps and does nto confuse the crap out of you, im still waiting tog et my own copy of Process Heat Transfer, id borrowed a former co-workers copy when i made this spread sheet. (Those silly out of print 1950's books are so hard to get)

 

[IRstuff]

100 W/m^2-K is definitely not natural convection, so the equations on Efunda would be the ones to use.

GreenRoger, how are these generators cooled terrestrially?

TTFN

FAQ731-376

 

[AllHandlesTaken]

Jmiles:

Did you try Amazon for "Process Heat Transfer"

http://www.amazon.com/s/ref=nb_ss_0...+transfer+kern&sprefix=process+heat+transfer+

They have many copies ranging from a new $90 reprint to a like-new original $44.95

 

[IRstuff]

Well, if you want to go cheap, there's free:

http://web.mit.edu/lienhard/

http://www/ahtt.html

TTFN

FAQ731-376

 

[jmiles]

Damn, you know i did look on amazon, but when i had looked before there were only expensive softcover reprints from foreign countries, and they were over 100 a piece, so i opted for ordering an original new one from a local technical book store, they just have not come through with it yet, 167 CAD though, so not too bad all in all.

 

[GreenRoger]

The Efunda calculator unfortunately only works for laminar flow. My flow is turbulent.
I'm not convinced by Jmiles formula; it asks for the conductivity of the external fluid while I am looking at the fluid inside the pipe.
I've downloaded Lienhard but haven't found a useable formula for this in it.

 

[ginsoakedboy]

GreenRoger,

You can also try the Thermal Wizard website here:

http://www.thermal-wizard.com/tmwiz/default.htm

The web pages are coded so they determine laminar/turbulent flow domains before computing.

 

[jmiles]

Sorry the formulas i gave were for external convection coefficients.

For the internal convection coefficient i use

Nu*k/ID

Nu = Nusselt NUmber
k = Thermal conductivity in W/m°K (Internal Fluid)
ID = Internal Diameter in meters

 

[GreenRoger]

Thanks Htlyst, the Thermal Wizard website has given me an answer.
I'd been trying to calculate the Nusselt number for myself but haven't managed that yet.

 

[GreenRoger]

I've put the turbulent equations from Thermal Wizard into Excel (attached) but I'm not getting the right values for friction.

 

[ginsoakedboy]

GreenRoger,

Your fluid velocity is too small to be in turbulent regime. (0.0001m/s through a 1m dia pipe) You should use the laminar formula.

Look at the theoretical stuff at:

http://www.thermal-wizard.com/tmwiz/convect/forced/fd-tube/about.htm

 

[GreenRoger]

The velocity should be higher eg 10 m/s

 

[GreenRoger]

Here's the spreadsheet with the correct speed. I think the friction formula is mis-typed.