Flow calculation

[rsr32]

Hello,

I have a 30,000 Gallon LPG tank about half full, that will be emptied into an empty second tank, through a piping arrangement. The first tank is pressurized using a compressor. The differential pressure created between the two tanks is 100 PSID

I am trying to figure out an expected flow rate (GPM)at the discharge of the piping going into the tank to be filled

I've attached a sketch, which has layout and sizing information.

also to note, there are excess flow valves at the piping intakes in each tanks, that will shut the flow off at a certain threshold. Unfortunately, the specs on the valves is not known, but I assume they must still be accounted for for friction loss in the head loss calcs.

I am not sure about how to account for the head loss in the entire system. I am using a bernoulli equation but am running into problems in solving it.

As can be seen by the diagram, the 2 inch piping dumps into a 3" header, then back into a 2 line to the second tank, with valves and elbows along the way. All valves, fittings, and piping seams are welded. An approximation is really all that is needed and in fact, all i need to prove is that the flow rate will be at least 250 GPM or more

Any help with this is greatly appreciated,
Thanks,
Dan

 

[zdas04]

As I say here often--Bernoulli's equation only works for situations where friction can be safely ignored, surface lengths must be in a few inches not several feet. There is something called "modified Bernoulli" that is taught in some schools, but is absolutely an abomination that should be avoided at all costs (it appears to work because friction is plug figure).

You didn't say if you expected the flow to be gas, liquid, or two-phase. Unless you're assuming 2-phase flow, you can use D'Arcy-Weisbach for liquid or the Isothermal gas equation for gas. If you expect flashing in the piping then no equation is going to give you a very good representation of flow.

David

 

[rsr32]

Thank you for the reply David,
the flow is assumed to be fully liquid, again all I really need is an approximation at best and to know that the flow is at least 250 GPM or more.
So it sounds like I need to use the D'Arcy Weisbach equation in this case, and i assume I would have to calculate the system losses and plug that in?

 

[Latexman]

That is correct, if you ASSuME no flashing. On your sketch, the 3" header has no length given. Now where are those excess flow valves again? They are not on the sketch. Is there one or two in the flow path? 2" or 3"? No wonder you are having trouble.

Good luck,
Latexman

 

[rsr32]

for the sake of simplicity, the flow can be considered to be all liquid. Again, all I'm looking for is a quick and dirty ball park figure, and that's all that's necessary to serve the purpose in this case.

The 3" header length is 142"

the two excess flow valve are on the end of 2" piping inside each tank. They are size 2".

 

[rsr32]

* the two excess flow valves are on the end of the 2" draw tube inside each tank.

 

[zdas04]

You can get viscosity at your conditions from something like REFPROP.exe (from NIST) or from their online program (I find REFPROP to be worth the nominal cost). Make sure you pick conditions that are close as liquid viscosity changes a lot over a pretty small pressure/temperature window.

David

 

[Latexman]

Malema literature says use the resistance of a fully open standard globe valve for their excess flow valves. They
use standard globe valve bodies to construct our EFVs.

Good luck,
Latexman

 

[Latexman]

Upon closer inspection and despite what they say, Malema has a graph of a 2" 300# EFV with a Cv = 10 (U.S. units).

http://www.malema.com/pdf/M_XF.pdf

Good luck,
Latexman

 

[SNORGY]

You can probably use DeGuzman-Andrade [m = Ae^(B/T)] for viscosity with close enough accuracy if you know the viscosity at two different temperatures; it's probably close enough to a single component Newtonian liquid if it's not flashing. On the other hand, this correlation is poor for multi-component HC. For my calculations I typically use Darcy-Weisbach with a Newton-Rhapson iteration of the Colebrook-White equation for friction factor, but in the flow rates coincident with the line sizes described, you would probably be close enough with f = 0.015. The most important consideration for the hydraulics in the whole set up will be establishing good K=fL/D values for the components that you mention, such as the excess flow valves.

Regards,

SNORGY.

 

[chicopee]

Aside from the material presented by the above responders, you should have a return line between the two tanks so that you have equal pressure in both tank during the transfer period.

Take a look at suppliers filling bulk storage tanks from their tankers, you will see two lines between a tanker and a bulk storage tanks during the transfer of product for the reason stated above and to control fugitive emissions.

 

[rsr32]

thanks all,
I was hoping there was a quick and dirty way to get a rough approximation but it looks like I'll have to dig up the old fluid mechanics books and look at it closer



Chicopee,

yes you are right, I omitted the vapor return arrangement in the sketch for simplicity's sake. the differential pressure (100psid in this case) from a compressor forces the liquid from the first tank into the second tank. it's not a typical standard procedure to transfer product from tank to tank at this site and the reason they are doing it is to drain the first tank to take it out of service and scrap it.

 

[Latexman]

I don't see what a vapor space equalization line gets you in this case, except to eliminate the dP and stop the transfer!

Good luck,
Latexman

 

[rsr32]

the vapors are actually controlled by a regulator, that feeds into a distribution system

 

[katmar]

A quick and dirty solution, if you do not have software available, probably means using an online calculator or downloadable spreadsheet. Google will find them. What you need to understand is the assumptions around the data that you feed into the calculator.

To be conservative, simply assume that all the piping is 2". The 3" section is short and is almost irrelevant, but taking it as 2" will not be far wrong.

The resistances of angle pattern globe valves vary considerably, but as a start take them as having an equivalent length of 150 pipe diameters. As a first guess assume the excess flow valves are the same.

In terms of the physical properties, it is fairly important to get the density close, but the viscosity is of little importance. It only affects the friction factor very slightly in the turbulent regime.

Plugging all this data into an online calculator will tell you that at 250 USgpm your pressure drop will be about 45 psi. BUT if you put 250 USgpm through a 2" Sched 80 pipe your velocity will be about 27 ft/s and this should raise red flags around the excess flow valves. I would guess they will be your limiting factor - not the rest of the piping.

Katmar Software - Engineering & Risk Analysis Software

http://katmarsoftware.com

"An undefined problem has an infinite number of solutions"

 

[rsr32]

thank you katmar, I will search for these spreadsheets and calculators. for my purposes as far as this case, I don't need a high level of accuracy, the rough estimation I can get by simplifying the system is enough. unfortunately I do not have specs on the excess flow valves but I will continue to look into it hoping I can dig up some info on them as it sounds like they will be the limiting factor.

 

[Latexman]

The least that can be done is examine the EFVs for any data/markings. The ones I've installed are well marked, but they may be quite old.

Good luck,
Latexman

 

[chicopee]

Latexman, the vapor line can also reduce the possibility of the propane from flashing into the vapor phase at the pump intake.

 

[Latexman]

What pump? The OP said a compressor pressurized tank # 1 and pushed the fluid to tank # 2.

Good luck,
Latexman

 

[rsr32]

any suggestions for a decent table of equivalent lengths for valves/fittings? can't seem to find any that list a value for 90deg angle globe valve..