Water Hammer Pump Sudden Shut off

[TechnoZA]

I would like to know what the effects would be if a pump had to suddenly be stopped due to, for example, a power failure. Is there significant risk of water hammer? What measures are recommended for preventing water hammer?

In my specific situation, I will be running 3 pumps, sucking and discharging into common manifolds, with each pump delivering around 870m3/hr. Suctions 18" and discharges 16".

 

[BigInch]

Its pretty good. My only comment is I think things might happen a lot faster than what you may have in mind. Quick acting swing check valves close very rapidly as well, so flow stops quickly and a lot of power gets expended during that short time. Another is that most pumps don't tend to coast very long against the discharge pressures of the product they just compressed and their moment of intertias are relatively small. Most pumps seem to stop in a sec or two. If pressures reduce below the vapor point of the product anywhere in the system, you start making vapor until (or if) pressures begin to recover (as downstream velocity head is converted to static head) and pop back up above vapor pressure. At that point, the vapor spaces collapse quite quickly, often violently. If that brings two separated columns of liquid together (you could have several hundred psi on one side of each column and only vapor pressure in the bubble in the middle, so acceleration to collision is quick), another pressure spike (could even be bigger than the initial event) is created when they crash and the cycle could start all over again, if sufficient energy remains.

"If everything seems under control, you're just not moving fast enough."
- Mario Andretti- When asked about transient hydraulics

http://virtualpipeline.spaces.msn.com

 

[stanier]

You do not guess these things.

Model the system with pumps at different speeds. You will soon see that the pump does not add to the system after a very small deterioration in speed.

If it was possible to guess engineering data there would not be the 15 or so transient analysis software packages out there that enable you to model such systems. All the graphical techniques would not have been developed to estimate the size of a pressure transients. Books would not have been written and in addition all the inquests undertaken for those killed in surge incidents would have been a waste of time.

I hope you have plenty of insurance and do not spending time in jail.

Down load a demo from AFT and look at a system response to what you are talking about.

The reason you pay for specialists to engineer things is the same as why you engage a lawyer, tax adviser, accountant, plumber, electrician etc etc. You cannot do the necessary work and need help. Judges frown on you not spending the money to engineer things.

Do you guess the wall thickness of a pressure vessel or pipe, adequacy of a foundation, the earthquake resistance of a building? If you do you do not belong on this forum.

 

[BigInch]

After all that talk above, it turns out that this system isn't too interesting from the transient perspective. Good news for TechnoAZ. The velocities are around 6-7 fps 2 m/s, the normal operating pressures arn't too high, not much change in elevation, so no high starting torques needed, the pumps are relatively slow rpms at 1440, main pump's power is not too great at 250 HP, only some check valves; no potential for over/under-shoot setpoint control problems, means that pressure transients (from a precursory look) shouldn't be much more than 30-40 psi (210-280 kPa) or so.

"If everything seems under control, you're just not moving fast enough."
- Mario Andretti- When asked about transient hydraulics

http://virtualpipeline.spaces.msn.com

 

[TechnoZA]

Thanks BigInch. Sort of what I expected, in my ignorance.

Stanier:
All I ask is the opportunity to gain understanding and knowledge. I'm not attempting to be an expert, or to make decisions on the matter. The stimulating discussions and ideas are useful to me. At least I get the opportunity to gain some understanding from those with much more experience than me. Your point about the seriousness of the matter and the need for professional advice is taken. Thank you.

 

[BRIS]

A pump stop is hydraulically instantaneous if it is less than the time of travel of the pressure wave to travel from the pump to the tank and back to pump. 2 x L/c

c = speed of the wave 600 -1100 m/sec (typically). L = pipe length

If the stop is instantaneous the pressure change (m head) Dh is DV x C/g (DV = change in velocity 2 m/sec, g = 9.81). Dh = 122 to 244 m head (12 to 24 bar). (Joukowsky equation)


Your pipe length is 4500m. The pipe period will be around 10 seconds (time of travel). Submersibles will stop almost instantaneously, the bosste pumps will stop in less than 10 seconds.

Pressure of +/- 12 to 25 bar are likely to be a problem !!.

Take Stanier's advice and run a model.

Brian

 

[BigInch]

Bris,

I modeled several pump stops both lift and mains and according to my results, its not happening. I don't believe the JoukowskI equation applies since there is no actual valve closure anywhere and I show water continues to be pumped, as the pump quickly slows down, so hydraulically instantaneous or not, pump stops don't get a big pressure build. I agree, as I have noted, my results should not be considered definitive and obviously TechnoZA needs to confirm that, but that's what I got.

"If everything seems under control, you're just not moving fast enough."
- Mario Andretti- When asked about transient hydraulics

http://virtualpipeline.spaces.msn.com

 

[BRIS]

If the pump stops before the reflected wave returns to the pump, which is the case here, then the Joukowsky equation applies along some length of the pipeline. Of course the negative pressure will only drop to vapour pressure but then the problems start!

(in fact the Joukowsky equation applies to each decrement/increment in velocity caused by pump stop/start)

A by pass around the booster station may be a simple and economic solution here, but I agree with Stanier you cannot risk not doing a study.