Steam collapse - pulling vaccum in a water tank - help please

[erbru]

Hi all

I need to check the sizing of an air flow into an insulated atmospheric water storage tank for the following situation and it is causing me a headache. I will then also need to size a new relief valve for the tank based on the scenario of the air flow system injecting too much air at high pressure.

The water tank (approx 35 m3, 5 metres high by 3 metres wide) is not foreseen for vacuum service nor pressurised service. That means from about -50 mbar to + 100 mbar is the maximum range. It is a very thin walled tank.

During hot 'sanitisation', the tank is totally emptied and filled with 2 m3 of water at approx 85°C.

The pressure in the tank increases rapidly during the filling process due to the hot vapour formed above the water surface.
To protect the tank, when the tank pressure reaches above 10 mbar, an automatic valve opens and releases the pressure to vent. This process is repeated as necessary as the tank is filled with hot water.

After the tank is filled with 2 m3 hot water, the tank water is pumped from the tank bottom through a piping loop (to clean the loop with the hot water). The pumped rate is approx 22 m3/h. All the water returns directly to the tank.

The loop had been full of water at ambient temperature and this 'cool' water returns to the 'hot' tank via a sprayball, immediately causing steam condensation and subsequent collapse in pressure.

The same pressure transmitter then opens a high pressure air circuit (8 bar) to the tank to equalise the pressure (set point for opening the air inlet valve is -20 mbar).

During the recirculation process, hot water continues to be added to the tank until 6 -8 m3 level is reached. During this time air is injected many times into the vessel in small bursts.

However it seems the air injection system is not correctly sized for the condensation and loss of pressure, since just after the start of the cold water recirculation, the pressure spikes to -40 mbar. The temperature (as measured on the side of the tank, and possible not fully representative) drops from about 85 °C to 70°C during this short space of time. During the subsequent phase of continuing the filling, the tank heats back up to 85 °C gradually and there are no more sudden spikes in low pressure.

The operators report a 'trembling' of the tank during this process.

I am trying to understand the correct air flow required for the worst case, which is as the cold water enters back into the tank and causes that sudden condensation.
How can the effect of steam condensation on pressure at this temperature be estimated?
Can I use the ideal gas law?
As for steam tables, I have looked but I don't understand what is happening to the enthalpy of the system here. The cold water spray is removing enthalpy in the form of latent heat (and some sensible heat) but how do I use that calculate the resulting pressure change.

Attached an output from the plant data logger showing the parameters of tank level, pressure and tank temperature.

The process starts between 13:48 and 17:36 on the time axis.
Green : pressure in bar (-0.05 to + 0.05 bar => - 50 to +50 mbar)
Blue : tank level (0 to 25 m3)
Black: tank temperature (0 to 100°C)

I suppose I could use this data to calculate the current air flow into the tank (pressure change in a known volume)...

 

[erbru]

Tank pressure variation during the initial heat up phase (addition of the first 1-2 m3 of water) and before recirculation. The steam is vented from the tank every 10 minutes or so as the vapour pressure reaches above 10 mbar. No air is injected. The venting happens via the filter on the tank and is supplemented by the pneumatically operated air valve once above 10 mbar. This shows how the current air filter doesn't allow the tank to fully breathe during this process. So the automatic vent is also essential.

 

[erbru]

A longer duration picture of the overall process from start of recirculation to about 6 m3 of level. Note the stabilising out of the temperature. I guess the thermowell is starting to get submerged at this point.

 

[EdStainless]

I would look into changing the sequence.
Can you be running full recirc before you start adding the hot water? This would average temperatures faster. It might require leaving a little more water in the loop, but it would help.
And I would set the air injection to always be adding air, never let it close. Yes you will be venting but that is better. I am assuming that the air valve is proportional and not just on/off. If you don't let it close then it won't be slamming as hard and should reduce the fluctuation.
Your control system just can't respond fast enough.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[georgeverghese]

Looks like your pressure controller for air inbreathing and the PC for outbreathing are both on - off controllers - what are their control setpoints ? I can see the reasoning for the on - off concept - a regular PID controller ( typically in PI mode only ) may be too slow to respond to the sudden collapse in tank vapor space pressure as the cold recirc water plug gets into the tank. A large gap between controller setpoints ( in and out breathing PCs') should help to reduce this pressure hunting.

Is there not a nozzle on the tank liquid side where you can return this cold water to, rather than the current one in the vapor space ? This would help to resolve both the pressure control and the safety issues.

 

[LittleInch]

erbru,

The context of this post is that you are still looking to deal with the symptoms rather than the cause.

Maybe it got lost in the midst, but I thought the idea of insulating and trace heating this cold water return line, which is a transient event and not part of the required procedure, deserved a bit of a look in??

Then the initial spray of water is actually warm / hot water and your sudden condensation of warm vapour goes away or reduces a lot, or am I missing something.

Seemed simple and easy to install to me.

However the change to empty the loop and start with hot water is a good move and should have a significant impact.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[georgeverghese]

If there isnt a nozzle you could use for this cold water recirc. stream, then another alternate would be to add an internal fullsized dip leg from the current recirc return nozzle that goes down to below the minimum liquid level in the tank during this sanitation operation. Drill a small 1/2inch syphon break hole at the high point of this dipleg.

 

[georgeverghese]

If you can return this cold stream to the liquid side, would suggest a guesstimate condensing heat transfer coeff of say 2850 w/m2/degK (500btu/hr/ft2/degF) over the exposed vapor liquid interface area of 7m2 at a temp differential of 85-30 = 55degC to compute the peak condensing relief rate, assuming total failure of the hot water temp makeup stream. See if this gives you some sensible results.

 

[georgeverghese]

One possible cause for the hunting in this on-off pressure control scheme may be that there is no deadband provided beween the vent and inbreathing pressure controllers.
An example setup with a deadband would be : the vent PC could be on OPEN command at 80mbar and CLOSE command at 20mbar; while the inbreathing PC could be on OPEN command at -40mbar and close command at -10mbar. So the tank pressure free floats between -10mbar and 20mbar, and the 2 pressure controllers dont "step on each others toes".

 

[Jordan Hribar]

When a sudden vapor collapse starts, how much air must be injected to arrest the pressure decrease and prevent the tank from failing? In my opinion, that is an academic calculation that has little to no practical value. During a worst-case vapor collapse scenario (e.g. high flow rate of low temperature cold water, finely dispersed by the spray ball into the tank) there's no practical way to stop the tank from failing. For vessels that are not designed for full vacuum, the way to defend against sudden vapor collapse scenarios is to prevent them from occurring rather than attempting to stop them once they have started. That's done by preventing the vessel's vapor space from being enriched with condensable vapor. Ensure that there is always a sufficient amount of non-condensable gas present, such that this scenario can't occur. When air/N2 injection is used as a layer of defense, you should start the injection as soon as possible. That is, start at a set point that is as high as possible, but below the normal operating pressure. For example, if you normally operate a 5 psig, then start at say 1-2 psig rather than waiting until the pressure has dropped into the vacuum range. And use a quick opening valve rather than a pressure regulator. Of course, you'll also need to make sure the vessel's pressure relief device has enough capacity so that this air/N2 flow rate can't overpressure the vessel.

Well said!

« Rien ne se perd, rien ne se crée : tout se transforme ».
— Antoine Laurent de Lavoisier (1743-1794)

 

[erbru]

Hi again
I totally agree that it is better to stop the event occurring. The fact is that the vessel must be sanitised due to the pharmaceutical industry's requirement. And changing the water vessel storage temperature to shock the microorganisms is the way this is done (chemical sanitisation isn't acceptable). And the tank has not failed in 15 years despite having gone through this procedure once a week. That in itself is something of a miracle!

I am going to seriously suggest about the two following operational changes:
- operating in a higher pressure zone (eg 0 to 30 mbar) and avoiding a dead band between the zones (but for this I need to knwo what the maximum allowable overpressure is, and I don't have the design files of the tank)
- emptying the loop and recirculating only hot water
- reduce the sanitisation temperature as far as possible (even a few degrees could help)

I am also in discussion with a rupture disc manufacturer but in these ranges the rupture discs tend to be huge and we don't have an appropriate nozzle.

George - I can't easily return the cold stream to the liquid size without very significant re-engineering. There simply isn't a nozzle available. Adding a dipleg to the current nozzle may be possible but would mean the vessel wouldn't be cleaned by the spray ball on loop return.

The dead band idea is worth looking into.
The control setpoints are :
Pressure = +10 mbar - open vent valve (& close again at 0 mbar I guess)
Pressure = -20 mbar - open air inlet valve (& close again at 0 mbar)

LittleInch - insulating and trace heating the cold water return line - could be done but it wouldn't totally cure the problem. Even an insulated vessel with only hot water added or sprayed into it, would have still a condensation case when the hot water addition stops and as the outside air cools the vessel. I'm not sure how the two condensation rates compare - obviously the cold spray is the worst case, but as this article on design for external pressure notes 'It is good practice to design any vessel exposed to steam service for full vacuum, the cool down rate can be very fast overloading vacuum protection equipment. '. Indeed is good practice!

http://pveng.com/asme-code-design/comments/external-pressure/

 

[LittleInch]

Great feedback and thanks for the responses.

I agree it won't stop the entire problem, but I think the sudden cool down and spray was your worst case and the other more gradual cool down could be dealt with by the current set-up. Maybe a mixture of the two approaches could be used - I just thought it was something quite simple to do, but I guess re-setting the set points is a relatively simple task in itself, but you would still be getting the sudden change from over pressure ( above atm) to below atm. Draining the system and filling with hot water just before spraying does the same thing I guess.

Yes, 750+ operations and no collapse is quite good, but as time progresses you may just stop getting lucky, especially as this tank appears to be vital to the ongoing operation.

You might need to be careful when working this close to atmospheric to make sure your instruments are measuring relative pressure, i.e. as the atmospheric pressure changes then modify accordingly or allow at least 5mbar positive pressure to always flow out of the vent and not inbreathe. Be careful not to compromise your overall sanitation requirements. Usually tanks are better at over pressure than vacuum collapse.

I'm not surprised your bursting disc is getting very big at this sort of pressure, but you can always work off a smaller nozzle and then add reducers to get to a larger size. A bit unorthodox, but if you haven't got a big nozzle then it's better than nothing.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[georgeverghese]

From a final containment point of view, (if you believe you dont have a well protected system here in the event operators have failed to follow safety startup operating procedures), would be to make arrangements to prevent hot water scalding of the operators due to a total rupture / implosion of this tank. This would mean a containment dyke to hold at least 110% of the max normal holdup in this tank, and keeping operators out of this containment zone during the sanitisation operation. Check local jurisdiction rules on other details on how this dyke is to be set up.

Yes, does look like you have room widen the deadband here to reduce this pressure cycling.

 

[erbru]

A little update:
After hunting further I stumbled across this very handy calculation tool down by Art Montemayor (some of you may have heard of him) for the 'steam out' of tanks and he has even done a simple model of rain landing on the tank during this process.

I ran the numbers in his spreadsheet and compared it to what I had computed. I took the recorded measured instantaneous pressure drops over a few seconds from the site data logger and used the ideal gas law to calculate the molar changes in water and hence the volume loss. This assumes water vapour at these pressures behaves as an ideal gas.

The numbers are not world apart. I get a volume displacement of about 3 m3/s and his method calculates 6 m3/s. His main assumption is the overall heat transfer coefficient of condensation (400 Btu/hr-ft2-F)

Based on his method I would need a 6 inch nozzle (assuming critical flow) to accommodate this flow. I guess this could be reduced if I accept a very slight underpressure. My calculated flow would indicate a 4 inch nozzle. Slight problem being, I don't have either. I guess I could build a little manifold of rupture discs.

For info, you can find Art's spreadsheets here:

http://www.cheresources.com/invision/topic/573-sizing-of-vacuumrelief-valve-on-storage-tank/page-3