Lobe Pump Repeatedly Blowing Rupture Disc

[TiCl4]

We are having a somewhat perplexing problem.

We have an emulsion polymerization system in which a monomer emulsion is created in one tank, then pumped over to atmospheric reactors at 120-200 F over a 3-4 hour period. The reactor and the emulsion tank are at equal elevations, with a lobe pump used as the means of conveyance between the two. A rupture disc is on the discharge side of the lobe pump, and recirculates back to the emulsion tank. Between the lobe pump and the reactor are a check valve and an automatic valve. The disc is set for 115 psig. The line and pump between the emulsion tank and reactor is 2", and typical flows are ~15 gpm. Viscosity of the emulsion can vary, and it is shear thinning, but usually 1-3,000 cP. These tanks are drawn until empty (drawing air), then are washed down with water to clear the emulsion tank.

We are having an intermittent issue: Sometimes when a batch is finishing and the tank is about to be washed down (or is being washed down), the rupture disc blows and recirculates material back to the tank.

To my mind, there are only two possible culprits here - the check valve and the automatic valve. Due to the repeated nature of this incident (and the timing of the failure), I suspect something wonky may be going on with the check valve. It is a swing-type check valve (a TCI TC-CKF, 2" 150#, to be specific). Is there something with swing check designs that can cause pressure spikes if the liquid flowing through it suddenly changes to air? All I can see is that the check valve may chatter a bit due to pressure fluctuations, but with only 7-8 psi static head due to liquid level in the reactor, there isn't much pressure to overcome.

Liquid Flow:

Tank -> Lobe Pump -> Check Valve -> Actuated Valve -> Reactor.

 

[Latexman]

If the ABV and/or check valve close too quickly a pressure wave/spike could burst the RD. We don't use swing checks in emulsion service. It gets gummed up and acts erratically. Try a ball check. The transition from laminar (emulsion) to turbulent (water) can be very noisy in terms of pressure waves.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water

 

[don1980]

For reliable service from a rupture disk, that's a very problematic application (incomprssible filled system, pd pump, high viscosity). If the system didn't pose a high risk of plugging, the obvious solution would be to replace the RD with a PRV.

Your RDs are probably bursting due to pressure spikes. i would install a pulsation dampener (the bigger the better) near the RD, and slow down the closing/opening time for the automatic valve.

 

[gerhardl]

As the two previous posts indicate, this is quite probably a double problem, and the solution should take care of both probabel causes.

1. Pump with soft start and stop.
2. Pulsation dampener might help, better to prevent/avoid pulsation.
3. Can washing process create steam collapsing (when cooled) and sudden pressure changes? Check!
4. Pipeliena layout and placement of checkvalve OK ? (No fluid train rushing either way when closing)
5. Extra checkvalve(s)?
6. Ball and swing-check valves are the two checkvalves most likely to create the problems you describe. Change to soft-closing if available for the fluid. Spring- weight loaded swing-check might be suitable, correctly dimensioned. Costly but good.
7. Recheck set pressure of burstdisc. Higher possible? More narrow tolerence possible? Bursting in correct direction? (Check!)
8. PRV will not remove the spike problems, but maybee some or more of the points above.

Good luck!

 

[georgeverghese]

The selection of an RD for this application seems to be incorrect here. An RD works much like a conventional PSV - it ruptures upon detection of high dP. I suspect the RD is specified to rupture upon detection of a given dP when emulsion tank level is high. When this tank is drawn down, with the pump operating at normal discharge pressure, the RD blows since it sees a high dP, as pressure downstream of the RD is much lower at low emulsion tank level.

See if you can replace the RD with a balanced bellows type PSV - in normal operation, this would stay closed even when tank level is low. At high tank level, the built up backpressure ( static head + frictional drop in PSV discharge line back to tank) on the PSV should be kept at less than 50% of PSV set pressure.

 

[TiCl4]

All: Thanks for the responses. Maintenance took the system apart and found the swing checks to have a good amount of buildup on them. My guess is that this was enough to cause the check valve to occasionally stick closed when the pump was cycled. We have cleaned and replaced them for now, but will be replacing them with ball checks and put them on a cleaning PM. Latexman - we actually have ball checks on every other system, but somehow don't have it on these! I'll try to remember to update this in a few weeks if the issue disappears.

Gerhardl:

1. The pump is on a VFD, and is started from 0%.
2. Agreed, but I don't want to deal with dampeners in a plugging service if I don't have to. These failures are happening too quickly to be fatigue failure (we've run for a long time without blowing discs before).
3. No steam in this system. No condensable vapors.
4. Sketch is attached. The check valves are right before the automatic valves, which are directly bolted to the reactors.
5. Seems like it would cause more potential for overpressure if the failure mechanism is the check valve sticking shut.
6. We are going to see how the ball checks go. If the problem persists despite the PM and change, I'll look into that.
7. I have moved the set pressure to 141 psig - the highest allowed based on the pipe spec and the burst disc tolerances (+6% of SP). This disc is oriented correctly.
8. Not a fan of PRV in viscous, fouling application where buildup is possible.

George: The RD is to protect the piping system in the event of the pump deadheading. The piping system has a spec of 150 psig @ 230 F according to the pipe specification. There is no consideration of emulsion tank level. The selection of a RD over a PSV is due to two facts: the material can be fairly viscous, and the material is prone to plugging due to polymerization.

 

[Compositepro]

Controlling pressure on the output of a positive displacement pump with only a vfd will almost guarantee burst rupture discs. When a valve closes, the pressure will spike before a VFD can respond. You need a pressure relief valve or pressure damper.

 

[TiCl4]

The VFD is used for flow control, not pressure control. I am aware that electronics cannot respond quickly enough to deadhead pressure spikes. That is the entire reason for the rupture disk - to provide pressure protection upon the accidental closure of the valve when the pump is running.

 

[Compositepro]

So what are you complaining about? Everything is working exactly as designed.

 

[TiCl4]

Compositepro: Yes, the rupture disc and relief worked as it was designed. However, the rupture disk repeatedly blowing (several times a week) is a serious issue. A rupture disc failing that means normal control design has failed and the last line of defense has activated. In some companies, any relief device (disk or valve) relieving is an incident that needs to be investigated and cause analysis completed.

Also, rupture discs aren't cheap. Running $2-3k per week on replacement discs is an issue as well.

 

[dogtop]

TiCl4:

What software do you use for your P&ID?

dogtop
An Inversion of Top Dog

 

[TiCl4]

Dogtop: AutoCAD LT 2019. I've also used Draftsight 2017 (free version of AutoCAD) with fairly good liking. The picture that I posted was not a CAD drawing, though - it was sketched in Microsoft Word as part of the rupture disc design basis document (which requires a sketch of the system being protected).

 

[dogtop]

TiCl4: Thanks.

dogtop
An Inversion of Top Dog

 

[Compositepro]

You: We have rupture disc failures. What can we do about that?

Me: Vfd control of a positive displacement pump is guaranteed to produce pressure spikes.

You: That is what the rupture disc is for, to take care of pressure spikes.

Me: So your system is working as designed.

You: Yes but rupture disc blowing is a serious issue.

We seem to be having a failure to communicate. By the way the rupture discs are working, not failing.

 

[TiCl4]

Compositepro: I think we are indeed miscommunicating.

You: "When a valve closes, the pressure will spike before a VFD can respond."

Me: We don't close any valves when the pump is running. The pump is shut down before any valves are closed. Therefore, there should be no pressure spikes that can burst the disc. We are experiencing rupture discs bursting while running the pump and while all valves are open. This should not happen. Therefore, there is some issue that is causing the discs to burst.

You: Vfd control of a positive displacement pump is guaranteed to produce pressure spikes.

Me: I assume you mean in the control scheme as detailed in your first quote, where you seem to assume that we are closing a valve and deadheading a positive displacement pump. Do I misunderstand you? By themselves, VFDs have nothing to do with pressure spikes.

Also: I am using the term "failing" in the structural sense - as in "the structure integrity of the rupture disc is compromised and it breaks". It is not meant to connote that they are not performing as designed.

 

[georgeverghese]

Okay, with the discharge line going over the top of the tank, backpressure on the RD is constant, so in principle, the selection of the RD is acceptable.

Agreed, a sticking check valve may be the cause of pressure spikes. Swing checks are also notorious for chattering at low flows. Wafer checks are better-they should be correctly oriented if you choose to opt for wafer checks.
With two control valves inline, any excess pressure in the line would be dissipated through either one partially one control valve, provided there is some excess pressure margin available (or if there is a gas filled dampener inline to absorb the transient pressure spike).

Also check for temperature mismatch between RD specification and actual operating temp.

Are these lines heat traced? Are all tracing loops working well? Its well known that electric heat tracing works a lot better than steam tracing.

 

[Latexman]

VFD's usually allow one to set a max. speed or % of full speed. Set this at the lowest you think folks can tolerate from an operations (capacity?) stand point. 50%? 75%? You will know better than me. The frequency of bursting RDs should decrease immediately, maybe even stop. Then, as you find and install other solutions (pulsation dampener, different check valve, no check valve, etc.), % of full speed can be increased.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water

 

[Terrifyingtesttube]

It doesn't help you solve the problem as to why the pressure spike is happening, but 4-port pressure relief valves (see sketch) are commonly used in food & beverage to protect around PD pumps without introducing that dead-leg/fouling/contamination risk. They basically act like 2 separate pipes until the over-pressure is tripped (available to open up or down, mechanical/spring set, and can also come with an air-actuated option for forcing open during cleaning). I have used them for everything from watery to greek yogurty type fluids.

 

[TiCl4]

All,

Thank you for all of the responses. It has now been 1 month since we took the check valves out for inspection. They were found to have significant buildup on them. They were cleaned and replaced (Latexman, we will be doing the MoC to transition to ball-type soon - we have ball-type checks everywhere else but on this system!). We also put the check valves on cleaning PM.

Since that cleaning and institution of the cleaning PM, there have been no instances of this occurring. This is sufficient for me to reasonably conclude that the check valve was indeed sticking and causing a deadhead scenario.

Terrifyingtesttube - neat design. What manufacturer have you used in the past?

 

[Terrifyingtesttube]

GEA (Q type) and SPX (WCB W61) are the ones that I have used.