Compressor Blocked Discharge

[Samiran]

Do you think one should take credit of recycle flow in the design of PRV on a centrifugal compressor discharge for a blocked flow case ? The point to remember is this discharge flow can be the governing case for sizing of flare ( 2000 T/h). However, if we assume recycle valves will open on blocked case the relief flow reduces drastically. This may violate code requirement. If the recycle valves do not open the compressor goes into deep surge.

Any thoughts to share ?

Samiran

 

[TD2K]

Credit is not taken for control valves operating as they are intended to operate unless their normal action will result in the relief load increasing, then you assume they will operate. Basically, you assume worse case because if you ever want to get in that position, you want to ensure you have adequate relief.

If they are already open, then you can assume they will remain in their last known position but I would argue for a surge valve, it 'normally' should be closed.

You can use the arguement that its action should prevent the relief valves from reaching setpoint or full load, however the numbers work out, but I wouldn't take credit for an instrumentation loop. The exception would be a high integrity system such as HIPS but that brings in a whole new set of requirements. I understand your point about the compressor going into surge if the recycle valve doesn't open but I haven't seen credit taken for the recycle valve.

Other possible solutions:

- can you rerate the discharge system to the maximum pressure from the compressor? What's the weak link in the system from a pressure point of view? Can you just rerate that item and take the higher overpressure allowances for your piping allowed under code (if you meet the requirements and if the client allows it)?

 

[Montemayor]

Samiran:

TD2K is exactly correct. His first point is the basic one in considering all over-pressure scenarios: this is an emergency case being considered (if the centrifugal compressor's dead-heading poses an over-pressurization) and as such, it is unsafe logic to consider normal process instrumentation as a safe guard or mitigation of the incident. A PRV is the emergency instrument of design and it must work to fulfill the basic scenario needs at the worst condition - that is the basic scope of this instrument.

However, as TD2K further points out, this is a unique and relatively benign dead-heading of a gas transport machine if we are dealing with a centrifugal instead of a reciprocating compressor. In other words, this is analogous to what is normally done in a centrifugal pump: the discharge system is designed to safely withstand the maximum, dead-headed pressure the machine can put out - not PRV or over-pressurization device is needed UNLESS there is another over-pressurization scenario that defeats this as the worst case. Only a detailed and meticulous process analysis can yield the safe answer to your application - and that is something we on the forum are unable to do for the obvious reasons that we don't have all the details and the facts.

Normally, what an original design would have done (if the dead-heading is the ONLY worst scenario) is that the discharge system would have been designed with a MAWP (maximum allowable working pressure) in excess of the head-head. But, to repeat myself in the interest of safety, this is assuming the parenthetical phrase, above. We don't know the safety-related effects of a compressor surge or of the dead-heading in your application, and these have to be evaluated and handled - if need be.

I hope this helps to reinforce TD2K's response.

Art Montemayor
Spring, TX

 

[Samiran]

Thanks TD2K & Art for your comments. They are well appreciated.

TD2K has mentioned HIPPS. Yes, the compressor discharge has HIPPS protection and above this it has discharge PRV protection. I guess both have the same SIL but can we eliminate the PRVs if we provide HIPPS ? I am aware Offshore facilities adopt this kind of philosophy but that has more logic.

The max head can be determined based on the wheel map but the problem is pressure. Suction pressure rises in this case which means at the "dead-head" we can still get a pressure much above relief pressure which is set point + 16% overpressure ( multiple relief).

The problem at hand represents a rerate case of a compressor wheel. The main problem is the compressor when it sees a blocked discharge. If the recycle valves do not open of this multi staged compressor the compressor goes into deep surge -which means we would probably damage the machine unless HH Vibration trips it out on a timely fashion. The code requirement is that we cannot take credit of the recycle flow (which TD2K has rightly pointed out the normal position prior to blocked case will be CLOSED).

This has resulted in a process design challenge for this rerate case. On one hand we have a fixed flare capacity and on the other hand we have code violation. We, as process/unit engineers, know very well that enough protection exists but to put it down on paper becomes difficult.

One solution adopted by a different facility is they have installed PRVs across the recycle valves which then meets code requirements and have done away with the discharge PRVs which has resulted in tremendous savings because of reduced flare sizing. But this assumes dynamic simulation predictions which is catching up with industry quite fast.

This is a diifficult and interesting topic. I would still invite comments like TD2K & Art have.

Samiran

 

[TD2K]

If you try to put the compressor flow in a blocked discharge down the flare, what happens? Do you get excessive back pressure, velocities, radiation, etc? Just because your flare system has a nominal capacity doesn't mean it can't handle higher flowrates, it simply results in more backpressure back in the unit.

If the relief valves operating on the compressor are unrelated to other upsets, and the flare will physically handle the flow, you may be able to argue a) it's very unlikely to happen because of the 'enough protection' in the design you mentioned and b) if it does happen, the flare can handle it AND nothing else has to get into the flare at the same time (because you normally don't design for double jeopardy). Now, if it's part of a system wide failure, then the additional flow from the compressor may result in too much backpressure for other relief valves to get into the flare system and that would not be acceptable.

I'm somewhat surprised if this is a blocked discharge case that your flare isn't designed to handle that. Blocked discharge should just be the normal flow rate through the system and isn't that potentially/actually going to go to the flare during startups and shutdowns of your compressor and other equipment?

 

[Samiran]

TD2K,

The flare design IS based on the blocked flow discharge. But I mentioned a rerate case where you can change a compressor wheel but not the entire flare piping , tip etc !
We are struggling to defend ourselevs in convincing ourseleves of the adequacy of the flare system by saying that adequate protection exists but then we cannot comply with code. This flare is a common facility among various Trains hence even if double jeopardy is ignored, a startup or shutdown scenario can be imagined for a different Train when a blocked case happens on another Train. We have contacted John Zink and they have come up with a number which the flare tip can handle at the maximum. Beyond this mechanical problems like lifting of mechanical seal etc can happen which needs further study. Radiation can also be a problem.

Our solution to all of these problems is to establish that during a blocked discharge the compressor will relieve much less flow than we think.

I am not sure what you meant by your last statement. I am not sure why startups and shutdowns (not compressor blocked flow case shutdown) would result in normal flows. Usually we reduce rates before any of those planned events, don't we elsewhere ?

Samiran

 

[aikmeam]

Samiran,

Correct me if I'm wrong. You've been having prob with the discharge side of the compressor ONLY?

Blocked dischage can occur anywhere along the fluid transport line. And that includes the lines prior, leading to the compressors.

In those instances, there's no credit of anti-surge valve and it's the same relief flowrate which the existing flare has to take. Isn't that a prob as well?

So, if it is, a new flare (or flare system modification) seems inevitable.

 

[Samiran]

aikmeam,

I guess the situation what you're trying to describe is not the situation in this case. This is not an "in-line" compressor which increases the pressure in a transport line.

This is a process compressor where vapor generated from chillers are compressed and condensed and distributed back to chillers ( as you'd find in a refrigeration loop). Here blockage of suction ( which is infeasible )will not lead to a full blocked discharge flowrate of this multi-suction compressor.

Any way I appreciate the problem you're trying to indicate and you are very right in identifying that.

Samiran

 

[tkdhwjd]

My opinion...

Kickback cannot be used to offset the maximum flare rate in case of the blocked outlet contingency. I will assume that the safeties are located upstream of the discharge kickback line (since k/b's are normally desuperheated). Therefore, if an operator inadvertently closes a discharge block valve, 2 things will happen:

1. Discharge pressure will quickly reach RV set pressure
2. Compressor will speed up to meet the increased discharge pressure

It is likely that the RV will lift first before the kickback valves, and in this scenario, your flare system will see a full discharge flow. Of course, the kickbacks will quickly open, but as others stressed, this credit cannot be used to downsize the safety requirement. The flow can be calculated based on your compressor wheel curve analysis at the relieving pressure/temperature.