Common Header in Multiple-PSV Installations for Reciprocating Compress

[Trukero]

I am trying to calculate the proper header size for a multiple-stage reciprocating compressor relief device system. For this reason, I need to define the proper inlet conditions for the multiple PSV’s required for my application. These devices are generally attached to a common header. Quoting API 521, 5.4.1.3.1:

“Common headers systems and manifold in multiple-device installations are generally sized based on the worst-case cumulative required capacities of all devices that may reasonably be expected to discharge simultaneously in a single overpressure event”

It is my interpretation that the API 521 leaves open the possibility to considerate whatever is the required relieving capacity, based on good engineering judgment. In this particular application, each PSV is protecting each stage (for example, one PSV per stage). It would be reasonable to say that all PSV’s for each stage cannot relieve at the same time. Am I correct?

Is there any other code that regulates this type of design (e.g. ASME)?

Final notes:
- The fluid is natural gas.
- The PSV’s are already sized.
- The gas source comes from a single stream/inlet, and it is through the different compression stages that the pressure condition increases.

 

[dcasto]

If one relief goes off, how will any other relief valve be able to get a source of gas? In a fire case, there maybe a an instantaneous amount of gas over the maximum design rate.

So, yes it is reasonable for multiple stage units to size for the largest capacity on that unit. Now, could it be reasonable to expect that building with 10 units having all 10 units have a relief valve going open at the sametime making the header even larger? I'd look at maybe 1 or 2 adjacent units having multiple relief valves going off at the same time.

If you have a relief valve on the suction to the compressor station sized to protect the inlet line, it will be very large and it should have its own header or, it will dictate alone the size of the header.

 

[zdas04]

On a Multi-stage compressor, a high suction pressure can result in subsequent stage over-pressure. For example, if the normal suction on a 3-stage compressor is 15 psia, first stage discharge is 60 psia, second-stage discharge is 240 psia, and third-stage discharge is 960 psia. Lets say the MAWP of the first-stage discharge bottle is 250 psig, the second stage discharge bottle is 600 psig, and the third stage bottle is 1,400 psig.

Say all the kills are controlled through a PLC that develops a fault (one failure). The high-suction kill is out due to the failure and a normal upstream event causes the suction pressure to drift to 30 psia, first stage discharge goes to 120 psia, second stage discharge goes to 480 psig, and third stage PSV goes off before the stage can get to 1920 psia. In this scenario, the third-stage PSV going off does not impact the other stages, and if the suction pressure continues to increase, the second-stage PSV will go off in a minute, followed quickly by the first-stage PSV.

This is a bit made up and is a really good argument for not running all the kills through a PLC, but in this scenario the header would have to be sized for all three PSV's.

In a more reasonable scenario where the pressure/temperature-kills are independent, you would probably be justified in sizing the header for the largest credible volume from a single PSV.

David

 

[dcasto]

most compressors I work with would never run when the suction pressured doubled, the horsepower would double and the engine or motor would stop.

 

[Trukero]

Based on your comments, I presume that no additional codes or norms define the criteria to select how many PSV can be open for a particular process, which in this case is a recip. compressor. I am agreed with zdas04 and dcasto, it is good idea to calculate the flare line based on certain amount of PSV opening but not all of them at the same time.

Thanks