Control loop response effect on PSV relief.

[fdomin]

In a crude preheat train a valve is controlling desalter pressure by means of throttling pump discharge.
PSV´s exist both in pump discharge (downstream of control valve)and in the desalter.
When a block of this train occurs upstream of the desalter, the PSV in the pump discharge will relieve.
Which flow should I consider for relief valve analysis?
- Normal flow through pump.
- Increased flow due to pressure control response and control valve going 100% open as pressure in the desalter will lower.

 

[TD2K]

Typically, control loop responses are not considered where their effect would reduce the relief flow, that is, they are assumed to remain in their last position.

If a control loop's expected response would make the relief case worse, then it is assumed to react as designed.

From your write-up, I believe your system is a pump, control valve, PSV, desalter and another PSV. If the PSV downstream of the pump relieves, the relief flow rate is taken off the pump curve at the equivalent head for your suction pressure and the discharge pressure (PSV set pressure + 10% plus dP through the open control valve and piping).

For the suction pressure, some companies use the maximum expected operating pressure, some use the upstream pressure's design pressure.

If piping is the only thing that is exposed to the high pressure and it is designed to B31.3 (which would almost certainly be the case in North America), you might want to consider the flow rate with the PSV relieving at 133% of piping design pressure. The 133% is allowed for short term operation (10 hours per incident and I believe a total of 50 hours per year?). There are some other conditions that must exist for you to use this overpressure allowance and some companies do not allow it.

For the desalter, you will be limited of course to 110% of the vessel's MAWP but if you can depend on both the piping PSV and the vessel PSV, you can use 116% as you have multiple PSVs protecting the vessel (PSV set pressueres must be staggered per code allowances in this case).