delayed coker relief rate

[ReliefGuy]

For those refinery engineers with Delayed or fluid cokers, I'm interested in basis used for releif evaluations from main fractionator. The Sengupta and Staats approach leads to relief rates in excess of normal drum vapor rate (plus emergency steam). While this approach has technical merit, it leads to oversized relief valves for a variety of reasons (esp. drum self quench). Using the drum rate alone, discounts the potential impact of column liquid hold up considered by Sengupta and Staats.

Has anyone pursued rigorous reconcilliation of Sengupta/Staats with real performance by either dynamic simulations, or other analysis?

 

[ifreeman]

ReliefGuy,


I have done similar calculation for Xylene Column before.
Here goes the results of that calculation

-------------------------------------------------------------------------
Condenser Feed(normal drum vapor rate you mentioned) : 257351 kg/hr

Sengupta Method : 231022 kg/hr

Dynamic Model(Using HYSYS) : 275513 kg/hr

*As the reboiler was fired heater(not steam reboiler), I used constant heat input for the column to be on the safe side.
-------------------------------------------------------------------------

I think these differences come from the intermal stream asuumption for sengupta method and latent heat difference
due to relieving pressure(for dynamic and sengupta) at which the relief rate is calculated.
As you know, higher the pressure lower the latent heat.
With same heat input to the column, we will see the more vaporization(mass basis) at higher pressure.

My conclusions are...
1)It is the heat input rate to the column which dominates vapor generation. (In your case, vapor from coke drum. So I think you need to check the vapor rate from drum during relief.)
2)For Sengupta, the inventory stream assumption will affect the relieving rate if the feed to column has wide boiling point range.(I think this is your case.) Therefore such as crude distillation unit, FCC Main Fractionator or your case should be modeled by dynamic simulation instead of Sengupta or conventional method.

Regards,
ifreeman

 

[ReliefGuy]

Thanks for your thoughts. I've been reluctant to go through the rigor of dynamic simulation since the expected behavior of the drum vapor generation in this situation (lost feed and raising pressure) is poorly chracterized and clarely dominates. Thanks for your help.