Sizing PSV for Oil/Water/Gas Separator 1

[Kynon]

I am currently trying to size a relief valve (or valves) for a blocked-outlet/full flow case on a number of three-phase separators, for installation on a new-build production platform. Our client has given us a number of potential feed flow cases, with some physical property data for each of the constituent phases - vapour phase, liquid phase, and aqueous phase. The data includes viscosity, MW, compressibility, density, Cp/Cv and mass flowrate. No temperature has been specified, so I am currently assuming that the properties, as well as being approximate "bulk" properties (since the liquid/vapour phases are oil & gas from a production well) are at the upper operating temperature.

All so far so good, however I am a little unsure how to proceed with determining the required valve size, as my previous PSV sizing experience predominantly involved either vapour-only or two-phase LPG relief - I've never come across what appears to be three-phase sizing, and I'd be grateful if anyone with experience of this sort of process/scenario could help out.

There are 16 different cases, so I won't give details of the whole lot, but here is a "typical" one to give an idea of the properties:

Vapour phase:
viscosity: 0.0124 cP
MW: 20.81
Z: 0.9583
Density: 13.09 kg/m³
Cp/Cv: 1.291
Flowrate: 37590 kg/hr

Liquid phase:
Viscosity: 90-140 cP
MW: 238.2
Z: 0.167
Density: 860 kg/m³
Cp/Cv: 1.114
Flowrate: 251500 kg/hr

Aqueous phase:
Viscosity: 90-140 cP
MW: 18.02
Z: 0.011
Density: 988.8 kg/m³
Cp/Cv: 1.16
Flowrate: 303900 kg/hr


At present I have two possible approaches that seem reasonable to me:

1) Input the property information into HYSYS, and use the in-built PSV sizing functionality to determine the required orifice area.

2) Assume the aqueous & liquid phases form an emulsion, with physical properties that are a weighted average of the two individual phases, and then consider 2-phase flow along the lines of DIERS (Omega method).

Am I on the right lines with either of these? I've tried the first approach, which is suggesting a "T" orifice, however I'm naturally concerned about undersizing the valve, even though a "T" is pretty big. Also, I haven't used HYSYS for this sort of thing before, and have no idea as to its suitability.

Any advice?

 

[Latexman]

Yes, you are on the right track. Will the oil and water mix to form an emulsion? They are in the viscosity range that I think they would, but the proof is in the seeing. If they do form an emulsion, that's your one liquid phase, the emulsion. Measure it's properties, especially viscosity versus degree of turbulence expected. Then use those with the vapor flow and properties in the 2 phase calc.

Good luck,
Latexman

 

[Kynon]

Thanks Latexman, however your response throws up some further issues for me, at least as far as this project goes:

1) I work for a subcontractor who is only providing some of the Process Engineering and some of the hardware - our Client is the main EPC contractor for the field operating company. Within the available timescale, I doubt we will be able to get samples for any testing of the emulsion properties to be carried out. The Client's specification states that the oil & water will form emulsions, so I intend to proceed on the basis that this statement is correct. This leaves me once again with the 16 sets of data to go on.

2) I don't have any latent heat data for the well fluids, while using the Omega method for the sizing of the relief valves requires this information for flashing systems. However, if I consider the system to be gassy & non-flashing, the latent heat term disappears. Is this a reasonable approach, based on the liquid phase being an emulsion, and also that the vapour fraction of the feed stream is well over 80% by volume in every case bar one? If not, then how to do I estimate the latent heat of the liquid phase?

(Note regarding point 2 - I have done some preliminary calculations based on gassy & non-flashing flow, which has resulted in required areas that are of the same order as those calculated by HYSYS, albeit around 20% larger in each case)

 

[Latexman]

Kynon,

1) I would not hesitate to estimate the required physical properties of the emulsion using appropriate mixing rules, except for viscosity. Viscosity of an emulsion needs to be measured at several shear rates after the emulsion is prepared under application conditions to get the viscosity and rheology. This need not take a lot of time and money. An hour in the lab hood with an inexpensive Brookfield viscometer and you have what you need. The client should be able to furnish this data in a matter of days.

2) In a fire case you need the latent heat. I don't believe there's a way around it and still hold on to your credibility. The client has ultimate responsibility, so they should be asked to provide the needed data, if it cannot be discovered or estimated.

Good luck,
Latexman

 

[MortenA]

But dont you have some sort of time frame - 20 seconds until the ESDV closes - or is it just continuous indefinitely relief?

If there is a time limit then i think that you should check if the vessel will fill up with liquid completely within that time. Maybe it wont?

But you cant input "properties" INTO hysys (to my best knowledge) you must have a composition (of the overall stream).

Best regards

Morten