Sizing a Relief (Fire Case) for a Three-Phase Separator

[RWerth]

I am trying to size a PRV for the fire case on a three phase (oil/water/gas) using API 521. We have a customer with an oil stream whose concentrations are fairly well dispersed (C1-C10+), and it can be assumed by the separator design that the vessel will maintain something close to a 50/50 water/oil split. The customer has supplied a laboratory breakdown of the components in the oil and to my surprise also included a phase diagram relating the partial pressure of the oil mixture to the temperature of the system. My immediate thought was Clausius-Clapeyron to determine the heat of vaporization, but upon performing this calculation I am getting heat of vaporization values approximately 7.5 X smaller than the typical value used (150 Btu/lb). I believe this is due to the fact that we are using the total wetted surface area, but are not including the water in the heat of vaporization calculation. Important considerations:

* My company doesn't typically get into process simulation and as a result I have no access to commercial sim software.
* The customer is happy to upsize, and in current system, it would not be a large change as the system operates at high P. The same can not be said for future situations, and I am trying to establish a precedent.
* I am hoping to operate within RAGAGEP (of course) but also don't know if it is practical to assume the whole vessel is filled with oil and use the previously described Hvap. Again, I would like to establish a standard method here, but an unrealistic result could both oversize the PRV and get shot down by the people in charge of purchasing.

My questions:
* From the documents we have received from engineering firms, it seems as though most use a value of approximately 150 Btu/lb. This seems to be an industry standard, but I would like to understand if/why it applies instead of just taking it at face value.
* Are there any other good engineering practices to estimate the heat of vaporization? It seems like I keep circling back to thermodynamics, but the more I investigate, the deeper the rabbit-hole goes. I'm happy to continue this pursuit, but I would prefer to know that it is the only option before building my own thermo flash-calculation simulator .

Thank you for your time,
RWerth

 

[Latexman]

Clausius-Clapeyron works well for a pure component, because there is a unique value of latent heat for a unique boiling point temperature at the system pressure. For a mixture, this gets murky. You have a boiling point range and the effective latent heat is a combination of all the components. If you got 20 Btu/lb from their data, either the data is no good or you did it wrong, especially if water (~1000 Btu/lb) is present. There may be other reasons too.

From API 521, "If no accurate latent heat value is available for these hydrocarbons near the critical point, a minimum value of 115 kJ/kg (50 Btu/lb) is sometimes acceptable as an approximation." I've seen folks use 50 Btu/lb when they had no data, based on this quote.

Are you close to the critical point for some of the components?

Developing a three phase flash calculation is not easy, in fact, it is quite difficult to do from scratch. The VLL equilibrium model will be quite complex and require a lot of data. Binary parameters are needed for each pair of components. I think this is N² - N binary pairs. 10 components needs 90 binary parameters!

I've got Aspen, so I don't have a recommendation on a free simulator, but I recommend you look for one. I think there are some out there. Some are even open source/code. You may get some recommendations from other folks.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water.

 

[apetri]

if you are not sure about thermodynamics better to consult a specialist,
as mentioned by Latex, my copy (sixth, 2014) of API 521 in 4.4.13 (Fires) mentions a minimum value of 50 btu/lb for hydrocarbons near the critical point, API also suggests a dynamic simulation (with sequential / direct integration etc.) to evaluate how composition and fluid properties change, with hydrocarbons + (high fractions of) water you can obtain a phase diagram showing a separate water line

calculated with Prode Properties for a mixture C1-C15+water, the water is the last component to evaporate which means high T, P (high stress),
for these reasons (in my opinion) it's better to evaluate the different conditions instead to adopt a predefined value.

 

[RWerth]

Thank you for the replies Latexman and apetri,

Latexman,
I did not make it clear in my original post, but the data given did not include water. It only analyzed the composition of the oil, and that is where the conundrum began. I considered the fact that this was a mixture and not a pure component, but believed it to be an acceptably accurate approximation after finding a 2016 paper by Abernathy and Brown in which the method was used to determine the Hvap of 87 Octane gasoline. The study resulted in values that deviated ~5% from literature values. The PRV is set to open at 500 psig, and while I don't know exactly what temperature that will occur at without more VLE data, I suspect that it is not quite in the critical point range (feel free to correct me if that is a bad assumption). I feared as much regarding the development of a three phase flash, and might just need to lobby for an Aspen (or equivalent) license.

apetri,
While I agree that a specialist would be best, unfortunately it is not an option (out of my hands). Not sure if you were suggesting it, but I did consider using Prode Properties as a free thermo simulator. I ran into an issue when I saw some documentation that stated it was not to be used commercially. Again, trying to set a precedent, I want to make sure the results I find are 100% verifiable and do not violate any copyright or other legal claims. I agree with your opinion that it is best to look at the entire situation as opposed to a snapshot. The plan was to find the point with the lowest Hvap (highest vapor flux) and use that to size the orifice.

Again, thank you for your time and help. I've been through quite a few posts and have seen many replies by both of you. Your contribution is recognized and greatly appreciated.

 

[apetri]

I have Prode, as for similar products there are some limits with the free versions, anyway the price is low (~200 $ with 1600 chemicals),
as mentioned by Latex there is also free software as for example Coco simulator etc. but I have not direct experience

 

[Latexman]

Is DWSIM free? I've heard of that.

RWerth, I bet $200 is only 2-3 hours of your time.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water.

 

[RWerth]

Agreed. I was not aware that the licenses were that reasonable. A request for info has been sent!

 

[georgeverghese]

The firecase relief option in the batch depressuring unit operation in Pro II / Simsci (or similar in Hysis) is ideally suited for this. But both these simulators dont do well with oil-water mixes. I've often found relief valves sizes obtained from these routines to be smaller than that predicted with thumb rules in the API. See what results you get with 100% oil. Obviously you will have to use pseudo components for the higher mol wt components in the oil stream, so these should be listed out in the Owner's characterisation of the feed stream oil phase.

 

[Latexman]

One could also select the one component which has appreciable concentration and results in the largest relief area. Then, assume this pure component fills the separator to the normal liquid level. It greatly simplifies the problem, and is conservative.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water.