Flash drum separation calcs

[HarryB88]

Hi all,

I'm hoping someone can point me in the right direction with the following.

I'm calculating the separation of a flash drum for use in a produced water treatment facility (oil refining). The water has been through primary treatment and the components left to be removed are some volatile organics (o-xylene, m-xylene, e-benzene and 1,2,3trimethylbenzene).

In order to calculate the degree of flash and mol fractions/flows I have utilised the Antoine equation to get my ki values and Rachford Rice equation to find my percent vapourised. The drum will operate at 77 degrees Celsius and 402 mbar-a (I have found data that will fit in these ranges).

The problem I have is that water makes up over 99.99 % of my feed (mole %) and hence it dominates the calculations. This hence makes my fraction of feed that is vapourised massively incorrect. After the flash drum the vapour will be condensed and pass through to a vacuum system and after flash has occurred, sent to flare. Can anyone advise on how it would be best for me to proceed? Some assumptions perhaps? Thank you in advance.

 

[sshep]

Hey Harry,

I think a Henry's law assumption would be better than Antoines at such concentrations. You could also simplify your flashed vapor fraction to just assume water for any significant vaporization.

If the hydrocarbon concentration is enough to form a 2nd liquid phase then it will totally change the nature of your flash calculation.

Some real process conditions would make it easier to comment.

best wishes,
sshep

 

[jamesbanda]

Sshep is right

When you have dilute solutions you need henrys law.. but heres the rub.. I've stuggled to get henrys law constant data..So no point using it unless you have the constants - it wont change you answers ! .. that sounds obvious..but i've seen many simulations with people confident they have used henrys law.. then i ask thats great..where are your constants and did you regress / measure them.. or get them from a databank.. then i get a blank look !
if you have aspen 7.3 the NIST engine has a lot of data which can be regressed..

another way..its not ideal.. but an EOS model is better than idea because you dont need henrys constant data..it wont be as accurate but a good start point..

another method i've used. (physcial property people may look on with horror here..) but if i dont have any Henrys constant data i find a similar chemical and use that..for a estiamte to bound my solution.. it is better than no data..again used that with the EOS to bound my solution..

i've probably not helped loads but at least give some background.

 

[HarryB88]

Hi both,

Thank you for your input sshep and James. Apologies for being vague on the process conditions, I opted for assuming the feed was entirely water and the calcs started to behave. Thanks again for your help guys.

Harry