10 replies to this topic

[staffel]

I am comparing results from three phase flash operation (vapor+liquid+water) with multiphase (vapor+liquid1+liquid2+water) for hydrocarbon mixtures + water in Properties (Prode) , I see that in several cases an additional liquid phase appears (liquid1,liquid2,water) and this can affect vapor amount (found variations of up to 5%), since (generally) I do not take in account additional phases which may have influence on vapor amount I am in doubt if my way to proceed is correct, in my knowledge API TDB procedures do not consider additional phases in addition to decanting water, am I correct or missing some important information ?

[PaoloPemi]

multiphase equilibria is an active area of research with important works from authors as Michelsen,
there are many examples of mixtures originating three (or perhaps more) liquid phases, one is described here

http://www.prode.com/en/multiphase.htm

in my opinion with hydrocarbon mixtures + water things are complex because models as Soave-Redlich-Kwong or Peng-Robinson when coupled with traditional (van der Waals) mixing rules may fail to describe accurately the behaviour of the mixture, several authors have introduced a second parameter (see Kabaddi-Danner formulation) others as Huron & Vidal , Wong & Sandler have defined complex mixing rules able to predict the behaviour of mixtures including polar fluids as water.
If you have defined one of these models you should expect a reasonable accuracy, better than usual approach to calculate water as pure (separate) phase (see API procedures).

[staffel]

I have some literature (mainly from SPE) about multiphase but little practical experience, our models take in account gas+liquid+water.
With multiphase I see that appearance of new liquid phases produce modifications in phase compositions and phase amounts of both vapor and liquid phases and this may affect the process, for example a 5% variation in vapor fraction has a impact when modeling a tower, heat exchanger etc.
A point which is not clear to me is the time and specific conditions required for the new phases to separate, I would suppose it's a process similar to decanting (by difference of density) , if that is true then under the conditions of relatively high turbolence in piping we may consider liquid phase as a homogeneous mixture.

[PaoloPemi]

in many equipments there are the conditions for two (or more) liquid phases to separate, consider for example a three phase distillation where two liquid phases and a vapour phase coexist

[Chellani]

I am still trying to understand exact problem but following info may be useful to you. Procedures in API TDB estimate water soluble in hydrocarbon phase. If I understand correctly, HYSYS calculates amount of water which is soluble in HC phase using temperature and composition of HC phase; if total water in stream is more than this quantity, aqueous phase is created. There are few more rules on top of this for few specific systems but general procedures looks as per above rules.
I also believe that procedure in API TDB considers only VLL equilibrium i.e. there can be only two liquid phases i.e. HC and aqueous. In actual world, there could be more than two liquid phases but that is not being simulated in HYSYS.
After these phases are defined, separation is based only on this rule and density of streams doesn't play any role e.g. if three phase separator would route HC phase to light liquid and aqueous to heavy liquid irrespective of density.
Dynamics can be more tricky and other complications are being modeled using work-arounds and even in other simulators.
Simulating specific unit ops for two liquid phases can be more complicated and user should explore option available in indivudal ops e.g. columns should use sparse cont solver if liquid phase splits on stages as legacy ones would fail for this. Correlations for pipe should handle phase splitting and for exchangers things are more complicated. I don't think even HTRI handles two liquid phases.

Edited by Chellani, 22 January 2012 - 09:53 AM.

[staffel]

thank you for comments
I suppose multiphase requires much more calc's and that would slow down considerably execution time as, for example, with mixtures of many components, that could be the main reason why only a few sofwares do consider more than two liquid phases.

[PaoloPemi]

convergence speed is (in general) slower in multiphase flash (compared with two phases flash), consider a typical mixture of hydrocarbons with 15 components,
Prode Properties will solve the isothermal flash in a fraction of second (except for points very close to critical condition),
if you require a vapor-liquid-liquid flash you may find two or more liquid phases but with the cost of 1-4 seconds (in my portable with 2.5 GHZ processor).
if you require a vapor-liquid-solid flash the program does a series of local minimizations to find a stable vapor-liquid-solid equilibria, you may discover one or two liquid phases plus solid but this may require 2-5 seconds or more.
If you multiply these for the number of operations in a large plant it is evident the final cost in terms of time.
With the actual technology I think that multiphase flash is useful for testing one specific operation and I can do that with Prode Properties and Excel.
Still there are examples as pipelines simulation where it is extensively applied.

[Chellani]

It has to affect calculation speed, how much and would that be cause of not modeling it; I am not sure about this.
I haven't looked back into the entire history but I believe HYSYS was derived for refining (as it was and still is biggest market for simulation products; correct me if I am wrong) and I don't think we get more than two liquid phases in refinery.
Design for dynamics of HYSYS was even worse. Probably developers thought nobody would model solid and aqueous phases together (or probably nobody would model solid phase) and they fixed number of phases to be three because of which we are still trying to find work-around when there are four phases.

[staffel]

PaoloPemi,
thank you for the data, I di some tests on a 24 components mixture and convergence times are about the same.
Surprisingly, when I add water I see that solid model predicts freezing at reasonable temperatures (at least in my test cases), could you provide some details about the model ?
Actually I have the free version of Properties with limited technical support.

[PaoloPemi]

The base version of Prode Properties which is free for students includes a simplified solid solution model where the fugacity of solid phase is calculated from liquid fugacity and enthalpy difference as calculated by selected model.
I have applied this model to verify wax precipitation, it requires proper data fitting, Prode Properties has a utility to fit vapor/liquid , liquid/liquid and solid/liquid experimental data points, you need to spend some time to identify the correct BIP parameters in order to get reliable results.
Extended versions of Prode Properties can include two additionsl models for solid phase
-a derivation of van der Waals and Platteeuw for Hydrates
-a multilayer solidification model which I do not know.

[staffel]

thanks for the information,
I did some tests for Solid-Vapor and Solid-Liquid equilibria with CH4 , CO2 system comparing results against experimental points,
SRK model for fugacity + SSM (the only available in my version) for solids,
errors are reasonable (not much larger than GPA software) considering that I have not tuned the model, with Benzene-Methane results are poor (but they may improve tuning the model) It would be interesting to have something good for chemicals.