7 replies to this topic

[moein_omg]

Dear all,

 

I am developing a computational code capable of handling different hydrocarbons including flash operation. Its ultimate goal is to be directly integrated to simulate different processes in a software.

I am facing a problem in early stages of project: when heavy hydrocarbons (C10 and higher) are included in the mixture equation of state (either Peng Robinson or SRK) fails to produce releasable results comparing to softwares like hysys e.g. it results in one phase rather than two.

 

It should also be mentioned that I use a classic scheme of PR and SRK.

How can I remedy this problem? 

 

Your precious advice in this problem would be greatly appreciated and helpful.

 

Regards

 

Moein

[Cheluget]

Are you using similar values for the binary interaction parameters as are in the simulators?

[moein_omg]

exactly the same values for binary interaction parameters as well as critical properties and ecentricity.

I personally hazard a guess that simulators use a more advanced scheme comaring to ours. I also use an advanced peng robinson scheme where alpha parameter is estimated based on experimental data and not calculating from ecenteric dependent function. I found similar results.

maight introduction of an molar volume correction parameter (e.g. peneloux method) remedy this situation?

[serra]

moein_omg,

presuming you initialize correctly, have correct math and so on,

as far as I know the best way to verify your converged values
is to calculate a phase envelope and compare values against
T-P (isothermal), H-P , H-T etc. flash operations
see post #8 (from marchem) in this thread

http://www.cheresour...-in-excel-2010/
 

with the help of a phase diagram you can verify if your

calculated values are on the left or the right of reference values

(which may be acceptable if you adopt some different model)

also a phase diagram can help to verify if convergence is reliable

in critical area (near critical points) or, for multiphase,

near liquid-liquid separation lines,

 

however,

if you plan to write industrial code keep in account that
commercial products are the result of many years of development,
not easy to replicate the features of these codes...

Edited by serra, 13 August 2014 - 02:24 AM.

[moein_omg]

Dear Serra,

thanks for your comments.

Actually I have validated my results against hysys results for C7 and lighter hydrocarbon for vapor/liquid equilibrium as well as phase envelope and densities. Although slight deviation exists , the trend are quite satisfying. The problem arise when I go for heavier hydrocarbons and the heavier the hydrocarbon, the worse results.

 

I also have to admit that I cannot duplicate codes like Hysys as there is long hystory of efforts and experiences behind it.

[serra]

it seems that you wish to replicate the results
of a commercial code
but in my opinion, if you know the matter,
and you must know to write industrial code,
you should be able to validate yourself the code
and, only as last step, compare against simulators.

To validate a model, you may, for example,
evaluate densities, fugacities and derivatives (Gibbs),
note that if a model based on PR works (i.e. predicts
correct densities and fugacities) for light
components it should work for heavy components, too.
If you have doubts about calculated fugacities try
to validate yourself the values (see above)
or you may validate against a different code.

Possibly you may get the values from a simulator,
the software which I utilize, PRODE PROPERTIES,
exports fugacities and derivatives vs. T,P,W
(temperature, pressure, composition)
and, when I write a new model I can easily compare values.

Once you have validated values of fugacities
you have to solve other problems as proper
initialization (with phase stability or equivalent method)
and converging the problem.

Commercial codes do include complex procedures
to initialize values (there are many papers
discussing the matter), this is an area where
a in-house code may result a bit limited compared
to commercial codes.

 

Finally, as last step, you may utilize the procedure

based on phase envelope (see my post #4)

to verify convergence in critical areas.

[Cheluget]

For the Peng-Robinson EOS, HYSYS uses the (1978) form (as opposed to the 1976) form - the difference is the alpha function form and the two equations are identical until an acentric factor of 0.49 (corresponding to n-C10) - this would explain a difference for PR EOS if you are using the 1976 form, but for the SRK I do not understand...

[moein_omg]

cheluget,

the differnce that you mentioned is for heavy highdro carbons (ecenteric factor > .5) where a different alpha function equation is used. I have tried both schemes but the same results were obtained. I even fitted an alpha function (as a function of reduced temprature) for each heavy component. I also tied different ecenteric factors from different resounces but all resulted similar.