Vapor Liquid Equilibria

[7604]

Hi folks!
Does anyone knows how to determine the parameters A12 and A21 fro VLE data i.e. from X and Y?

Cheers

 

[siretb]

you are refering to some binary interaction parameters, for an activity model, I assume. Why don't you tell us what model it is?
generally, you get the parameters by a regression pragram, running the model against actual experimental VLE data. Some software packages, like ASPEN have that pre-programmed. Or find it in a book, like the DECHEMA series.

 

[Guidoo]

Following link shows you how A12 and A21 are determined for Van Laar model based on VLE data:

http://www.geocities.com/geneticoptimization/GOALVLE.html

 

[UmeshMathur]

Hi 7604:

The best reference is "Computer Calculations for Multicomponent Vapor-Liquid and Liquid-Liquid Equilibria" by Prausnitz, Anderson, Eckert, Grens, Hsieh, and O'Connell (Prentice-Hall, 1980). This provides extensive computer programs, which are also available from Professor Prausnitz (U. of California, Berkeley) for a nominal fee, I believe.

You need to learn the fundamentals from this book. FOr the actual parameter estimation, I suggest that you use the activity coefficient regression package that comes with your simulator (so that you're sure the physical properties of the pure components are consistent with what you'll end up using when you do simulations).

The interaction parameters listed in the DECHEMA books, referred to by "siretb", are good only if you also use their physical properties, especially for vapor pressure. Sometimes, these do not quite agree with what your simulator is using.

Also, picking the right activity coefficient model is important. I recommend the UNIQUAC model that, unlike Wilson or NRTL, can handle both VLE and LLE data with two interaction parameters per binary. Wilson cannot handle LLE and NRTL requires 3 interaction parameters per binary.

If you have a VLLE system, be careful as you then need to regress VLE and LLE data simultaneously. Code for that option is discussed in the book but is not printed there. However, it was available from Professor Anderson (U. of Connecticut) a few years ago.

For VLE of gases (i.e., components above their critical temperature) in solvents, you must use the "unsymmetric" activity coefficient convention. This is also discussed in the book I have cited.

If you have binaries in your mixture for which no VLE data is available, you will have to use the UNIFAC group contribution method to estimate their VLE or LLE. This is a whole separate subject. Generally, the UNIFAC method is also available with the major commercial simulation packages. It should be used only as a last resort for estimating VLE or LLE when no measured binary VLE or LLE data is available.

Finally, use the regression package that comes with your simulator carefully and do not extrapolate physical properties beyond their range of guaranteed accuracy. Check your results by doing some multicomponent bubble and dew point calculations to make sure you get good results before starting serious simulation work.

In general, this work requires some expertise and experience and should not be assigned to novices in thermodynamics, especially if you're responsible for plant design or something similar.

Have fun.

 

[7604]

For those following my threads the methods are Van laae and Margules.
Txs siretb,guidoo and umeshmathur.
Cheers

 

[UmeshMathur]

7604:

It is quite feasible to estimate the parameters for simple activity coefficient models such as van Laar and Margules using Excel. The built-in non-linear optimizer (called the "solver") can be used to search for the optimal values of the interaction parameters, based on least-squares minimization of the error in total pressure (Barker's method). Models such as Wilson, NRTL, and UNIQUAC require more work, but it's not prohibitive. In my opinion, you should resort to such home-grown methods only if your process simulator doesn't come with a built-in VLE regression package.

For van Laar and Margules, you only have to provide the pure component vapor pressure equations.

 

[Lblopez]

Hello, I supposed that you have laboratory data for VLE of your mix sistem. If you have acces to simulation package, like PRO/II, CHEMCAD or HYSYS you may do it so easy. In this programs you can obtain BIP for Van Laar and Margules and any that you choice. If you want, I can send you the procedure to do it. Write me for send you the procedure.

Good Luck
Lblopez

 

[7604]

Lblopez
Thanks for your help. unfortunately I dont have access to any simulation package.

cheers

 

[UmeshMathur]

7604:

If you send me your pure component vapor pressures and VLE data, I will happily regress your van Laar or Margules VLE parameters for you as long as you are not in a big hurry. I recommend van Laar over Margules, as it covers a wider range of compositions pretty accurately, even for really non-ideal mixtures, with only two parameters per binary.

 

[siretb]

If you send VLE data for a regression we'll need :
list of ALL components
X (liquid mass or mole fraction), for each component
Y (vapor mass or mole fraction)
temperature and pressure.

 

[UmeshMathur]

A minor technical point to follow up on siretb's message:

If you only have T-P-X data (instead of the full set of T-P-X-Y data), you can still regress the liquid phase activity coefficient model parameters. However, you will lose the ability to do a comparison of the calculated versus experimental Y values, which can often be significant for systems where the vapor phase is non-ideal (e.g., for polar, associating, or solvating mixtures). This comparison is important also for checking the thermodynamic consistency of the data set - see the previously cited "Computer Calculations for Multicomponent Vapor-Liquid and Liquid-Liquid Equilibria" by Prausnitz, Anderson, Eckert, Grens, Hsieh, and O'Connell (Prentice-Hall, 1980).

In fact, failure to model the vapor phase non-ideality can lead to serious design errors, especially for mixtures containing significant amounts of organic acids that are known to dimerize significantly, and so have apparent vapor phase fugacity coefficients very far from unity (in the range of 0.4 or even lower). Also, in such cases, the calculated vapor density - needed for column design or rating calculations - can be too low by over 50%, which can lead to seriously over-sized columns in the design case.

One of the quickest methods of measuring VLE data is using static cells in which the vapor phase composition is not measured. This method should be used only for systems where you KNOW that the vapor phase is close to ideal. This is true for most chemical systems at low pressures (typically below 50 psia), provided you don't have the aforementioned polar, associating, or solvating components to contend with.

The moral of the story is: if you're not sure about vapor phase non-ideality, measure the full T-P-X-Y data set, not just T-P-X. Also, pay special attention to accuracy of measurements in the dilute regions as these affect your infinite dilution activity coefficients the most.