separation process

[pmoneysh]

Hello everyone, I am currently working on a project for 2-propanol separation. Feed is multicomponent consisting of methanol, acetone, diisopropyl ether, 2-propanol (97 wt%)and water. We are using two-column distillation process to obtain 99.99 wt% IPA. Now my problem is choosing the right thermodynamic model for the system. Regression of VLE seems quite difficult as it is very difficult to obtain the vle data for these many components(+solvent) used (I tried to look for it). How can I choose the best thermodynamic model to describe this process? So far, I think NRTL model would be good (just my opinion). But is there any way I could verify it?
Any suggestions would be very helpful.
Thank you.

 

[apetri]

kindly post on one group only,
simulators can include many models, the operating manual can help with selection or you can consult books as Thermodynamic models for industrial applications etc.
NRTL can be Ok for low pressures or you can, for example, select some EOS with complex mixing rules, note that you can optimize BIPs to fit measured vle, enthalpy etc. data points...

 

[TiCl4]

In the (limited) experience I've had with model selection, I've researched and used decision trees (like in the link below) to get an initial selection of EOS/activity coefficient model. I suggest you take a look and look up "Thermodynamic Model Selection" on Google to see other papers written regarding model selection.

http://www.cchem.berkeley.edu/cbe150b/docs/VLE/Guidelines.pdf

To verify the model, I imagine you could do the following: Prepare a liquid mixture in the lab at a known liquid composition that is similar to your feed composition. Fill a vessel and heat it to a specific temp. Bleed off all inert gases. Measure final temp/pressure, and take a gas-phase sample. Run the sample through a GC to determine vapor phase composition. Compare those values against a flash calc in ASPEN or whatever you are using to see how well the model predicts the composition at that given temp/pressure.

Repeat this for your the temp, composition, and pressure ranges you expect to see in the actual distillation unit.

 

[Latexman]

Methanol, acetone, diisopropyl ether, isopropanol, and water are well studied and there should be a lot of VLE/VLLE/LLE data available. This data is usually collected and entered for a model as binary pairs. With 5 components I believe you have 10 binary pairs. Look for data in binary pairs. Fit a model to that. Then enter all into your model.

Good Luck,
Latexman
Pats' Pub's Proprietor

 

[georgeverghese]

You could give this a shot in the dark with the NRTL model, but beware that these components are all polar, and azeotropes are very likely, which may not be predicted by run of the mill simulators in the public domain. If this is for some critical design work, get in touch with the plant licensor for simulation run results.

 

[Latexman]

And with the diisopropyl ether and water be sure the model can handle two liquid phases, VLLE.

Good Luck,
Latexman
Pats' Pub's Proprietor

 

[pmoneysh]

Thank you all for the response. @apetri copied "kindly post on one group only,". @georgeverghese, of course, there is azeotrope formation. We have used a solvent (extractive distillation). @latexman sure we took account of that.
Anyway, this feasibility study is for industrial application. That's why I am worried about the thermodynamic model. Going through the comments, I guess NRTL would be okay. Thank you all for your response.