2 replies to this topic

[Dwake]

Hey folks,

I'm doing a university design project to separate 1-butene and 1-pentene using distillation, and I need to find data to support my fluid package choice in Aspen Hysys. I'm trying to learn how to use Aspen Properties to compare data obtained using these different fluid packages.

 

There is no experimental VLE data for the 1-butene/1-pentene binary available to me, since it isn't the NIST or Dechema databases included with Aspen.

Some questions, which would help me understand this program better (I simulated these graphs using PR):

1) How is it still able to generate a a T-xy or P-xy graph without experimental data?

2) Under Parameters -> Binary Interaction I'm unable to view anything, but I'm able to view parameters for other systems like methane/ethylene (presumably because there's existing data). However, the categories "MLQKIJ-1 and RKTKIJ-1" still appear blue (but they're blank). What do these categories represent?

Thank you!
 

[MrShorty]

According to this: http://www.diquima.u...elsV732-Ref.pdf(I hope this is not a breech of copyright) MLQKIJ is a viscosity mixing rule parameter. RKTKIJ is a mixing rule parameter in the Chueh-Prausnitz liquid molar volume model.

 

As for how to generate PTxy data without experimental data?

 

Using the PR EOS (or RK or SRK or other phi-phi model based on a Van der Waals type EOS with a Van der Waals type mixing rule), one could assume the kij for the Van der Waals mixing rule is 0.

 

Or, using any other technique you like, make a good guess at the kij for this system (I am not familiar with Aspen, so I cannot say why it won't let you see what it is using for kij). I think this is what would be going on in something like the "predictive SRK" model mentioned in the linked document.

 

Looking outside of the phi-phi models

 

One could assume Raoult's law.

One could use a gamma-phi model and assume "ideal" interaction parameters. (NRTL tauij=0 or Wilson lambdaij=1, for example).

One could use UNIFAC (again, I don't know Aspen and whether UNIFAC is one of their built in activity coefficient models, but I see it listed in the attached document, so I expect it is).

 

In the absence of experimental data, we can make these kinds of assumptions or use these predictive models. We will still have questions about our assumptions (how good is the assumption that kij=0 in the PR equation or how good is the "solution of groups" concept that UNIFAC is built on). A lot of our confidence in the resulting VLE calculations will depend on our confidence in the assumptions and predictive model(s) that were used.

[serra]

for a separation you should verify that selected models (for fugacity and enthalpy) give proper results (at operating t, p),

best option would be data regression (experimental data points as input) , when no experimental points are available and VLE is far from that predicted by a EOS with BIPs = 0, you may generate (low pressure) points with predictive models such as UNIFAC and use these points as input for data regression procedure  (but accuracy may be questionable...)

another possibility is to consider a model based on a EOS + UNIFAC or equivalent predictive method,

(my copy of Prode Properties includes several models based on EOS + UNIFAC, possibly there are are similar options in your software).