7 replies to this topic

[kaptoxa]

I'm trying to simulate the process of washing Acetone from HC stream (octane) with water.

I'm using mixer (Octane and Acetone mixture + Water) and 3 phase separator. I've already tried many fluid packages (Wilson, NRTL, UNIQUAC) and it doesn't look like it works. As soon as I add Acetone separator stops showing correct results. It just not separating water from octane 

 

Any recommendations regarding Fluid package (Thermodynamic model) chose?

Or any ways to calculate partitioning without UniSim?

 

Kind regards.

[PaoloPemi]

not sure about the feasibility and the results you wish to obtain but... 
make sure that selected thermodynamic  models in your simulator includes the correct BIPs for the different pairs... otherways try predictive models such as UNIFAC or similar..
 
see as example a liquid-liquid equilibria at 15 C 1 Bar predicted with an extended UNIFAC

 

 separator.bmp   1022.49KB   13 downloads

Edited by PaoloPemi, 21 October 2022 - 03:26 AM.

[MrShorty]

One thing I notice in your question is that you claim that it is showing incorrect results without explaining exactly why you think the results are incorrect or what you believe the correct results should be. From there, you assume that the problem is with Unisim or with the chosen thermo models (all common, well tested models). Assuming, as PaoloPemi recommends checking, that your copy of Unisim is using believable values for the interaction parameters, I would expect your simulator with those thermo models to do at least a qualitatively correct job and predicting the behavior of your separator. Some things I would look at:

 

1) When you say the "results are not correct" what exactly do you mean? How are you determining that they are not correct?

2) As PaoloPemi suggests, what is your simulator using for BIPs and are they believable?

3) (This is one I encounter with clients with some frequency), be sure you aren't letting your own bias or wants interfere with the judgement of what is "correct." I have had numerous cases where the simulator made correct predictions, but they weren't the kinds of predictions that the client wanted to see. For example, in a case like this, the simulator might correctly predict that 5 (random number, I don't know if 5 are required for your specific scenario) water washes would be necessary to get the desired result, but the client was only interested in the process if it could be done in 3 or fewer. In some of these cases, the client will blame the simulator when the real problem is with the client's expectations. Just make sure that when you claim the results are not correct it is because they are really not correct and not because you don't like what the simulator is telling you.

[kaptoxa]

not sure about the feasibility and the results you wish to obtain but... 
make sure that selected thermodynamic  models in your simulator includes the correct BIPs for the different pairs... otherways try predictive models such as UNIFAC or similar..
 
see as example a liquid-liquid equilibria at 15 C 1 Bar predicted with an extended UNIFAC

 

separator.bmp

 

Thank you for the answer.

I was able to ger results using UNIQUAC activity model. The was no BIPs for water - octane pair so I used VVE coefficients estimation. And as a result I got almost equal partitioning of Acetone between octane and water. Can't tell that what I wanted to see. 

Do you think I can trust these results?

 

One thing I notice in your question is that you claim that it is showing incorrect results without explaining exactly why you think the results are incorrect or what you believe the correct results should be. From there, you assume that the problem is with Unisim or with the chosen thermo models (all common, well tested models). Assuming, as PaoloPemi recommends checking, that your copy of Unisim is using believable values for the interaction parameters, I would expect your simulator with those thermo models to do at least a qualitatively correct job and predicting the behavior of your separator. Some things I would look at:

 

1) When you say the "results are not correct" what exactly do you mean? How are you determining that they are not correct?

2) As PaoloPemi suggests, what is your simulator using for BIPs and are they believable?

3) (This is one I encounter with clients with some frequency), be sure you aren't letting your own bias or wants interfere with the judgement of what is "correct." I have had numerous cases where the simulator made correct predictions, but they weren't the kinds of predictions that the client wanted to see. For example, in a case like this, the simulator might correctly predict that 5 (random number, I don't know if 5 are required for your specific scenario) water washes would be necessary to get the desired result, but the client was only interested in the process if it could be done in 3 or fewer. In some of these cases, the client will blame the simulator when the real problem is with the client's expectations. Just make sure that when you claim the results are not correct it is because they are really not correct and not because you don't like what the simulator is telling you.

 

1)  At first the 3 phase separator didn't actually show any separation with any thermodynamic model. I was able to see separation only with UNIFAC + some BIPs estimation. As I replied to PaoloPemi i'm still not sure if I trust these results.

2) UniSim Design. Standart set of thermodynamic models.

3) Agree. Now Unisim shows that I need 400% water to wash Acetone. Not exactly what I expected to see.

 

As a conclusion now I have simulation showing almost equal partitioning of Aceton between octane and water. I was not able to find any experimental data to prove it. The only thing I was able to find is Log Kow = - 0.24 (standart conditions?) which also show almost equal partitioning (but between water and octanol, both polar). Any ways I can check my results?

 

Many thanks.

[MrShorty]

The best check would be measured ternary LLE/partition coefficient data to compare to. You say that you have not been able to find any.

 

Since you are using an activity coefficient model, if you can get Unisim to tell you what it is calculating for the activity  coefficient of acetone in each phase, you can check against literature VLE and/or LLE data to see if you think those activity coefficients are believable.

 

If it helps, a quick look at my limited literature database shows that the activity coefficients of acetone in water and alkanes are about the same, so having an LLE calculation that shows little preference for acetone to be in one phase over the other seems reasonable.

[kaptoxa]

The best check would be measured ternary LLE/partition coefficient data to compare to. You say that you have not been able to find any.

 

Since you are using an activity coefficient model, if you can get Unisim to tell you what it is calculating for the activity  coefficient of acetone in each phase, you can check against literature VLE and/or LLE data to see if you think those activity coefficients are believable.

 

If it helps, a quick look at my limited literature database shows that the activity coefficients of acetone in water and alkanes are about the same, so having an LLE calculation that shows little preference for acetone to be in one phase over the other seems reasonable.

Thank you for your contribution. 

 

Yes, it looks like this simulation is correct.

 

One final question. I'm using 3 phase separator to separate water from octane (vapor phase flowrate = 0 under process pressure) and separation is weird. The lighter octane phase is not separated from water till it's 10% wt saturated with water. So while adding up to 10% wt water separation is not happening at all. Any ways to avoid this?

[MrShorty]

It probably depends on why it is telling you that 10% water is needed.

 

If it is calculating that, at equilibrium, the organic phase can hold 10% water, then one would need to understand why the simulator believes that water is that soluble in the organic/hydrocarbon phase.

 

If it is something else, you would need to understand that reason for needing 10% water, and then figure out what to do differently.

[SilverShaded]

UNIFAC/Uniquac should not be relied on for serious work without tuning the interaction paremeters against published test data for the mixture or binary pairs your dealing with. 

 

VLE data should not be used for LLE simulation, it's known to be unreliable.