Jump to content



Featured Articles

Check out the latest featured articles.

File Library

Check out the latest downloads available in the File Library.

New Article

Product Viscosity vs. Shear

Featured File

Vertical Tank Selection

New Blog Entry

Low Flow in Pipes- posted in Ankur's blog

0

How To Choose Thermodynamic Package.

property package proii simulations water separation

3 replies to this topic
Share this topic:
| More

#1 donotdisplayname

donotdisplayname

    Brand New Member

  • Members
  • 3 posts

Posted 24 July 2017 - 03:39 PM

Hello all,

 

I recently moved from systems engineering to process simulations.  Please help me with the following basic questions.

 

Have a feed comprising (I-butene - 54%, Isoprene, 6%, tert-butyl alcohol 8%, Pyran <1%, C5Ohs < 1% and traces of formaldehyde (<1%) with remaining water -30%).  Need to recover all the products > 99%. 

 

Question -  As water present in significant amount - whats the optimum method to remove this first.  Noted formaldehyde is highly miscible with water and I need to have nearly 99% recovery.  (Commercial formaldehyde as product - 37 wt %).  

I understand property package defines the interaction parameters, viscosity, enthalpy, entropy calculations etc. 

 

The feed was preheated to XX C in exchanger and directed to a flash drum where vapors in overhead, liquid and water in bottoms. 

 

I tried following of few options:

 

Option A - Using SRK.   Pure water gets separated in bottom of flash drum while remaining vapor and liquid streams obtained as per equilibrium conditions of flash drum. Formaldehyde (that is highly miscible with water goes into vapor stream). The other liquid stream can be fed to a column  for further separation.

 

Option B - Using NRTL (K-Value VLE-NRTL and no other changes for property estimations in thermo package). 

There is NO water separation from bottoms, and  only vapor and liquid. 

 

Option C - Using NRTL (K-Value VLE-NRTL and K-Value LLE-NRTL). 

There is small flow in liquid and MOST of the components in water separated from bottoms

 

I am wondeirng

1 - how do I know which property package to use

2. how do i separate water (>30% wt in feed)

3. basis to use other thermodynamic properties e.g. Liquid / vapor  enthalpy, liquid / vapor entropy, liquid / vapor density etc.

 

Thanks in advance for your replies.  

 


#2 serra

serra

    Gold Member

  • Members
  • 310 posts

Posted 25 July 2017 - 01:10 AM

you are facing different problems,

1) the definition of a proper process (to separate the components in mixture),

2) selection of accurate models (for fugacity, enthalpy, volume plus transport properties) to simulate the process

 

it seems you are considering distillation but you get very different values from different thermodynamic models,

that is normal, it is good practice, with non-ideal phase equilibria, to verify if selected models allow the required accuray,

for fugacities you can verify VLE data, data regression (for BIPs) is required in many cases,

since there is a relation between fugacities and enthalpies this step can result very useful,

 

you mention a few models as SRK and NRTL, 

if I remember correctly formaldehyde can react with water , I don't  know if that is true in your case but for a mixture including formaldehyde, water and a few other components I have used the CPA model (with the reactive flash) available in Prode Properties,

selection of non-EOS based models can be Ok at low pressures (you do not specify separation conditions)

 

if you decide to select a EOS based model (for liquid fugacity) make sure that alpha function has specific values  for water,  formaldehyde etc. the alpha in SRK std. is not suitable.


Edited by serra, 25 July 2017 - 01:10 AM.


#3 donotdisplayname

donotdisplayname

    Brand New Member

  • Members
  • 3 posts

Posted 25 July 2017 - 08:42 AM

Serra, 
Thanks for responding regarding formaldehyde separation.  I have basic questions, if you can please help.
 
How do I decide thermodynamic model for separation? So far from reading material, I understand that - model should be chosen such that  i) vapor and liquid properties fugacity, enthalpy, density, etc and ii) behavior i.e vapor and liqid equilibrium curve is predicted as close as possible. 
Under certain situations e.g mixture of isobutylene, isoprene, formaldehye, C5OH and water; a single EOS of model is not fit to predict right results.
 
As such, we modify existing models (or EOS) to achieve the right results.  So - 
1. Do we specify separate EOS (or model) for vapor and liquid each?
2. You mentioned " can verify VLE data, data regression (for BIPs) is required in many cases..".   How do I check it?
3. What is the difference - when I choose NRTL and update - K-VLE to NRTL and K-LLE to NRTL.   
Similarly how to decide what model to use for each property e.g. Liquid Enthalpy , vapor enthalpy, liquid density, vapor density.
 
If I choose, K-VLE-SRK and K-LLE-NRTL...Does this mean -
a) For K vales between vapor and liquid interaction; for all components; SRK will be applied.  Example - isobutylene and remaining hydrocarbons.
B) For K-vales between liquid and liquid interactions; for all  components; NRTL will be applied.  Example - formaldehyde and water and ketones are miscible?
 
Please share a thought process on how to approach a separation.  I really want to perform this simulation with better accuracy.
 
Thanks in advance.  


#4 serra

serra

    Gold Member

  • Members
  • 310 posts

Posted 25 July 2017 - 10:08 AM

chapter 6 in The Properties and Gases and Liquids gives detailed information about the different models etc.

 

depending from application (components in the mixture and operating conditions) different models may be required,

 

many models require to define BIPS (from a database or data-regression) but there are predictive models (UNIFAC, PSRK etc.)

 

consult the operating manual / help for the models available in your simulator and selection rules,

 

if you have access to experimental data you can verify which model gives the best results,

otherways predictive models may be a solution (but take care of possible errors)






Similar Topics