8 replies to this topic

[viral desai]

Hi

I am a new bie for this forum.

Often it is said that when the simulation results prediction is not good we have to regress the binary interaction parameters.

Can anyone elaborate this?How to do regression in aspen plus?What data are required


regards
viral desai

[herrani]

Hi viral desai

The precise use of the binary interaction parameters depends on the actual thermodynamic model you are using. However, they are just one additional parameter that is added in the calculation of the parameters of the mixture.

When the results of thethermodynamic model are different from the actual experimental data, you can use this parameter to account for the interactions between the binary pair that have not been taken into account in the thermodynamic package.

Ususally, you can find these parameters in the open literature if you are working with some common thermo package, and mixtures.

As far as I know, it is not possible to actually regress these parameters using aspen plus directly (you might have to write a VBA code instead), but I am not sure on this point. If I remember correclty, it was actually possible if you used aspen properties though.

The information you need is experimental data on Vapour-Liquid equilibrium on the binary pair you are interested in over a wide range of temperatures and/or pressures.

[abhi_agrawa]

Viral,

If you have the Vapor-Liquid equilibrium data for the binary mixture, you can regress the data by using Aspen Plus. You can not do data regression in Hysys. Alternatively, you can estimate the BIP's by writing your own program.

Alternatively, if the BIP's are available in any databank like Dechema etc. you can input them in Aspen Plus.

-Abhishek

[eghbali66]

Hi
I tried data regression in electrolyte systems. But i always recieved error messages.
is anybody have an example of DRS of aspen plus in electrolyte systems, for example MEA+H2O+CO2?

[marchem]

you can use a different software for regressing BIPs,
for example the student version of Prode Properties has these values for Water-EthylAmine (consider VLE, not reactions)

Attached Files

•  BIPS.jpg   89.82KB   60 downloads

Edited by marchem, 04 October 2011 - 12:04 PM.

[Chesps]

For estimating correct interactions between the components of the mixture when the software does not have accurate interaction parameters, you can use practically estimated VLE (Vapour Liquid) data to estimate interaction parameters (often called BIPs). Also you have to select thermodynamic Property Model for which you want to regress the BIPs. To be sure about the correctness of estimated paramters you have to have enough data points. How to regress them can be found in the help of softwares or Tutorials provided by ASPEN PLUS, CHEMCAD etc.
Dictonary meaning of Regression: The relation between selected values of x and observed values of y (from which the most probable value of y can be predicted for any value of x)
So you are just finding out unknown constants in the equation using discrete practical data to use equation to interpolate it at all desired points to predict VLE.

[chemdoc]

I would be prudent with strong solvating mixtures, the data provided by marchem are quite in agreement with my values but you must select a suitable thermodynamic model.

[Chellani]

I would disagree with "Often it is said that when the simulation results prediction is not good we have to regress the binary interaction parameters". If simulation results are not good, you should check your inputs first. Probablity of incorrect inputs has always been higher than software / database issues. BIPs available in database of most of the simulation tools are good enough in 'general' operating range.

In case you are really interested in regressing, below link and file uploaded in that may help you.

http://www.cheresour...__fromsearch__1

[PaoloPemi]

I would recommend as first step in process simulation to select suitable thermodynamic models for each operation, I agree with Chellani that in many cases a specific data regression is not required and one can utilize the values stored in software when these are available, for models based on liquid activities (NRTL etc.) BIPs are required, differently most tools (for example my Prode Properties) assumes ideal behaviour, the same for EOS models with non-ideal / polar fluids, for example water/ammonia etc.
The availability of a large database is a bonus but one can regress from experimental data, estimate VLE values (UNIFAC) , adopt a predictive model (PSRK) etc.
I case of doubt a VLE PXY/TXY diagram can be very useful, I create regularly these diagrams in Excel and they give me confidence with the model.