Propose Best Practices

For developing an adequate model, three stages need to be valicated: 

At first, data must be available When data exist, they may be unreliable or far away from the operational conditions. For this, it is necessary to perform concsistency analysis: internal consistency and external consistency. 

The second step consists in extrapolating the data to the operational conditions. This can be done in several ways: - use of graphical presentations to understand the trends of the exixsting data - use of theoretical models that use fundamental physical principles. 

Finally, a model that is available within a commercial simulator should be calibrated to the best possible data: those that are consistent and close to the operational conditions.
The number of parameters is often large in this stage, so that simplifying assumptions may be needed to optain the best results.

The JIP aims at promoting the investigation of elevtrolyte thermodynamics in view of practical applications.
In a previous edition of EleTher, following issues have been investigated:

Data analysis:
trends

Considering that data rarely come in the conditions of industrial applications, it is important to start by analysing trends so that educated extrapolations can be made. 
As an example, we have investigated the mean activity coefficients of the alkali halides as a function of temperature. 
To this end, Bromley's equation was used in order to reduce the full MIAC curve into a single number.


Doing so, following trends are observed:
Read more here

Parameterization

Parameterization implies several steps that are of crucial importance:
  • Selection of the data
  • Construction of the objective function
  • Selection of regressed parameters
The eNRTL model was used to parameterize the ternary system containing water, methanol and NaCl. It was observed that the model could describe equally well the Mean Ionic Activity Coefficients, the Osmotic Coefficient and the Vapour-Liquid Equilibria. The enthalpies of mixing however required more care and the optimal parameter set was not the same. 

The regression tool used allows visualizing the sensitivity of the objective function towards the parameters. This plot clearly shows that the parameters are degenareted: a valley of equivalent solutions exist:

Speciation

In the ternary system containing water, acetic acid and sodium hydroxide, the reactive flash allows visualizing the species as a function of water content.
The parameterization might yield very different pictures. 

An example looks like this: