Welcome to the Aspen Plus V8.6 teaching module on Regression of UNIQUAC Binary Parameters from Liquid-Liquid-Equilibrium Data. For information on navigating this module, please refer to Navigation Hints located above the slide. Click the Next button on the bottom right hand corner to begin.

In this example we have a system containing ethyl-acetate, water and ethanol.

We show how to determine binary parameters for the UNIQUAC activity coefficient model using liquid-liquid equilibrium data.

You can find the UNIQUAC Activity coefficient model details in the Online help. The model equations are referenced in the online help and this is the part of the equation related to binary parameters.

The table shows how the UNIQUAC parameters used in Aspen Plus (parameter name = UNIQ) relate to the parameters used in the expression above.
Binary LLE data available cover a range of temperatures (x y?), so we will be regressing both UNIQ/1 and UNIQ/2 parameters to account for temperature dependency. We will not use the other parameters in the equation since they are not necessary to adequately represent the data of interest. It is important to not over-fit the data.
Ternary data is at constant temperature so we will only regress UNIQ/2 in this case. UNIQ/1 will remain at a fixed value.

To formulate a regression case we need to have experimental data and define the parameters to be regressed.
Data can be obtained from several sources, such as your own experiments, literature or databases such as NIST TDE. NIST TDE data can be accessed directly from Aspen Plus.
NIST TDE contains experimental data for pure component and binary systems only. We will search for binary LLE data for our systems in the NIST TDE database. For ternary data, we have access to them from a literature reference and will enter them manually. We are interested in a good parameter fit for a temperature range of operation between 0 C and 70 C.

We will guide you through the different steps required to set up the problem.
First we will define the components in the mixture and select the property method.
Next we will enable data regression.
Then we will show you how to retrieve and enter the experimental data and formulate the regression cases.
Finally we will show you how to analyze the results to validate the obtained parameters.
Click the Next button on the bottom right hand corner to walk through the solution or click on a step to skip to it.

We start by selecting the components and a property method, and then switch to regression mode.
Select the blank simulation template in Aspen Plus to start defining the case.
We define the first component in the mixture by entering water. In this case the ID is matching a component in the databank so Aspen Plus retrieves the component name and alias. The row information is now complete.
Next we enter ETOH for component ethanol. The ID is found and the row is now complete.
Finally we enter ethyl acetate, giving it an ID of ETOAC. This ID does not match anything in the databanks, so no information is retrieved. Place the mouse back in the field where you entered ETOAC and click the Find button. There are different criteria you can use to search for components. Enter ethyl acetate in “Name or Alias” using “Contains” as the search type. A list of databank components found is displayed in the table below. Click on ETHYL-ACETATE in the list to select and click Add selected compounds button. Click the Close button. The ethyl acetate component row is now complete.
The 3 components are now completely defined and mapped to the databank.
Select Method
Click the Next button or press the F4 key to move to the next required input sheet.
We are now going to specify the Property method we want to use in this case. There are several property methods using UNIQUAC. Certain options contain a hyphen and a suffix after the hyphen.
The convention is that the activity coefficient method name is followed by initials identifying the equation of state used for vapor phase interactions. Where there is no hyphen, the vapor phase is assumed ideal.
When the components in the vapor phase interact in a markedly non ideal way, we can select an equation of state to model those interactions. This is one such case, given the compounds’ polarity. We need to define Redlich-Kwong equation of state for vapor interactions. In the Method name we select UNIQ-RK.
When we click the Next button again we are taken to the binary interaction parameters sheet. Aspen Plus always displays the binary-parameters it retrieves from its databanks for the user to review. We want to fine tune the parameters that are already available in the databank. Depending on the pair, we have a range of variation starting at 293.15oK. We also notice that the ethyl acetate/ ethanol pair uses parameters valid for a higher range of temperature.

Click on the N.I.S.T. button to review binary data we have in N.I.S.T. that we can use in the temperature range of interest.
Select binary mixture, water, and ethyl acetate as the components of interest.
Click Retrieve Data.
Look for Binary LLE, click it and review the temperature ranges of the data sets.
Dataset Binary LLE 053 contains data for the temperature range we are interested in. Double click it to select it
Click the “Save data” button. Save both data and its uncertainty, as this could be useful during regression. Click OK to place the data in the data forms and again to go directly to the data form generated.
All binary data have been automatically saved to the form.
We will now have to create another data set for our ternary data. Under data click New and name it TRI-LLE.
In the Setup tab use Phase Equilibrium Data as the Category. Select TXX as data type for liquid-liquid equilibrium data, and select all three components by clicking the >> .

Next we check that the composition basis is Mole Percent.
Finally we enter the experimental data. These data have been supplied in the form of an Excel spreadsheet, available in this module’s Resources. We select the data and copy and paste it onto the data sheet.
Notice that the composition of ethyl acetate is grayed out and the standard deviation is zero. This is because the sum of compositions is constrained and has to add up to 1. Therefore, one of the components will have a calculated value.

Click the Run Mode Regression button in the Home ribbon to switch to regression mode. There are now incomplete regression related forms in the data browser, which we will be working on next.
We will be creating two regression cases, one for each data set.
Go to regression forms in Explorer and create a new regression case. The pre-selected UNIQ-RK property method is already defined in the Setup and we do want to perform data Regression to determine the binary parameters. Therefore, we will leave the Regression calculation option alone. The Evaluation calculation option would be useful to ascertain how well current parameters in use fit data entered in the Data sheets. This is not what we want to do.
In this case we do not have Henry components nor Chemistry defined so we leave those blank.
Once we pick the binary data set in the table and all other fields in the row are automatically defined. For LLE data, thermodynamic consistency does not apply.
Default choices are adequate so we move on to defining the parameters to regress.
We want to regress UNIQUAC binary parameters 1 and 2 for both ethyl acetate and water pairs, which we have binary data for.
Note that we can enter an initial value, lower and upper bounds and a scale factor to aid convergence, but that is not part of required input. We can also assume that the parameters are symmetric; however, that is not the case here.
In the report tab we can select certain additional properties to be reported. This will help visualize and evaluate regression results. We select TXX data to be reported.
The second regression case we set up uses ternary data. We go through the same set up as in the first case, but in this case we define all binary parameters UNIQ 2 for regression and fix UNIQ 1 for the water and ethyl acetate pair, rather than regressing it, as it is a constant in the equation.
We are now ready to run the regression cases.
Note that you can run both cases simultaneously or select an individual case to regress.
It is useful to have control when looking for the best fit, so we run the first regression case to fine tune the binary parameters and then run again for the ternary data.
Here we will run both cases simultaneously to save time.
When the run completes the control panel shows information on convergence. As there are already parameters in the binary parameter form, we are asked if we want to overwrite them. We can choose to do so and these new parameters will be in the right forms ready to use in simulations. The existing parameters are still available and we can easily switch between parameters from different sources.

Results are now available for both regression cases.
Looking at the binary parameter case, we can see the current values calculated for the UNIQ 1 and 2 parameters and their standard deviation.
There are other tables showing property residuals and estimated against experimental values for comparison. However, it is more interesting to visualize the same information in graphic form. We can use the Plot buttons in the ribbon to look at the residual for each component’s composition.
The residuals are the difference between the experimental value and the estimated (or calculate) value. They should be randomly distributed and within the experimental uncertainty.
Another useful graph shows the experimental versus estimated values (calculated from regressed parameters). In this case the points should fall very close to the plot diagonal for a good fit.
The predefined TXX plot shows the boundaries between the two co-existing liquid phases. We can see that the fit for binary LLE data is very good.
The regression case using ternary data has similar plots available but it also includes a triangular diagram to show the equilibrium tie-lines. The fitting looks very reasonable, except for the small deviation at the lower end of the organic phase.
Regressed parameters are automatically added to the property parameters input forms, so they are ready to use in further analysis and simulation work.