Information Technology Reference
In-Depth Information
Figure 6.5
Activity coefficient plot.
This is illustrated from Aspen Plus regression runs, which give the following results.
When
b
ij
=
0:
a
12
=
1
.
435
a
21
=
2
.
298
When
a
ij
=
0:
b
12
=
498
.
8
b
21
=
791
.
1
Virtually identical results were obtained with least squares and maximum likelihood
regressions.
When setting up multiple regressions in one run, care must be taken to set the
unused parameters equal to zero because Aspen Plus will use older or stored val-
ues in the regression. The ideas illustrated above are appropriate for any choice of
activity coefficient equation. Details of the regression runs above are available at
Examples/chipavanLaar and Examples/chipwilson.
6.5.2 Regression of LLE Data
When fitting liquid - liquid equilibrium data the situation is somewhat different. Most
extraction processes are isothermal; therefore, if binary mutual solubility data are avail-
able, consistent with ternary (or higher) LLE data at the same temperature, the
a
i,j
can
be obtained by using one data point for regression. Since each data point can generate
two equations,
(
γ
1
x
1
)
I
=
(
γ
1
x
1
)
II
(6.30)
an exact solution is possible.
