PCP-SAFT Parameters of Pure Substances Using Large Experimental Databases

This work reports pure component parameters for the PCP-SAFT
equation
of state for 1842 substances using a total of approximately 551 172
experimental data points for vapor pressure and liquid density. We
utilize data from commercial and public databases in combination with
an automated workflow to assign chemical identifiers to all substances,
remove duplicate data sets, and filter unsuited data. The use of raw
experimental data, as opposed to pseudoexperimental data from empirical
correlations, requires means to identify and remove outliers, especially
for vapor pressure data. We apply robust regression using a Huber
loss function. For identifying and removing outliers, the empirical
Wagner equation for vapor pressure is adjusted to experimental data,
because the Wagner equation is mathematically rather flexible and
is thus not subject to a systematic model bias. For adjusting model
parameters of the PCP-SAFT model, nonpolar, dipolar and associating
substances are distinguished. The resulting substance-specific parameters
of the PCP-SAFT equation of state yield in a mean absolute relative
deviation of the of 2.73% for vapor pressure and 0.52% for liquid
densities (2.56% and 0.47% for nonpolar substances, 2.67% and 0.61%
for dipolar substances, and 3.24% and 0.54% for associating substances)
when evaluated against outlier-removed data. All parameters are provided
as JSON and CSV files.