Discussion Of Estimation Methods For Pure Component Constants

This chapter has described a variety of methods for predicting critical properties, acentric factor and normal boiling and freezing temperatures. Unlike previous edi tions where most of the methods were of the group /bond/atom type and were limited in the classes of properties or types of substances treated, recent work in these techniques has both improved their accuracy and broadened their range. Thus, there are now two methods (Joback and Constantinou/Gani) that provide all pure component constants and heat capacities and properties of formation of ideal gases (see Chap. 3) with a single group formulation.

All of the group /bond/atom methods examined here were set up on spreadsheets since their application was the same regardless of the property and component. Some methods required larger data bases than others, but implementation and execution for new substances and properties would be straightforward. It is also possible to obtain a complete suite of estimation methods in the program Cranium (1998). The Constantinou/Gani method for the broadest set of properties is available directly from CAPEC (1999). It is expected that methods of currently limited application, such as the Marrero/Pardillo approach, will be expanded to include other properties. There are still enough errors and limitations in the methods that new research will continue with this approach. It is likely that an individual user with a typical individual computer will be able to use both current and future versions of these methods.

This edition has also introduced the molecular descriptor and QSPR relations which add another dimension to the methodology since they can be applied not only to pure component constants but to a variety of solution systems (Mitchell and Jurs, 1998; Katritzsky, et al., 1998). This presents users with opportunities to obtain more reliable values, but also may require greater expertise and investment in the selection of computer software for estimations. As mentioned above, there is no tabulation of contributions for these methods since the molecular structure and descriptors of each new substance are computed from molecular and quantum mechanical programs. While complex, the estimation methods are established by a generally agreed upon process of fitting limited data (Mitchell and Jurs, 1996; St. Cholakov, et al., 1999) to establish the weights of the significant descriptors from a large set of possibilities. As described above, the results can be very good and it is likely that further improvements in computational techniques will add even greater reliability and applicability. However, the computational power required is extensive and care must be exercised to use the same computational programs as the developer in order to insure that the values for the descriptors will be consistent with those fitted. This is likely to require expertise and computers of a large organization and beyond that of an individual. At this time, these methods have not been implemented in process simulation software, but that would certainly be possible in the future.

It will be important that users follow the developments in this area so that the most prudent decisions about investment and commitment can be made.

Was this article helpful?

0 0

Post a comment