Property prediction and consistency analysis by a reference series method

Abstract

in Wiley Online Library (wileyonlinelibrary.com).Data analysis and prediction of pure component properties of long-chain substances is considered. The emphasis is on homologous series and properties for which insufficient data are available. A two-stage procedure is recommended, whereby a linear (or nonlinear) quantitative structure-property relationship (QSPR) is fitted to a ''reference'' series, for which an adequate amount of precise data is available. This QSPR should represent correctly both the available data and the asymptotic behavior of the property. In the second stage a quantitative property-property relationship (QPPR) is derived to represent the predicted property values of a ''target'' series in terms of the property values of the reference series. The procedure is applied for properties which are highly correlated with the number methylene groups in homologous series: ÁH 0 f and DS 0 f . It is shown that the method is very useful for consistency analysis of property data and enables a reliable prediction of DH 0 f and DS 0 f , and, thus, also of DG 0 f for long-chain substances.Homologous series have the general formula H(CH 2 ) n R, where ACH 2 A is the methylene group, n is the number of methylene groups, and R is an end (or functional) group, Correspondence concerning this article should be addressed to M. Shacham at shacham@bgu.ac.il.

Prediction of normal melting point of pure substances by a reference series method

“…The target series are related to the reference series by means of a quantitative propertyproperty relationship (QPPR). This method was recently used by Shacham et al 17 to successfully predict the enthalpy, entropy, and Gibbs energy of formation of ideal gases.…”

“…This similarity can be used for improving the prediction for series for which an insufficient amount of and/or low-precision data are available. For such situations, relying on the suggestions of Peterson 15,16 , Shacham et al 17 developed the "Reference Series" method. Using this method, experimental (if available) or predicted data of a "reference" homologous series for which the largest amount and highest precision experimental data are available, are used as the basis for prediction of properties for other (target) series.…”

“…Shacham et al 17 have recently shown that properties of compounds in two homologous series can be represented (at least locally) by a linear QPPR y t 5b 0 1b 1 y r n C n C;min (1) where y r is the property value of a compound in the reference series, y t is the property of a compound (related to the reference compounds in terms of the number of carbon atoms n C ) in the target series, and b 0 and b 1 are parameters.…”

Challenges in the Modeling of Thermodynamic Properties and Phase Equilibrium Calculations for Biofuels Process Design

“…For homologous series the TQSPR algorithm searches for a descriptor vector collinear with the studied property and develops a linear correlation, which predicts within experimental precision from only several measured data. Definition and identification of the importance of dominant descriptors led to refinement of the TQSPR method (Kahrs et al, 2008) and the development of the references series method (Shacham et al, 2013a) and the TQSPR1 method (Shacham et al, 2013b).…”