Improved structure–property relationship models for prediction of critical properties

New chemical process design strategies utilizing computer-aided molecular design (CAMD) can provide significant improvements in process economics by identifying the chemicals with optimum function attributes. Robust CAMD algorithms rely on thermo-physical property models for process simulation; hence, reliable property models are required to realize the full potential of these algorithms. Further, models which can predict thermo-physical behavior of diverse molecular species based solely on chemical structure information are particularly valuable in many applications.In this study, we present new structure–property relationships (SPR) models for the prediction of critical properties (critical temperature, pressure and volume) of a diverse organic dataset containing over 1230 molecules involving 73 classes of hydrocarbons. A number of approaches, including linear, non-linear and genetic algorithms, have been employed for model development. In addition, the models benefited from (a) the inclusion of descriptors from three different commercial QSPR software packages, (b) literature descriptors identified to be significant and (c) new descriptor combinations we have developed to account for the non-linear behavior exhibited by structural descriptors. The resultant QSPR models are capable of predicting critical properties of the diverse set of molecules considered with an average absolute percent deviation (%AAD) of 0.9, 1.5 and 1.7 for critical temperature, critical pressure and critical volume, respectively.