PREDICTION OF THERMOPHYSICAL PROPERTIES BY METHODS BASED ON SIMILARITY OF MOLECULAR STRUCTURES

摘要

Prediction of thermophysical properties required for heat transfer calculation and the design and development of thermal systems is considered. Newly developed computational methods for property prediction are described and their use is demonstrated for the prediction of various constant properties, such as normal boiling and melting temperature, critical properties, heats of formation, etc., as well as for the temperature dependent properties: vapour pressure and viscosity of liquids. The computational methods discussed include the Quantitative Structure-Structure-Property Relationship (QS2PR), the short-cut QS2PR method (SC-QS2PR) and the targeted QSPR method (TQSPR). These methods are based on the use of molecular descriptors (calculated from the molecular structure) for predicting properties. However, unlike in the traditional property prediction methods, these new methods are targeted to a particular compound, or a group of compounds, and rely on the identification of a relatively small number of structurally similar compounds. Hence, they can provide accurate predictions and estimates of the prediction error. In the examples presented it is demonstrated that proper combination of the proposed method can provide property prediction within the experimental error level.