Accurate prediction of the standard net heat of combustion from molecular structure

作者：Albahri Tareq A^*

来源：Journal of Loss Prevention in the Process Industries, 2014, 32: 377-386.

DOI：10.1016/j.jlp.2014.10.005

• Heat of combustion
• Group contribution
• Molecular modeling
• Neural networks
• QSPR
• Quantitative structure property relation

A Quantitative Structure Property Relation (QSPR) is developed to predict the standard net heat of combustion (Delta H-c(o)) of chemical compounds based only in their molecular structures. A Structural Group Contribution (SGC) method is used to determine Delta H-c(o) through two models: a Multi-Variable Regression (MVR) based on least squares and an Artificial Neural Network (ANN). The SGC method was used to probe the structural groups that have significant contribution to the overall Delta H-c(o) and concluded that 47 atom-type structural groups can represent the Delta H-c(o) for 586 pure substances. The input parameters of the SGC method are the number of occurrence of each of the 47 structural groups in each molecule. The ANN was the more accurate of the two models; it can predict Delta H-c(o) with an overall correlation coefficient of 0.999 and an average relative error of 0.89%. The MVR model is less accurate but is also simple and practical and provides reliable estimates. The results of both models are compared to others in the literature. The SGC method presented is very useful and convenient to assess the hazardous risks of chemicals.

• 出版日期2014-11