Computation of upper flash point of chemical compounds using a chemical structure-based model

Farhad Gharagheizi (1) , Poorandokht Ilani-Kashkouli (2) , Seyyed Alireza Mirkhani (1) , Amir H. Mohammadi (3)

1 Department of Chemical Engineering
2 Department of chemical engineering
3 CEP/Fontainebleau

In this communication, a quantitative structure-property relationship (QSPR) is presented for an estimation of the upper flash point of pure compounds. The model is a multilinear equation that has eight parameters. All the parameters are solely computed based on chemical structure. To develop this model, more than 3000 parameters were evaluated using the Genetic Algorithm Multivariate Linear Regression (GA-MLR) method to select the most statistically effective ones. The maximum average absolute relative deviation (mARD), ARD, squared correlation coefficient, and root mean squares of error of the model from database (DIPPR 801) values for 1294 pure compounds are 25.76%, 3.56%, 0.95, and 17.42 K, respectively.

Mots clés: Average absolute relative deviations; Error of the models; Flash points; Multivariate linear regressions; Pure compounds; Quantitative structure property relationships; Root Mean Square; Squared correlation coefficients; Structure-based

Domaines: Génie des procédés

Farhad Gharagheizi, Poorandokht Ilani-Kashkouli, Seyyed Alireza Mirkhani, Amir H. Mohammadi. Computation of upper flash point of chemical compounds using a chemical structure-based model. Industrial and engineering chemistry research, 2012, 51 (13), pp.5103-5107. ⟨10.1021/ie202868v⟩. ⟨hal-00796847⟩