چكيده به لاتين
. In this study, the safety properties such as autoignition temperature, flash point, lower and upper flammability temperature, lower flammability limit(V%), and upper flammability limit(V%) of any alkyl ester available in the DIPPR database were gathered and studied to develop models for predicting these properties for fatty acid alkyl esters; the quantitative structure-property relationship was the method which was applied for creating these models. In this method, the main objective is to create a reliable mathematical relationship between the desired property (here, safety properties) and the structure of components gathered to develop the model (here, alkyl esters). The number of data points gathered for each safety property is 126, 179, 179,179,178,179 for AIT, flashpoint, LFLT, UFLT, LFL(V%), and UFL(V%), respectively. A total number of 4 models (3 nonlinear models created with machine learning methods such as genetic programming, support vector machine, and random forest and one linear model created with multiple linear regression) were developed for each property. The statistic parameter R2 calculated from examining the best models’ reliability for the whole dataset are reported as below: 0.735 for AIT, 0.977 for flashpoint, 0.979 for lower flammability limit temperature, 0.979 for upper flammability temperature, 0.988 for lower flammability limit(V%) and 0.922 for upper flammability limit(V%).
