Fatty Acid Methyl Ester (FAME) composition used for estimation of biodiesel cetane number employing random forest and artificial neural networks: A new approach

Abstract

Cetane Number (CN) is the property used to evaluate the quality of biodiesels. The CN is mainly affected by the Fatty Acids Methyl Ester (FAME) composition of the biodiesel. The common experimental methods of determination of CN is expensive and time consuming and are not always accurate, so it is vital to use other methods to predict CN. In this work, Random Forest (RF) and Artificial Neural Networks (ANN) assisted by 10-fold cross validation were employed to present appropriate, reliable and more generalized models for the prediction of CN based on experimental data of 131 different FAMEs collected from literature. Two different regression models obtained based on these methods. The Root Mean Squared Error (RMSE) and the coefficient of determination (R2) of 0.95, 2.53 for ANN model, and 0.92, 3.09 for RF model showed the high accuracy of these models. In term of accuracy, ANN model showed better results compared to RF model. On the other hand, in term of transparency and ease of interpretation, the RF model could be widely applied in CN determination. The positive effect of FAMEs on CN was obtained if Stearic acid or Myristic acid was higher than 51.95% or 44.95% regardless of other FAME acid percentages. In addition, values greater than 68.4% in Linolenic acid could lead to a negative effect of that acid on CN.