Prediction and Optimization of Fish Biodiesel Characteristics Using Permittivity Properties

The purpose of this research was to predict and optimize the fish biodiesel characteristics using its permittivity properties. The parameters of biodiesel permittivity properties such as έ, dielectric constant, and ε″, loss factor at microwave frequencies of 434, 915, and 2,450 MHz, were used as input variables. The fish biodiesel characteristics, as Fatty Acid Methyl Ester (FAME) content and flash point at three different levels of reaction time 3, 9, and 27 min and catalyst concentrations 1, 1.5, and 2% w woil-1, were selected as output parameters for the models. Linear Regression (LR), the Multi-Layer Perceptron (MLP), and the Radial Basis Function (RBF) as the methods of Artificial Neural Networks (ANN), and the response surface methodology were compared for prediction and optimization of FAME content and flash point. A comparison of the results showed that the RBF recorded higher coefficient of determination at frequency of 2,450 MHz as 0.999 and 0.988 and lower root mean square error as 0.009 and 0.023 for FAME content and flash point, respectively. The optimum condition was obtained using RSM by FAME content of 89.88% and flash point of 152.7°C with desirability of 0.998.