Optimization of biodiesel production from cottonseed oil using response surface methodology and artificial neural network techniques

Abstract

The depletion of fossil fuels and increasing environmental concerns necessitate the development of sustainable energy alternatives. Biodiesel, derived from renewable feedstocks, is a promising solution to address these challenges. This study focuses on optimizing biodiesel production from cottonseed oil using a hybrid modelling approach that integrates response surface methodology (RSM) and artificial neural networks (ANN). Unlike conventional studies that rely on a single optimization technique, this study combines RSM with Box-Behnken design and ANN to enhance predictive accuracy and process efficiency. Biodiesel was synthesized through transesterification using methanol and KOH catalyst, with optimization based on reaction time (40, 60, and 80 min.), concentration of catalyst (0.5, 1, and 1.5 wt. %), and methanol to oil ratio (1:4, 1:6, and 1:8). The Box-Behnken design of RSM generated an experimental design matrix, while ANN featured a 3-10-1 architecture to evaluate process variables. The highest biodiesel yield of 94.66 % was achieved at catalyst concentration (1 wt. %), a reaction duration (60 min.), and a methanol to oil ratio (1:6). The RSM quadratic model achieved an R² of 0.970 and an adj-R² of 0.968. The ANN model, trained using the Levenberg-Marquardt approach, achieved a mean squared error of 4.963e-18 and an R-value of 0.9957 at epoch 3. Gas chromatography-mass spectroscopy (GC–MS) confirmed several fatty acid concentrations in the methyl ester. Furthermore, the biodiesel's key physicochemical properties meet EN 14214 and ASTM D6751 standards. This study contributes to advancing renewable energy sources by utilizing cottonseed oil, thereby promoting environmental sustainability.