Process Optimization and Characterization of Cottonseed Biodiesel using Box Behnken Design and Artificial Neural Network Approaches

The production of biodiesel requires effective and predictive optimization strategies due to the growing demand for sustainable energy and the environmental constraints of fossil diesel. The present study was amid optimize biodiesel yield from cottonseed feed stock using BBD and ANN methods. The combined of statistical interpretability of BBD with the nonlinear predictive capability of ANN, were enhanced the reliability of optimization and prediction accuracy. The effects of methanol to oil ratio (3:1-9:1), KOH catalyst concentration (1-2 wt. %), and reaction time (60-80 min) on biodiesel yield were systematically investigated. BBD was used to design the experimental matrix and evaluate parameter interactions, while ANN with a 3-10-1 architecture was applied to predict biodiesel yield and assess model accuracy. The maximum cottonseed oil methyl ester yield was obtained at a methanol to oil ratio of 6:1, 1 wt. % of KOH, and 60 min reaction time. Biodiesel yields of 95.5% by experimental, 96.71% by BBD, and 95.48% ANN were attained, thereby validating the efficacy and reliability of the proposed dual optimization approach. Gas Chromatography-Mass Spectrometry (GC-MS) analysis confirmed methyl linoleate as the dominant ester (44.06%), and the produced biodiesel satisfied ASTM D651 and EN 14214 fuel standards. The model was confirmed P-value of 0.00 with Adj R 2 , R 2 , of 98.6 and 98.7% respectively. ANN surpassed BBD by achieving a higher coefficient of determination (R 2 = 99.3) alongside lower mean squared error (MSE = 72) and root mean squared error (RMSE = 26.8) in contrast to BBD (R 2 = 98.6, MSE = 81, RMSE = 28.5). The findings demonstrate that the combined BBD-ANN framework is a robust and viable tool for biodiesel process optimization, offering high efficiency, reduced processing time, and strong potential for environmentally sustainable fuel production.