Box–Behnken optimization and <scp>ANN</scp> ‐ <scp>ANFIS</scp> integrated modelling for transesterification using thermally activated basic oxygen furnace slag catalyst

Abstract

Abstract Global energy demand and environmental concerns have intensified the search for renewable and sustainable energy sources. This study thus, focuses on optimizing the transesterification process of waste cooking oil (WCO) using thermally activated basic oxygen furnace slag catalyst calcined at 850°C (BOF 850). The optimization and modelling were conducted using Box–Behnken design (BBD), artificial neural networks (ANN), and adaptive neuro‐fuzzy inference system (ANFIS). Among these, ANN produced the most accurate model with the least root mean square error (RMSE) of 0.0985 and a coefficient of determination ( R 2 ) value of 0.9710, outperforming ANFIS (RMSE: 0.1780, R 2 : 0.9604) and BBD (RMSE: 0.4973, R 2 : 0.98230). While ANFIS has historically demonstrated strong predictive power in transesterification reactions, its performance in this study was limited by the small dataset (17 data points), highlighting its sensitivity to data size. The optimal conditions determined via BBD were a methanol‐to‐oil ratio of 23.5:1, a catalyst amount of 20.4%, and a reaction time of 10.3 h. Experimental validation at these conditions resulted in a biodiesel yield of 91.73%. Analysis of variance (ANOVA) showed that catalyst amount had the highest statistical significance ( F ‐value: 58.08, p ‐value: 0.0001), while the methanol‐to‐oil ratio and reaction time exhibited the strongest interaction effects ( F ‐value: 42.84, p ‐value: 0.0003). Gas chromatography–mass spectrometry (GC–MS) analysis of the produced biodiesel revealed a composition of 52.7% saturated and 46.1% unsaturated fatty acids, influencing the fuel properties. The study thus shows that ANN can effectively model transesterification of WCO using BOF 850 catalyst.