Optimization and predictive modelling of biodiesel production from waste cooking oil catalyzed by blast furnace slag geopolymer using RSM and machine learning

Abstract

Biodiesel production from waste cooking oil (WCO) has emerged owing to growing interest in sustainable energy sources. Geopolymers synthesized from industrial wastes, such as blast furnace slag (BFS), are promising catalysts because of their environmental benefits and catalytic properties. However, a knowledge gap exists in the application of machine learning (ML) for the transesterification of WCO catalyzed by geopolymer. This study aimed to optimize and predict biodiesel yield using a numerical approach, response surface methodology (RSM), and two ML algorithms: artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). Four input process parameters were investigated: methanol-to-oil ratio (20–50 wt.%), catalyst ratio (5–15 wt.%), reaction time (4–8 h), and reaction temperature (30–70°C), with biodiesel yield as the response. Central composite design (CCD) was used to evaluate the effects of process parameters, and models were evaluated using R2, root mean squared error (RMSE), mean absolute error (MAE), mean average percent error (MAPE), and average relative error (ARE). The optimum yield of 98.635% was achieved at 11.103 wt.% catalyst, 44.068 wt.% methanol to oil, 6.704 h reaction time and 57.493°C reaction temperature. ANFIS displayed the best predictive performance (R2: 0.996, RMSE: 1.429, MAE: 0.684, and MAPE: 1.548). Analysis of variance (ANOVA) results indicated the methanol-to-oil ratio had the most significant impact (F-value: 91.77), followed by the catalyst ratio (F-value: 51.58). The produced biodiesel met ASTM D6571 and EN 14214 standards. Future research should focus on catalyst reusability and catalyst synthesis optimization for industrial applications. This study contributes to global efforts toward sustainable biodiesel production by addressing waste disposal and green fuel development.