Machine learning-based predictive modelling of biodiesel production from animal fats catalysed by a blast furnace slag geopolymer

Abstract

• Biodiesel production was modelled using ANN and ANFIS to optimise the process. • Key parameters influencing biodiesel yield were methanol-to-oil ratio, catalyst ratio, time, and temperature. • ANFIS demonstrated superior predictive performance compared to ANN, with a higher R² of 0.9857 and a lower MSE of 2.9386. • The study emphasises the potential of machine learning techniques for optimising biodiesel production. • Future research should explore the application of other machine learning models and their optimisation capabilities. Climate change and fossil fuel depletion have driven the need for environmentally friendly, low-carbon-emitting fuels. Biodiesel has emerged as a cleaner and sustainable renewable energy source. Yet its production is confronted by challenges such as high feedstock costs and wastewater generation from conventional catalysts. To address these issues, alternative feedstocks like animal fats and eco-friendly catalysts such as geopolymers have gained attention. However, a knowledge gap exists in machine learning (ML) applications for biodiesel production from animal fats catalysed by blast furnace slag (BFS) geopolymer. This study bridges this gap by employing predictive modelling methods: artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) to predict biodiesel yield. Four input parameters were studied: methanol-to-oil ratio (6–18 mol/mol), catalyst ratio (5–15 wt. %), reaction time (3–7 h) and temperature (30–70 °C), with yield as output variable. The highest yield, 97.06 %, was achieved. MATLAB 2024b's NN toolbox and neuro-fuzzy designer toolbox were used to implement the models. Evaluation metrics were used to assess the models' effectiveness; both ANN and ANFIS demonstrated reliable performances, with ANFIS outperforming ANN. ANFIS achieved an R 2 of 0.9857 and a lower mean squared error (MSE) of 2.9386, compared to ANN's R 2 of 0.9781 and MSE of 6.2106. This study highlights the potential use of ML in biodiesel production from animal fats catalysed by BFS geopolymer, offering a significant contribution to the field while reducing environmental impacts. Future research should explore biodiesel's physicochemical properties and BFS geopolymer catalysts' reusability.

Topics & Concepts