Application of Musa paradisiaca catalyst on synthesis of biodiesel and optimization of process parameters using random forest machine learning model

This study explores the sustainable synthesis of biodiesel using a heterogeneous catalyst derived from Musa paradisiaca (banana peel) biomass and a blended feedstock of Quality Neem Oil (QNO) and Spent Frying Oil (SFO). Agricultural waste biomass was valorized into a low-cost, eco-friendly catalyst, contributing to waste reduction and green fuel production. The prepared catalyst was comprehensively characterized using X-Ray Diffraction (XRD), Energy-Dispersive X-ray spectroscopy (EDX), Scanning Electron Microscopy (SEM), and Fourier Transform Infrared Spectroscopy (FTIR), confirming its crystalline nature, elemental composition, surface morphology, and functional groups. The blending of SFO, rich in saturated fatty acids, with QNO containing higher unsaturated fatty acid content effectively improved biodiesel properties, particularly viscosity and cold flow behaviour. To optimize the transesterification process, a Random Forest (RF) machine learning model was employed due to its ability to capture complex nonlinear relationships and operate efficiently in high-dimensional feature spaces with strong predictive accuracy. The RF model achieved a maximum biodiesel yield of 94.37% under optimal conditions of 56.5 °C reaction temperature, 7.5:1 methanol-to-oil molar ratio, and 4.4 wt% catalyst loading. Catalyst characterization revealed enhanced surface area and porosity, which significantly contributed to improved catalytic activity and reaction efficiency. Overall, this research demonstrates the effective integration of a biomass-derived catalyst and Random Forest-based optimization as a promising approach for cost-effective, high-yield, and environmentally sustainable biodiesel production.