ABSTRACT
This study investigated the synthesis of biodiesel from the transesterification of an oil blend made from an optimum combination of Waste Vegetable Oil (WVO), waste crude palm oil (WPO) and waste animal fat (WAF) using a composite heterogeneous catalyst synthesised from waste marble tiles and plantain peduncle. The catalyst and the precursors were characterized with the aid of a scanning electron microscope (SEM) and various analytical techniques, including energy dispersive X-ray (EDX), X-ray diffraction (XRD), Fourier transform infrared (FTIR) spectroscopy, Brunauer, Emmett and Teller (BET), and X-ray fluorescence (XRF) analysis. The individual oil feedstock and the blend as well as the biodiesel produced were also characterized to determine their physicochemical characteristics. The blending of the oil was carried out using a mixture design while the transesterification experiments for biodiesel production were carried out using a BoxBehnken design (BBD). The biodiesel yield was assessed by investigating the impact of its processing parameters which are the reaction time, reaction temperature, methanol-oil ratio, and catalyst loading. Two independent expert systems namely Response Surface Methodology (RSM) and Artificial Neural Network (ANN) were employed in the modelling of the biodiesel yield prediction. The results showed that catalyst and oil blend characterization showed high suitability for use in the transesterification experiment. Furthermore, the result of experimental biodiesel yield showed that the reaction temperature, methanol-oil ratio and catalyst loading significantly affected the yield of biodiesel while a much lesser impact was observed for biodiesel yield concerning reaction time. The response prediction results from RSM and ANN were assessed based on their performances in terms of xv statistical parameters such as R 2 , predicted R2 , adjusted R2 , and RMSE. The ANN model performed better than the RSM model in terms of the statistical parameters considered. This indicates the robustness of the ANN modelling over RSM in predictions and solutions of complex iterative problems such as those encountered in biodiesel production by transesterification. The optimization results showed that for the RSM optimization, the optimal value of the biodiesel yield was 77.60%, observed at a reaction time of 117.046 minutes, reaction temperature of 66.86oC, catalyst loading of 1.24 wt%, and methanol-oil ratio of 11.76 while for the ANN-GA optimization, the optimum value of the biodiesel yield was 88.28%, obtained at ideal reaction conditions of 66.36oC temperature, 99.92 minutes reaction time, 1.267 wt% catalyst loading, and 11.379 methanol-oil ratio. This indicates a higher optimization prediction by ANN-GA over RSM. The biodiesel produced after characterization was found to constitute principally saturated fatty acids and methyl esters which is indicative of high performance and lower emission loads and ideal for compression ignition engines.
