Crude rubber seed oil esterification using a solid catalyst: Optimization by hybrid adaptive neuro-fuzzy inference system and response surface methodology

Abstract

Esterifying high free fatty acid (FFA) oil with acid is necessary to avoid soap formation during biodiesel production. Thus, this study evaluated the efficacies of response surface methodology (RSM) and adaptive neuro-fuzzy inference system (ANFIS) in modeling the esterification process for crude rubber seed oil (CRSO) with a high FFA catalyzed by dehydrated Fe2(SO4)3. A central composite design (CCD) with three factors and five levels was applied to examine the influence of methanol:CRSO molar ratio (25:1–75:1), Fe2(SO4)3 loading (8–16 wt %), and time (3–4 h) on reduction of the high FFA (22.2%) of CRSO. The performance of the particle swarm optimization (PSO), genetic algorithm (GA), and RSM were assessed in optimizing the process variables. Statistics for the ANFIS and RSM models showed that both could reliably describe the esterification process with low mean relative percent deviation (MRPD) of 1.77 and 4.98; and high coefficients of determination (R2) of 0.9838 and 0.9730, respectively. The optimization results are in this order: ANFIS-PSO, ANFIS-GA, RSM-PSO, RSM, and RSM-GA. The ANFIS-PSO hybrid predicted the best optimal condition as Fe2(SO4)3 loading of 16.97 wt%, methanol:CRSO molar ratio of 44.21:1, and time of 3.39 h with the lowest FFA of 0.56%.