Adaptive neuro-fuzzy inference system-genetic algorithm versus response surface methodology-desirability function algorithm modelling and optimization of biodiesel synthesis from waste chicken fat

Abstract

BackgroundLarge and wide availability of waste chicken fat (WCF) and its underutilization make its application as cheap and sustainable biodiesel feedstock, a viable approach to slaughter-house waste control and environmental protection. MethodsThree-step process:wet-rendering, esterification and transesterification were employed for the biodiesel production. Adaptive neuro-fuzzy inference system integrated with genetic algorithm (ANFIS-GA) and response surface methodology(RSM)-desirability function were applied for modelling and optimization of the transesterification of WCF. Impacts of input process conditions: reaction time, reaction temperature, methanol: WCF molar ratio, and catalyst concentration on the response were studied. Biodiesel fuel quality was evaluated using FTIR, GC-MS, AOAC, ASTM D standard methods. Significant FindingsThe RSM and ANFISmodels described the process accurately through high R2 of 0.9361 and 0.9713 and low RMSE of 1.1399 and 1.059 respectively. RSM and ANFIS-GA optimum conditions were 92.70% and 94.89% yield at reaction time of 51.91 and 50.21 min, temperature of 51.74 and 50.20ºC, methanol: WCF molar ratio of 8.37:1 and 6.91:1 and catalyst concentration of 0.85 and 0.78wt% respectively. ANFIS-GAgavebetter predictive capability and economic conditions than the numerical tool of RSM. Fuel quality of the biodiesel satisfied international specifications.