Optimized biodiesel production from C. innophyllum bio-oil using Kriging and Ann predicitive models

Abstract

This work aimed at optimizing the two-stage transesterification efficiency of the production of Calophyllum inophyllum biodiesel using Artificial Neural Network and Kriging predictive models. Response Surface Methodology was used to develop the central rotatable composite design of twenty-seven trial experimental runs with variations in the input process parameters like methanol to oil molar ratio, potassium hydroxide catalyst loading, and reaction time. A multi-layered non-linear regressive Artificial Neural Network model with feed-forward propagation and a numerical surrogate Kriging model was used to predict the Calophyllum inophyllum biodiesel yield. The efficacy of the developed model was verified using Analysis of Variance by comparing its coefficient of determination and the mean relative percentage deviation values. The optimized Calophyllum inophyllum biodiesel as 98.1% is derived with 0.94 v/v of methanol to oil molar ratio, 0.98 wt % of potassium hydroxide catalyst loading, and 80 minutes reaction time with 70oC constant reaction temperature as predicted by Kriging model. The optimized parameters were also verified experimentally.