Comparative Analysis of Conventional Optimization Techniques with RSM and ANN Models for Extracting Oil from Sterculia urens Seeds for Biodiesel Applications

Abstract In this work, oil was extracted from Sterculia urens (S. urens) seeds in a batch reactor and the parameters affecting this process were optimized. For this study, a suitable solvent for extraction was identified and its solvent to seed meal ratio was determined as petroleum ether solvent and 8:1, respectively. The total oil content of the seed by Soxhlet extraction is 38.9 wt%. Other parameters which affect the oil yield, such as the meal size of seed, extraction temperature and extraction time, were optimized and their values were 0.25 mm, 60°C and 240 min, respectively, to get 38.2 wt% oil. Moreover, Response Surface Methodology (RSM) and Artificial Neural Network (ANN) techniques were compared with conventional extraction and used for predicting the optimum values. The correlation regression coefficient (R2) values for RSM and ANN were 0.822 and 0.99, respectively. The predicted optimum values obtained in both the tools are approximately similar for oil extraction as 0.5 mm meal size, 60°C temperature and 180 min to get optimum yield. Finally, the physiochemical parameters of the oil were determined by standard methods and predicted the properties of biodiesel made from this. Thus these results suggest that the oil obtained can be used as a potential second-generation non-edible feedstock for biodiesel production.