Yellow Oleander Seed Oil Extraction Modeling and Process Parameters Optimization: Performance Evaluation of Artificial Neural Network and Response Surface Methodology

Abstract

The effects of sample weight, time and solvent type on YOSO yield were evaluated using ANN and RSM. The predicted optimal condition for the extraction process was found to be the same for the ANN and RSM models developed: sample weight of 20 g, time of 3 h and petroleum ether. The models predictions of YOSO yield (ANN [77.42%] and RSM [78.64%]) at optimum levels were verified experimentally (ANN [77.63%] and RSM [76.64%]). Evaluation of the models by R2 and AAD showed that the ANN model was better (R2 = 1.00, AAD = 0.61%) than the RSM model (R2 = 0.98, AAD = 3.19%) in predicting YOSO yield. Physicochemical properties of the YOSO indicated that it was nonedible and the fatty acids profile showed that the oil was highly unsaturated (76.13%). Practical Applications This study demonstrated modeling of YOSO extraction and optimization of process parameters that are involved. The performance evaluation results showed that both the ANN and RSM could be used for modeling and optimization of YOSO extraction process. Also, the characterization of the oil showed that it could serve as raw material for many chemical industries such as biodiesel production, soap, cosmetic and pharmaceutical industrials. The results from this study can be successfully scaled up to pilot scale. Also, the results could be extended to the extraction of other oilseeds.