Multi-objective optimization of mechanical oil extraction from Camellia oleifera seeds using Kriging regression and NSGA-II

摘要

The objective of this study is to optimize the mechanical extraction process of crude camellia oil regarding yield and quality. The effects of process parameters namely applied pressure (10-45 MPa), heating temperature (40-100 degrees C), heating time (10-40 min), and moisture content (0-9%) on oil yield, free fatty acid (FFA), and peroxide value (PV) were investigated by modeling the oil extraction process using Kriging regression (KR) models; optimum process parameters were then determined via non-dominated sorting genetic algorithm II (NSGA-II). The results showed that KR model is effective to predict the yield (R-2= 0.9595 and A-RMSE = 1.94), FFA (R-2= 0.9983 and A-RMSE = 0.0517), and PV (R-2= 0.8733 and A-RMSE = 0.0358). The relationship between process parameters and each objective is non-linear. In addition, the maximum yield (42.20%) and best quality parameters (minimum FFA = 0.2917 KOH mg/g and minimum PV = 0.1316 g/100 g) are contradictory, which suggested applying Pareto-optimal front to balance the three objectives to a satisfied degree. Practical Applications Heating temperature, applied pressure, heating time and moisture content are primary operating conditions affecting the yield and quality of mechanically extracted fats and oil. Selecting the suitable values of these parameters can therefore improve the oil yield and oil quality. This could be realized by process optimization based on experimental and mathematical modeling method. The KR model and NSGA-II were, therefore, adopted to optimize the process parameters to obtain maximum yield, minimum FFA and PV. In addition, parametric analysis was conducted to give insight into the effects of process parameters on the yield and quality, which is beneficial to understand the mechanism behind the mechanical extraction of camellia oil. The Pareto-optimal front obtained from NSGA-II provides engineers with several decision parameters to design the equipment and to obtain better end products more efficiently.