Modeling and Optimization of Subcritical CO₂ Extraction of Safflower Seed Oil Using Response Surface Methodology and Artificial Neural Networks

This article aimed to find effective modeling methods to predict the extraction rate of safflower seed oil by subcritical CO₂ extraction, and optimize its extraction process conditions. Based on single-factor experiments, Box-Behnken experimental design was adopted to study the effects of extraction pressure, separation temperature, and extraction time on the extraction rate of safflower seed oil. Response surface methodology (RSM) and artificial neural network (ANN) were used to model and analyze the same experiment. The process conditions were optimized using RSM numerical optimization, ANN, and the combination of artificial neural network and genetic algorithm (ANN-GA). The results showed that both RSM and ANN models could accurately predict the extraction rate. However, by comparing the determination coefficient (R²), mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) values of the two models, it was concluded that the ANN model (R²=0.9966) had a better predictive effect than the RSM model (R²=0.9950). The optimal extraction conditions and extraction rate determined by ANN-GA were as follows: Extraction pressure of 19.04 MPa, separation temperature of 55.50 °C, extraction time of 134.98 min, and extraction rate of 23.52%. The study showed that both RSM and ANN methods could be used for modeling and optimization of subcritical CO₂ extraction of safflower seed oil, but ANN had better prediction accuracy and fitting ability.