Response Surface Methodology and Artificial Neural Network Optimization of Subcritical CO2 Isobaric Extraction Process and Kinetics of Grape Seed Oil
This article studies the modeling and optimization of subcritical CO2 isobaric extraction of grape seed oil using response surface methodology(RSM)and artificial neural network(ANN),with the extraction rate of grape seed oil as the indicator.The main process parameters(extraction pressure,separation temperature,extraction time)are usd to investigate the impact on the extraction rate of grape seed oil,And the optimal process conditions were optimized and validated through RSM and artificial neural network coupled genetic algorithm(ANN-GA).At the same time,during the extraction process,the optimal kinetic model for grape seed oil extraction was fitted and validated by analyzing the changes in grape seed oil extraction rate under different extraction pressure and time conditions.The results show that both RSM and ANN methods can accurately predict,and it is concluded that the RSM model(R2=0.9940)has better prediction performance than the ANN model(R2=0.9879).And RSM and ANN-GA optimized the optimal extraction conditions and extraction rate as follows:extraction pressure 16.13 MPa,separation temperature 59.55 ℃,extraction time 100.6 minutes,extraction rate 11.36％;The extraction pressure is 16.5 MPa,the separation temperature is 60.95 ℃,the extraction time is 87.3 minutes,and the extraction rate is 11.32％.After experimental verification,the predicted values of the two methods are basically consistent with the experimental values.In addition,the Logistic model can well fit the kinetic process of subcritical CO2 isobaric extraction of grape seed oil(R2≥0.9990),and the model validation values and predicted values have a high degree of fit(R2≥0.9443).The research results provide theoretical and technical reference for the development and utilization of grape seed oil resources.
response surface methodologyartificial neural networksubcritical CO2 isobaric extractiongrape seed oildynamics
NETL
NSTL
万方数据
