Process Parameter Optimization Of Dual-Frequency Ultrasound Extraction Based On Small Sampel

The active components of traditional Chinese medicine(TCM)have a wide range of applications and practical values in chemical,medical,food and other fields.In the process of extraction,a set of excellent extraction conditions can ensure a higher extraction rate and save resources to the greatest extent.However,the principle of dual-frequency ultrasonic extraction process is complicated,so it is difficult to obtain the optimal process parameters.In the current process parameter optimization research,the optimal process parameters are generally obtained by analyzing the process parameter prediction model established.Therefore,it is very important to establish an accurate process parameter prediction model.However,it is hard to obtain a large number of effective data.Because,in the extraction field,data is usually obtained by manual acquisition.And the small amount of data means that it cannot fully express the characteristic information in the sample space.So the prediction model based on the limited sample data is usually poor in performance,resulting in a large error in the predicted results.To solve these problems,this paper proposes a method to generate virtual samples to expand the sample data set.In this method,Box-Behnken Design(BBD)was used to design the experiment and collect the original small sample data.Secondly,the prediction models of Extreme Learning Machine optimized by genetic algorithm(GA-ELM)and Support Vector Regression(SVR)were established respectively with the collected data,and the data were generated on the basis of the above two models.Then,according to the generated data,the virtual samples with better quality were obtained and selected in the iterative process through the Fireworks-Differential Evolution with Levy flight(L-FW-DE)algorithm proposed in this paper.The generated virtual sample and the original small sample were used to form the synthetic sample,and the prediction model of dual-frequency ultrasonic extraction was established according to the SVR principle.The SVR prediction model established by the original sample,GAELM prediction model built by the original sample and the SVR prediction model established by the synthetic sample were compared and analyzed.The results show that the prediction model based on the synthetic sample has the best performance.Finally,according to the SVR prediction model established by the synthetic samples,the optimal combination of process parameters was obtained by fireworks algorithm(FWA).The extraction of liquiritin from liquorice and liquiritin and paeoniflorin from compound Shaoyaogancao decoction by dual-frequency ultrasonic was carried out,and it was verified that the extraction rate was the highest under the optimized parameters.The experimental results show that the virtual sample generation method proposed in this paper can effectively generate reasonable virtual samples,and the synthetic samples composed of small samples and virtual samples can improve the accuracy of the prediction model,so as to get the optimal extraction conditions,and realize the optimization of the process parameters of dual-frequency ultrasonic extraction of active components in single TCM and compound TCM.The research has theoretical guidance and application value for industrial extraction process.