Study On Modeling And Frequency Optimization Of Ultrasonic Extraction Process Of Natural Product And Its Application

Ultrasonic extraction of effective component from natural products(they are mainly referred to natural plants such as Chinese medicinal materials and food materials in this thesis)is a green extraction technology with high efficiency,less residue and non-pollution.In recent years,the rapid development of continuous expansion of ultrasonic natural extraction industry scale has led to urgent need optimization and control of the extraction process.The modeling of ultrasonic extraction process is the basis and prerequisite of process optimization and control.Currently,more in-depth studies on the dynamic modeling of ultrasonic natural product extraction process have been done by many domestic and foreign scholars.However,the study of ultrasonic extraction process by using machine learning theory has not been well investigated.In addition,in terms of extraction efficiency,aiming at the optimization problem of ultrasonic frequency whithin a continuous wide frequency range is also rarely reported.Therefore,it is of great scientific and practical value to establish practical extraction model according to the actual situation of ultrasonic natural product extraction process and to optimize the ultrasonic frequency of ultrasonic natural product extraction process.In view of this,some important issues are deeply studied in this thesis,mainly including dynamics model,soft sensor model based on support vector machine theory,the optimization of ultrasonic frequency and its application for ultrasonic natural product extraction process.The main research concents are summarized as follows.(1)Aiming at the lact of ultrasonic frequency for the current dynamic model,based on the mass transfer dynamic model and the mechanism of ultrasonic extraction,an improved ultrasonic dynamic model is proposed by introducing the ultrasonic frequency.Furthermore,to verify the validity of the improved model,based on the optimized extraction variables by ultrasonic extraction experiment of glycyrrhizic acid from glycyrrhiza uralensis,dynamic models of the relationship between glycyrrhizic acid concentration and ultrasonic power,ultrasonic frequency and extraction temperature are established.The experimental and simulation results show that the proposed model is feasible and effective.(2)Aiming at the problem of the dynamic model with poor universality and portability,a prediction model of the extraction of glycyrrhizic acid from glycyrrhiza uralensis based on SVR is established.The prediction results verify the effectiveness of the proposed model.In addition,aiming at the problem of SVR model with long training time,based on the analysis of least squares support vector machine(LSSVM)theory,a prediction model of the extraction of glycyrrhizic acid based on LSSVM is established.Finally,the prediction results are compared with SVR model.(3)Aiming at the problem of parameters optimization of LSSVM model,an improved fruit fly optimization algorithm is proposed by introducing a self-adaptive step factor and chaos optimization algorithm.Based on the Markovian convergence analysis theory,it is proved that the proposed algorithm converges to the global optimal solution and the effectiveness is verified by simulation tests.Based on this,a prediction model of the extraction process of ultrasonic glycyrrhizic acid based on CDSFOA-LSSVM is established after parameters are optimized by using the CDSFOA algorithm.Finally,compared with the prediction results of SVR and LSSVM models,the proposed model has faster training speed and higher prediction accuracy.(4)Aiming at the problem of the aforementioned models with no online forecasting ability,an online learning non-bias LSSVM was proposed by using the unbiased LSSVM and the online LSSVM.Further,for multi-target output problem,a multiple-input multiple-output online learning non-bias LSSVM algorithm is proposed based on the combination of ONBLSSVM and multi-input multi-output LSSVM.To further improve the speed and accuracy of the proposed algorithm,it is optimized by introducing the weighting factor and using the on-line recursive learning method.Based on this,to predict the concentration of two effective components,such as rutin and quercetin from the stalk of Euonymus Alatus(Thumb.)Sieb,a multi-objective prediction model of ultrasonic extraction process based on the improved MIMO-ONBLSSVM is established,and the effectiveness of this model is verified.(5)Aiming at the problem that the used ultrasonic frequency during the current ultrasonic extraction process is too few and can not be continuously selected,an optimization method of ultrasonic frequency is proposed.This is,coarse search of the optimum frequency band in a wide band is performed,and then,fine search of the optimum frequency in the obtained optimum narrow band is performed.Thus,the purpose of improving the extraction efficiency of ultrasonic natural product is achieved by continuously searching the optimal ultrasonic frequency.Based on this concept,by using automatic control,power electronics and computer technologies,a visual ultrasonic extraction system is designed to continuously automatically search for the optimal ultrasonic frequency in a wide frequency range.By using the designed extraction system based on the frequency optimization method,extraction experiments and frequency optimization of two natural products,such as tomato and sophora japonica are researched,respetively.Finally,the optimal ultrasonic frequency is obtained and the extraction efficiency is improved.