Juan Chen*,XiaoyiHuang,Yan leiQi,Xin Qi,Qing Guo

College of Information Science&Technology,Beijing University of Chemical Technology,Beijing 100029,China

Process Optimization of Ultrasonic Extraction of Puerarin Based on Support Vector Machine☆

Juan Chen*,XiaoyiHuang,Yan leiQi,Xin Qi,Qing Guo

College of Information Science&Technology,Beijing University of Chemical Technology,Beijing 100029,China

A R T I C L E I N F O

Article history:

Received 25 December 2013

Received in revised form 10 February 2014 Accepted 1 April2014

Available on line 20 June 2014

In ultrasonic extraction technology,optimization of technical parameters often considers extraction medium only,without including ultrasonic parameters.This paper focuses on controlling the ultrasonic extraction process of puerarin,investigating the influence of ultrasonic parameters on extraction rate,and empirically analyzing the main components of Pueraria,i.e.,isoflavone compounds.A method is presented combining orthogonal experimental design with a support vector machine and a predictive model is established for optimization of technical parameters.From the analysis with the predictive model,appropriate process parameters are achieved for higher extraction rate.With these parameters in the ultrasonic extraction of puerarin,the experimental result is satisfactory.This method is of significance to the study of extracting root-stock plant medicines.

?2014 Chemical Industry and Engineering Society of China,and Chemical Industry Press.All rights reserved.

1.Introduction

Isoflavone compounds,the main componen ts of Pueraria,have functions of decreasing blood sugar,expanding peripheral vascular, especially dilating coronary artery blood vessels and inhibiting aggregation of platelets,all of which have positive effects on the treatment of diabetes and on the improvement of peripheral nerve injury,retinopathy and kidney lesions resulted from their microangiopathy[1].Traditional extraction and separation methods(such as decocting method)give lowextraction rate of active ingredients and low impurity clearance,restricting the progress in utilization of puerarin.Thus new methods to extract active ingredients in puerarin have emerged,among which ultrasonic extraction has become a hot area.Ultrasonic extraction has the advantages of low temperance,short time and high extraction rate[2].In the traditional ultrasonic extraction of plant components,the optimization of technical parameters only considers the relationship among several factors,such as extraction rate,concentration of extraction solvent, liquid-so lid ratio,extraction temperature,extraction time,and extraction times.The mechanism of ultrasonic extraction and the mass and heat transfer process in the extraction are not taken in to account.Thus the optimization has considerable limitations.In this work,the effect of technical parameters on extraction rate is analyzed and the effects of these technical parameters and medium parameters on the extraction rate are determined experimentally.Based on the intensified single-frequency ultrasonic extraction,a 20 kHz and 40 kHz double-frequency ultrasonic extraction device is designed and the effect of double-ultrasonic intensification on the extraction of active ingredients is studied.

The process optimization for puerarin extraction focuseson the process factors in the extraction experiment and determines their optimum combination.A commonly used optimization method is the orthogonal experimental design.This method involves finding the best combination of factors[3]with the following shortcomings.(1) The orthogonal experimental design uses part of the experiment instead of the whole one,so it cannot analyze the effect and their interaction one by one as the whole experiment does,and if the interaction exists the confounding of interactions will appear.(2)The orthogonal experimental design can only lead to a more appropriate combination of parameters but not their quantitative relationship with extraction efficiency.(3)Although the variance analysis of orthogonal experimental design can evaluate the experimental errors and the importance degree of experimental factors,it cannot provide an accurate evaluation for the significance of every factor.Thus,there are some limitations associated with the use of orthogonal experimental design to optimize the process. In this study,we combine the support vector machine(SVM)method and the orthogonal experimental design and employ them in modeling and simulation the extraction of puerarin to obtain the optimum combination of factors and provide quantitative relation of influencing factors to experimental result.The optimum combination of factors is verified and compared by experiments.

2.Mechanism Analysis of Ultrasonic Extraction of Puerarin

In the ultrasonic extraction process,technical parameters of ultrasonic wave and medium parameters affect the extraction rate significantly.The leaching and extraction of active ingredients from plants can be accelerated by mechanical and cavitation actions and mass transfer of ultrasonic wave.Ultrasonic mechanical vibration can break cell walls and lead explosion of cavitation bubbles,and the seepage is related to heat and mass transfer in the extraction process.The effect of parameters,such as concentration of extraction liquid C,liquidsolid ratio β,extraction temperature T,extraction time t,extraction times m,the frequency required for ultrasonic energy to break cell walls f1,broken strength I1,the frequency of ultrasonic cavitation f2, and the strength of cavitation I2,on the optimization of extraction process is comprehensive,and the extraction rate can be written as

An ultrasonic extraction involves heating,mechanical and cavitation mechanisms.The ultrasonic cavitation effect and its strength and the ultrasonic frequency are related to the ultrasonic strength,so studying the cavitation mechanism in the extraction of plant component is of great importance.Experiments have found that when the ultrasonic frequency is lower thanthe natural resonance frequency of forming cavitation bubbles,bubble nuclei will disappear and ultrasonic cavitation will occur.When the ultrasonic frequency is higher thanthe natural resonance frequency,bubble nuclei will undergo complex nonlinear vibration instead of disappearance.Only if the ultrasonic frequency equals the natural resonance frequency canthe ultrasonic bubbles and the transmission attain the most effective energy coupling and canthe ultrasonic cavitation occur.

2.1.Single-frequency ultrasonic cavitation bubble

Cavitation involves the motion of bubble surface.When oscillatory ultrasonic waveswith sound p ressu re PAsin2πft spread in a liquid,tiny cavitation bubbles in the liquid will be stretched and compressed.The equation of motion for bubble surface with the effect of sound waves is[4]

where R is the radius(m)of cavitation bubble at time t,R0is the initial radius of bubble,andare the acceleration(m·s?2)and velocity (m·s?1)of particle on the bubble surface,ρlis the density(kg·m?3) of continuous medium,P h is the static pressure(Pa)of the fluid,K is an index(K=1 under isothermal condition),σis the surface tension (N·m?1),μis the liquid viscosity(Pa·s),P0is the initial sound pressure ofm edium(Pa),Pvis the vapor pressure in the bubble(Pa),PAis the amplitude of sound pressure(Pa),and f is the ultrasonic frequency.

In the experiment,some parameters are set as follows.In water at 25°C,R0=2.0μm,σ=7.25×10?2N·m?1,μ=10?3Pa·s,K=1, Ph=1.013×105Pa,P0=1.013×105Pa,Pv=3.2718×103Pa,ρl= 1000 kg·m?3,PA=131.7×103Pa,and f is 20,25,40,and 80 kHz.

Fig.1 shows the change of bubble radius with time.Under the same sound pressure and experimental conditions,the maximum radius of cavitation bubble decreases gradually with the increase of ultrasonic frequency.At certain frequency,the process changes from transient cavitation to steady-state cavitation.Therefore,to enhance the effect of ultrasonic cavitation in an extraction process,low frequency should be used.

Fig.1.The radius of cavitation bubbles under different frequencies.

For initial bobble radius of20μm,frequency of20 kHz,and different amplitudes of sound pressure PAwith other parameters constant,the radius of cavitation bubble varies with time,as shown in Fig.2.As the amplitude of sound pressure increases,the ultrasonic cavitation is changed from steady to transient one.The bubble en larges to a size with its radius several times the initial radius.The grow th and collapse become more violent and the survival time of cavitation bubble is longer.

The simulation result and analysis show that higher amplitude of sound pressure will make cavitation proceed more violently.In this experiment,ultrasonic wave has a frequency of 20 kHz,power of 900W, sound intensity of 796.2W·cm?2,and amplitude of sound pressure of 4802.5 kPa.

2.2.Double-frequency ultrasonic cavitation bubble

In a continuous medium,the bubble surface under doubled frequency f1and f2satisfies the equation of motion[4]:

Fig.2.Transient cavitation.

Fig.3.The radius of cavitation bubble with time under single and double frequencies.

In the experiment,the double-frequency ultrasound generator gives frequency f1of 20 kHz(power of 900W),sound in tensity of 796.2W·cm?2and amplitude of sound pressure of 4802.5 kPa;and frequency f2of 40 kHz(power of 80 W),sound in tensity of 0.71 W·cm?2and amplitude of sound pressure of 143.5 kPa.With initial radius of cavitation bubble R0=2.0μm,the simulation result is shown in Fig.3.In the single-frequency ultrasonic field with a frequency of 20 kHz,the radius of bubble increases to almost 700R0.In the double-frequency ultrasonic field with a frequency of 20 kHz+40 kHz,the radius of bubble increases to around 710R0while the time of collapse is the same.Therefore,the double-frequency ultrasonic generates better cavitation effect than single-frequency.

With the same process parameters and extraction time of 40 min, the metallographic microscopic photographs for ultrasonic extraction of puerarin are shown in Fig.4.With double-frequency ultrasonic, cell wall crack appears and material seepage is obvious,as shown in Fig.4(b).It is mainly because of the explosion of cavitation bubbles caused by one ultrasound of double-frequency,with a new cavitation nucleus from self-reactivation and another from the other ultrasound. On the one hand,this effect facilitates faster mass transfer in the solution,enhances the uniformity of ultrasonic effect,and promotes chemical compositions of protoplasm to dissolve in the solution.On the other hand,the double-frequency radiation enhances the mechanical disturbance in the medium and more air enters the liquid through the liquid surface,increasing cavitation nuclei and intensifying the cavitation. Thus the use of double-frequency ultrasonic extraction can increase the extraction rate.

3.Experimental

3.1.Apparatus and chemicals

The apparatus and chemicals used in this work include:UV-VIS spectrophotometer(UV-2102),analytical balance,reference puerarin substance(China Food and Drug Verification Institute,batch number 110752-200912),reagents of chemical or analytical grade,and distilled water.The puerarin used was obtained from Xinxiang district,Henan Province.The ultrasonic extracting device[5]was made in our laboratory.

3.2.Methodology

For preparation of reference substance solution,5.4 mg puerarin, dried to constant mass,was put in a 25 ml volumetric flask with 95% ethanol content,producing a 216m g·L?1reference substance solution.

The standard puerarin reaches them axim um absorp tion[6]at UV w avelength of250 nm.A certain volume of standard puerarin reference solution,i.e.,0.1,0.2,0.4,0.6,0.8 and 1.0m l,was placed in a 10ml volumetric flask and ethanol content of95%was added as the reference solution.A UV spectrophotometer at 250 nm was used to measure the absorbance.The standard curve of puerarin is shown in Fig.5.

When the absorbance measured by the UV spectrophotometer is factored into the equation A=0.0901C?0.0025,concentration C(m g·L-1)and extraction rate η can be calculated.

where M(g)is the total mass of puerarin and V(L)is the total volume of extraction liquid.

Fig.5.Standard curve of puerarin.

Fig.4.The metallographic photos for ultrasonic extraction of puerarin for 40m in.

Fig.6.Effect of ethano l con ten t on the extraction rate.

4.Analysis

Extraction of puerarin using double-frequency ultrasound is studied under these conditions:breaking w all frequency 40 kHz,breaking w all strength 0.71W·cm?2,ultrasonic frequency 20 kHz,and cavitation intensity 796.2W·cm?2.The factors affecting the extraction process include:extraction time,extraction temperature,ultrasonic parameters (change in cavitation in tensity alters ultrasonic parameters and achieves a duty cycle through cavitation power),ethanol con ten t,and liquid-so lid ratio[7].In an ultrasonic extraction process,temperature is a parameter that increases with time and gives a set of optimum parameters through optimization of the process.These optimum parameters will be used in the extraction,while temperature and time cannot be used as the determining factors for the design of orthogonal experiment.The experiment is designed as follows.

4.1.Effect of ethanol content on eχtraction rate

With the same puerarin medium,ultrasonic parameters and environmental parameters,puerarin is extracted with different ethanol con ten t from 50%to 90%.The result is shown in Fig.6,in which the absorbance reaches a maximum value at ethanol con ten t of 60%.The relationship between absorbance and extraction rate is calculated by Eq.(4)and a maximum extraction rate of 0.812% is obtained.

4.2.Effect of liquid-solid ratio on eχtraction rate

Fig.7 shows that the absorbance decreases as the liquid-solid ratio increases from 20:1 to 60:1.The extraction rate attains the highest value,0.895%,at liquid-solid ratio of 50:1,calculated from the equation relating the absorbance and extraction rate.

4.3.Effect of double-frequency operating time on eχtraction rate

Fig.7.Effect of liquid-solid ratio on the absorbance.

Fig.8.Effect of double-frequency action.

The operating time with double-frequency(f1=40 kHz,f2= 20 kHz)is referred to as the alternate operating time rate of breakingwall ultrasonic and cavitation ultrasonic.The alternate action time of the double-frequency ultrasonic is set as 2,5,10,and 20 min,and the experiment is stopped when the combined operating time is 40m in. Fig.8 gives the relationship between ethanol content and absorbance. The change of absorbance with double-frequency operating time rate is not considerable.In the design of experiment,this factor is not considered.In the entire experimental process,the double frequency functions simultaneously,so only the effect of other static factors on the extraction rate is taken in to account.

4.4.Effect of duty cycle of power ultrasonic on eχtraction rate

The duty cycle(referred to the working gap rate)of cavitation power ultrasonic changes within tensity I2of cavitation power ultrasonic,and the changing range of the time duty cycle rate is from 1 s:3 s to 6 s:1 s.As shown in Fig.9,the absorbance reaches the maximum value at the duty cycle of5 s:1 s.According to the relationship between absorbance and extraction rate,the extraction rate can reach 0.938%at most.

4.5.Design and an alysis of orthogonal eχperiment

Based on single-factor experiment,ethanol content,liquid-solid ratio and duty cycle of cavitation power ultrasonic are considered. Three levels are set for each factor,as shown in Table 1.

From different solutions in the design of orthogonal experiment, referred to as the orthogonal table L9(34)and the determination formula of absorbency,the content of puerarin in each solution can be obtained and measurement results of the sample are shown in Table 2.K1,K2, and K3refer to the sum of extraction rates of these three level factors, and k1=K1/3,R refers to the difference in maximum and minimum values of K1,K2,and K3.It can be deduced that the effect of each factor on puerarin is as follows:liquid-solid ratio＞ethanol con ten t＞duty cycle.The final choice for the optimum process condition is A2B3C3, with liquid-solid ratio of 50:1,ethanol content of 70%and duty cycle of 6 s:1 s.

Fig.9.Effect of intensity of cavitation.

Table 1The factors and levels that affect extraction rate

From the optimization result of orthogonal experimental design, some combinations are used in the experiment and the result does not exceed the scope considered,suggesting that the experimental result cannot be easily obtained through direct analysis.The rest of this paper will combine the SVM regression analysis and orthogonal experiment to obtain a mathematical model for extraction process of Pueraria and predict the best parameters for ultrasonic doublefrequency extraction.

5.A Soft Sensor Modeling Based on LS-SVM

5.1.SVM theory

SVM is a new supervised learning method based on statistical learning theory proposed by Vapnik and his research team in 1995[8].SVM has many excellent features[9],such as sparse solution,use of kernel technology that obviates the need for local minimum solution,and strong generalization ability with small sample.The basic idea of SVM regression is that,with a nonlinear mapping,map the input data in to a high dimensional feature space,then makes linear regression and solves the convex optimization problem in this space.The essence of SVM is to convert a nonlinear function estimation problem in original space to that in a high dimensional space.

5.2.LS-SVM

LS-SVM[10]converts the inequality constraints in SVM to equality constraints and the empirical risk error changes from 1-norm to 2-norm,so the quadratic optimal problem-solving is transform ed to the linear equation problem-solving.LS-SVM greatly increases the convergence speed[11].

For easy measurement of auxiliary variables,the training samples are(χi,yi),…,(χl,yl),whereχi∈Rmwith mas the input space dimension and yi∈R is the output vector(i=1,2,…,l).Characteristic sample data are mapped in to a high dimensional feature space where SVM regression function is constructed.The fitting function is

where w is the weight vector with high dimension,?(χ)is the mapping of input vector χ in the feature space,and bis the threshold value. Compared with the standard SVM,LS-SVM is an improvement based on the minimum regularization theory and greatly simplifies the SVM solution process.Optimal w and b can be obtained by minimizing the following objective function

with constraint condition

5.3.Predicting model with LS-SVM for optimized eχtraction process

This study focuses on the content of isoflavones in puerarin.Combining the SVM method and orthogonal experimental design,we decide the experimental condition before experiment.Normalized experimental data are used as the learning sample of LS-SVM to train and establish a predicting model,as shown in Fig.10.The normalization method,the difference of data and average value divided by standard deviation,is adopted.The kernel function[12]of SVM adopts a polynomial form.

5.4.Simulation analysis of eχtraction processes

Extraction processes are simulated with the prediction model of process optimization.The ethanol content changes from 50%to 70%at liquid-so lid ratio of 60 and duty cycle of 6 s:1 s.The effect of ethanol content on the extraction rate is shown in Fig.11.The extraction rate decreases monotonously with the increase of ethanol content.Thus we select the extraction medium with ethanol con ten t of 50%.

With the ethanol content kept at 50%,the liquid-solid ratio changes from 40:1 to 60:1 with the highest duty cycle of ultrasonic.Fig.12 shows that the extraction rate increases with liquid-solid ratio,so the ratio of 60:1 is selected.

With ethanol con ten t of 50%and liquid-solid ratio of 60:1,the power ultrasonic duty cycle changes from 4 s:1 s to 6 s:1 s.Fig.13 shows that with the increase of duty cycle,the extraction rate first decreases and then increases,with the highest extraction rate at the duty cycle of 4 s:1 s.

Table 2Design and result of orthogonal experiment

Fig.10.Combination of SVM and orthogonal experiment.

Fig.11.Effect of ethanol content on extraction rate.

Th rough the stimulation,it is found that the extraction rate of puerarin reaches the highest point at ethanol content of 50%,liquidsolid ratio of 60:1,and duty cycle of 4 s:1s.

Fig.12.Effect of liquid-solid ratio extraction rate.

5.5.Result

The model trained by the SVM is used for prediction and the result is compared to that optimized by the orthogonal experiment with the best parameters.The predicting result is 1.1128 with the SVM optimization, while it is1.0984 with the orthogonal experiment.The effect of parameters on the extraction rate is as follows,ethanol content＞duty cycle＞liquid-solid ratio.

Table 3 gives verification results of the two optimization methods. The predicting method based on the orthogonal experiment and SVM can comprehensively reflect the effect of parameters on the extraction rate and avoid the disturbance brought by other uncertain factors.The experimental verification indicates that the predicting value with the LS-SVM method is higher thanthat of the orthogonal method.The relative error is within 1%.

6.Conclusions

The mechanism,cavitation effect,and transient effect in ultrasonic extraction of active ingredients of plants are analyzed.The effect of single-frequency 20 kHz ultrasonic is simulated and the maxim um value of cavitation collapse changes from one to two peak values. Double-frequency 40 kHz/20 kHz increases them ass transfer,enhances the uniformity of ultrasonic effect,and accelerates the dissolution of chemical ingredients in the protoplasm.The double-frequency irradiation intensifies the mechanical disturbance in the medium and moreair enters the liquid through the surface,increasing the number of cavitation nuclei.The cavitation effect increases continuously,raising the extraction rate.

Fig.13.Effect of power ultrasonic duty cycle on the extraction rate.

Table 3Experimental verification

Based on the orthogonal experiment,appropriate parameters in the experiment are decided and the experimental data are used to build a prediction model.This model is effective in process optimization.

The influence of process parameters and medium parameters on the extraction rate of active ingredient in puerarin is analyzed.In the ultrasonic double-frequency extraction,the influencing factors are the cavitation power of ultrasound,liquid-solid ratio,and the volume factor of ethanol solution.Their effects on the extraction rate are as follows:liquid-solid ratio＞ethanol con ten t＞duty cycle.For the method combining the SVM and orthogonal experiment,the optimum parameters are obtained:ethanol content 50%,liquid-solid ratio 60:1,and duty cycle 4 s:1 s,giving the predicting result of 1.1128.The experimental verification shows that the error of prediction is within 1%.Thus it is feasible to design the orthogonal experiment by the LS-SVM method,which is of significance to future research on traditional root plant medicine.

[1]H.Y.Wang,D.C.Tang,The analysis of the effect of the active ingredient in the puerarin on the treatment of diabetes and its complications,Chin.J.Clin.Rehabil.9 (27)(2005)134-135(in Chinese).

[2]L.Wang,D.Li,C.L.Bao,J.Y.You,Z.M.Wang,Y.H.Sh i,H.Q.Zhang,“Ultrasonic extraction and separation of anthraquinones from Rheum palmatum L.”,Ultrason. Sonochem.155(2008)738-746(in Chinese).

[3]H.Y.W u,B.G.Chang,C.C.Zhu,A special case of the genetic algorithm—the method of orthogonal experiment,J.Softw.12(1)(2001)148-153.

[4]T.Q.Qiu,R.H.Zeng,X.Y.Zhang,“Mechanism of extraction enhanced by dualfrequency ultrasonic”,J.South China Univ.Technol.(Nat.Sci.Ed.)34(8)(2006) 89-93(in Chinese).

[5]X.Qi,X.Y.Zheng,J.Chen,“A double frequency and bidirectional turbulent ultrasonic device for extraction”,Proceedings of the IEEE International Conference on Automation and Logistics,IEEE Press,Zhidong Wang(Publication Chair)Qingdao,China, 2008,pp.2710-2713.

[6]J.P.Guo,Q.Sun,Q.Zhou,T.Y.Zhao,Y.T.Gao,The discussion of the different extracting process of flavonoid in puerarin,Chin.Herb.Med.26(10)(1995) 522-523(in Chinese).

[7]X.G.Xu,F.Y.Lu,X.C.Sui,N.Lu,“The optimization of the extracting process of puerarin through orthogonal experiment”,LishiZhen Med.Mater.Med.Res.11(09) (2000)780-781(in Chinese).

[8]V.N.Vapnik,The Nature of Statistical Learning Theory,Sp ringer Verlag,New York, 1995.

[9]X.G.Zhang,The statistical learning theory and support vector machine(SVM),Acta Autom.Sin.26(01)(2000)32-42.

[10]J.A.K.Suykens,T.Van Gestel,J.De.Brabanter,B.De.Moor,J.Vandewalle,Least Squares Support Vector Machines,World Scientific,Singapore,2002.

[11]H.F.Wang,D.J.Hu,Comparison of SVM and LS-SVM for regression,Proceeding of IEEE:2005 International Conference on Neural Networks and Brain,IEEE Press, Beijing,China,2005,pp.279-283.

[12]T.Kavzoglu,I.Colkesen,“A kernel functions analysis for support vector machines for land cover classification”,Int.J.Appl.Earth Obs.Geoinf.11(2009)352-359.

☆Supported by the National Natural Science Foundation of Ch ina(21146009, 21376014).

*Corresponding author.

E-mailaddress:jchen@m ail.buct.edu.cn(J.Chen).

Ultrasonic extraction

Orthogonal experimental design Support vector machine

Extraction rate