Research On Quality Evaluation For Lingnan Aromatic Chinese Medicinal Materials Based On Bionic Olfaction With Fusion Techniques

Chinese medicinal materials(CMMs)have been used for the treatment and prevention of human diseases for thousands of years in China.It has been confirmed that CMMs has played an important role in preventing and treating the chronic,infectious diseases and epidemic.So CMMs are getting more and more international attention and have been widely recognized by the international community.However,due to the diverse specifications of the herb components and complex sources of CMMs,occurrences of fake and poor quality products have emerged in the CMMs market.This has seriously restricted the global use of CMMs.Therefore,the study of effective methods for quality evaluation of CMMs is becoming a peculiarly important issue.Traditional manual quality identification for CMMs has strong subjectivity and poor repeatability,while physicochemical quality identification for CMMs is cumbersome,time-consuming and complicated.So it is difficult to popularize them to the identification of CMMs.The bionic olfactory system also called electronic nose(E-nose)can detect one or mixed odor by simulating biological olfaction system.Compared with traditional odor analysis technology,such as gas chromatography,mass spectrometry and flame ionization detection,bionic olfaction technology has the advantages of fast,simple,high accuracy.Thus it has been widely used in the fields of food,agriculture,medicine and environmental detection.In this paper,the application status of bionic olfaction technology in detection,processing and identification of odor information in CMMs was analyzed.Combined with the analysis technology of gas chromatography-Mass Spectrometer(GC-MS),the paper has studied the identification mechanism of bionic olfaction technology in odor processing and odor identification for CMMs.In order to explore the correlation between odor characteristics of electronic nose sensor and odor components in CMMs,the performance of different pattern recognition methods for the classification of CMMs was compared.Prediction models for key odor components in CMMs based on the bionic olfaction technology and a grade evaluation method for CMMs were proposed.Research work and conclusions are as follows:(1)In this paper,10 kinds of CMMs were selected as the objects.Taking Amomum villosum and Galangal from different habitats as examples,the basic principle and data processing methods of odor identification detected by PEN3 were studied.By comparing the experimental results of three pattern recognition methods:partial least squares(PLS),support vector machine(SVM)and convolutional neural network(CNN),the classification accuracy of SVM and CNN is similar and the accuracys of training set and test set are more than 98%,which are higher than those of PLS model.It shows that SVM and CNN models have more advantages over than PLS model in classifying non-linear data.(2)With the help of the mass spectrometry library in Institute of National Standards and Technology(NIST 14.L),the volatile constituents of Amomum villosum and Galangal from different habitats were analyzed using GC-MS.the results showed that there were 65 constituents identified in Amomum villosum samples.The 65 constituents belonged to six different chemical categories:terpenes(42),alcohols(12),esters(4),ketones(3),alkanes(2)and aldehydes(2).Among them,37 constituents were identified in Guangdong Amomumvillosum,accounting for 93.51%of the total volatile components;40 constituents were identified from Guangxi Amomum villosum,accounting for 97.13%of the total volatile components;and 47 constituents were identified from Hainan Amomum villosum,accounting for 97.41%of the total volatile components.The average oil yield of Amomum villosum was about 0.0137 m L/g(1.37%,V/W).There were 56 constituents identified in Galangal samples.The 56 constituents belonged to four different chemical categories:terpenes(37),alcohols(12),esters(4)and others(3).Among them,52 constituents were identified in Guangdong Galangal,accounting for 95.25%of the total volatile components;43 constituents were identified from Guangxi Galangal,accounting for 93.12%of the total volatile components;and 42 constituents were identified from Yunnan Galangal,accounting for 92.25%of the total volatile components.The average oil yield of Galangal was about 0.0083 m L/g(0.83%,V/W).Lastly,the common volatile components and specific volatile components in the samples were analyzed,which verified the response changes in different sensors caused by these components.(3)The correlation between the characteristics of electronic nose sensor and the odor components in CMMs has been studied.Principal component analysis(PCA)was used to study the correlations among sensor characteristics,sample origin and odor components.The results were displayed using Bioplot.It showed that terpenes,alcohols and esters in Amomum villosum were closely related to the six sensors(S1,S3,S5,S7,S8 and S9)in PEN3,terpenes,alcohols and esters in Galangal were closely related to the three sensors(S7,S8 and S9)in PEN3.According to these different response values caused by terpenes,alcohols and esters,the identification of CMMs from different habitats can be successfully carried out(4)According to the correlation analysis of sensors characteristics and odor components,three prediction models for key odor components in CMMs based on PLS,SVM and CNN algorithm were studied and constructed.Prediction results for 6 key odor components((lr,4r)-(+)-campho,?-caryophyllene,d-cadinene,camphene,bomeol and bomyl acetate)in Amomu,villosum have showed well.Most parameter values(Rc)2 and(Rp)2 were about 0.9 in PLS model.In SVM model,the parameter values(RC)2 and(Rp)2 for borneol were the biggest,(Rc)2=0.952 and(Rp)2= 0.941 and except(3-caryophyllene,the parameter values(Rc)2 and(Rp)2 were over 0.92.In CNN model,the parameter values(Rc)2 and(Rp)2 for camphene were the biggest,(Rc)2=0.963 and(Rp)2=0.931.The parameter values(Rc)2 and(Rp)2 for other components were over 0.93.Prediction results for 3 key odor components(?-caryophyllene,1,8-cineole and ?-farnesene)in Galangal showed well.In PLS model,all parameter values(Rc)2 and(Rp)2 were over 0.93.In SVM model,all parameter values(Rc)2 and(Rp)2 were over 0.945.In CNN model,the parameter values(Rc)2 and(Rp)2 were over 0.95.The prediction results show that the performance of SVM and CNN models are higher than that of PLS model,which further verifis that the performance of SVM and CNN are better than that of PLS method when optimizing non-linear data..At the same time,the research shows that,based on bionic olfactory technology,the key odor components in CMMs can be predicted well using some appropriate pattern recognition methods.(5)In order to study the odor characteristic in Amomum villosum from different grades,a quality identification model made up of classification,quantitative analysis and grade judgement was proposed.Firstly,PLS,SVM and CNN models were constructed to classify the different grades of Amomum villosum.Then,two key odor components,borneol acetate and(lr,4r)-(+)-campho,were predicted based on odor component analysis.Finally,the grading evaluation model of Amomum villosum was constructed according to three evaluation indexes:borneol acetate content,(1r,4r)-(+)-campho content and the content ratio of borneol acetate/(1r,4r)-(+)-campho.The CNN model was used as an example to evaluate the quality of Amomum villosum.The experimental results showed that the correct rate of training set was 95%and the correct rate of test set was 100%in first-class samples.In second-class samples,the correct rates were 85%and 80%,separately.In third-class samples,the correct rates of training set and test set were both 80%.