Construction Of Systematic Identification Methods For Mineral Traditional Chinese Medicines

Objective: To establish system identification method for rapid analysis of mineral Traditional Chinese Medicines(TCMs),and to provide new method for fast detection of mineral TCMs,the identification method combines the advantages of traditional identification methods and some modern analysis techniques such as X-ray diffraction(XRD),near infrared diffuse reflectance spectroscopy(NIRS)and Raman spectroscopy were explored.Methods: Based on the proposed thinking and technical structure form the system identification method of mineral TCMs,first,the applications of XRD,NIRS and Raman spectroscopy on the quality control of mineral TCMs were breadthways discussed.Then,Fluoritum,for example of practical application on the system identification method,was lengthways analyzed by using above-mentioned modern echnique combined with morphological identification and traditional physical and chemical analysis.The definite means are as follows:(1)The morphological identification and conventional physical and chemical analysis were used to identify all samples of Yangqishi and Yinqishi.XRD patterns which had been collected from sample powders were used to analyze phase composition.By comparison between XRD and morphological characteristic,the mineral origins of Yangqishi and Yinqishi were respectively clarified and identification of the two was guided.Then,the XRD fingerprint of Yangqishi was determined.And accord to the fingerprints and the intensity of each characteristic peak in XRD patterns of Yangqishi to calculate the similarity of different samples which use the cosine law and the correlation coefficient method,systematic cluster analysis was also used for the data.(2)In order to provide theoretical bases for rapid identification of mineral TCMs with NIRS,NIRS features from 51 kinds of familiar mineral TCMs were generalized and compared on the basis of the previous research,and then the characteristic spectral bands were determined and analyzed by referring to mineralogical and geological literatures.(3)Raman spectra of 67 batches of sample including Otolithum Sciaenae,Galaxeae Os,Ophicalcitum,Calcite,Stalactite and their mixture which had different content of Ca CO3 were collected,and the PLSR quantitative models were established by using an improved si PLS to optimize the characteristic spectral bands and using the Ca CO3 contents which were measured by EDTA titration method as references.The best model was also verified and evaluated.(4)58 batches of Fluoritum sample,a batch of quartz and a batch of calcite were collected from different sources.To preliminary identify their authenticity,the morphological characteristics and physical and chemical characteristics were inspected.And the EDTA titration method was used to determine the Ca F2 content in every sample for quality evaluation.Using the scanning electron microscope,the ultramicro characteristics of representative samples were observe.And using XRD technology,identification results were further collated.Raman spectra of sample including certified products,falsify products,doping products and easily-confused products were collected,and their spectral signature were respectively generalized and analysed;Then by matching the Raman spectra of light-colored sample(the Ca F2 content >75%)were with relevant Ca F2 content measured by EDTA titration method,the Raman spectral PLSR quantitative models of Fluoritum were established.Then a BP-ANN qualitative identification model of Fluoritum was established on the basis of the input data obtained from the multiple correlation coefficients with NIRS,and the model made it possible to distinguish various samples including raw Fluoritum,calcined Fluoritum,vinegar-quenching Fluoritum and spurious Fluoritum simultaneously through an integrated operation.NIRS from 74 batches of sample including Fluoritum and their mixture which had different Ca F2 content were collected and matched with relevant Ca F2 content measured by EDTA titration method,in order to establish the PLSR quantitative models of NIRS.Finally,compared and analysed the pros and cons in the applications of quality control for Fluoritum about above analysis methods,and probe into appropriatethinking about the system identification method for rapid analysis of Fluoritum.(5)By respectively using the SVM for classification(SVC)and SVM for regression(SCR)algorithm,the qualitative identification model for raw Fluoritum,calcined Fluoritum,vinegar-quenching Fluoritum and spurious Fluoritum and the quantitative models of Ca F2 content in Fluoritum sample were established with NIRS.And using the BP-ANN algorithm,the Raman spectral quantitative models of Fluoritum were established.Comparing above nonlinear models on the basic of intelligent algorithm with relevant conventional linear models,the preponderances of intelligent algorithms for the fast detection of mineral TCMs were discussed.Result:(1)According to the XRD study on Yangqishi and Yinqishi,it has been found experimentally that the mineral origin of Yinqishi is Talc schist and the mineral origin of Yangqishi is Tremolite and Actinolite.The geometric and topological characteristics of 10 certified products of Yangqishi are consistent.Using the XRD patterns of these 10 certified products obtain XRD fingerprint of Yangqishi which has 18 characteristic peaks.Similarityanalysis showed,similarity of XRD patterns common peak of certified products were higher(>98%),similarity of doping products is slightly lower(85%~97%),and the counterfeits has the lowest similarity(<30%).They can be distinguished with significant differences.And the results of cluster analysis are consistent with similarity analysis results.(2)It turned out that NIRS features of mineral TCMs were mainly at 8 000~4 000 cm-1,which vested in water,-OH and [CO32-] and so on.Absorption peaks of water had enough regularity.Generally,the structure water and-OH had a combined peak which was strong and keen-edged around 7 000 cm-1,the crystal water had two strong peak around 7 000 cm-1 and 5 100 cm-1,and water only has a broad peak around 5 100 cm-1.Because of differences in the crystal form and the content of water in mineral TCMs,NIR features of water in mineral TCMs which could be used to identify mineral TCMs were different.Mineral TCMs containing sulfate be rich in crystal water,mineral TCMs containing silicate generally had structure water,and mineral TCMs containing carbonate merely had a little of water,so it was reasonable for the NIR analysis to classify mineral TCMs with anionic type.In addition,because of differences in cationic type,impurities,crystal form and crystallinity,mineral TCMs had exclusive NIR features which vested in Al-OH,Mg-OH,Fe-OH,Si-OH,[CO32-] and so on,at 4 600~4 000 cm-1.Calcined mineral TCMs were often associated with water and main composition changes,also changes of the NIR features,which could be used for the monitoring of the process.(3)The established quantitative model for Ca CO3 content showed a prediction result that the average relative deviation of the prediction results is 2.71% and the average recovery rate was 100.46%,when the content is between 46.54%~99.97%,and when characteristic bands of 1 290~1 280,730~714,700~690,660~650,465~460,455~445,405~385 cm-1 had been optimized.The result also showed that the model using Raman spectroscopy and based on an improved si PLS can get a rapid determination for contents of 5 five kinds of Traditional Chinese Medicine containing Ca CO3.(4)The system identification of Fluoritum shows: by preliminary inspecting the morphological characteristics and physical and chemical characteristics in 58 batches of sample,there are 38 batches of certified products,15 batches of falsify products and 5 batches of doping products.Comparing the Ca F2 content measured by EDTA titration method,5 batches of sample have more impurities(the Ca F2 content<85%),in 38 batches of certified products,which should be doping products essentially;2 batches of doping product and a batch of falsify product have higher Ca F2 content(?85%),which should be certified products.The ultramicro characteristics of representative samples are different,and the detritus of certified products are blocky structure,which have sharp corners and smooth surface.The XRD analysis shows that certified products have vast Fluorite and little Quartz,on the contrary,the falsify products have vast and complicated impurities and little Fluorite.Fluoritum had 3 groups of characteristic peaks of Raman spectrum in the range of 310~325,720~1 500,1 700~1 900 cm-1,which can be used to identify its authenticity.The accuracy of BP-ANN qualitative identification model of Fluoritum with NIRS is 93.33% for 15 batches of sample in prediction set,which has a strong predictive ability.In the quantitative analysis of Raman spectroscopy,the established PLSR quantitative model for Ca F2 content shows a prediction result that RMSEP is 2.99%,R2 is 80.29%,the average relative deviation of the prediction results is 2.66% and the average recovery rate is 100.11%,when the content is between 77.76%~ 99.75%,and when characteristic bands of 1 550~1 540,1 530~1 520,1 510~1 500,540~535,530~525,475~450,300~250 cm-1 have been optimized.In the quantitative analysis of NIRS,the established PLSR model shows a prediction result that RMSEP is 4.17%,R2 is 95.26%,the average relative deviation of the prediction results is 4.99% and the average recovery rate is 102.86%,when the content is between 44.28%~99.95%,and when characteristic bands of 7 500~6 300,5 900~5 500,5 100~4 300 cm-1 have been optimized.The NIRS quantitative model also has a strong predictive ability,and is wider applicable scope of the Ca F2 content than the applicable scope of Raman spectral model.(5)Compared with conventional linear model,the predictive ability of models based on spectroscopy and intelligence algorithm is stronger.In practical terms,the accuracy of SVC model of Fluoritum with NIRS is 93.33%;The established SVR model of Ca F2 content in Fluoritum with NIRS shows a prediction result that RMSEP is 4.10%,R2 is 96.04%,the average relative deviation of the prediction results is 5.03% and the average recovery rate is 102.23%;And the BP-ANN model of Ca F2 content in Fluoritum with Raman spectroscopy shows a prediction result that RMSEP is 2.73%,R2 is 85.64%,the average relative deviation of the prediction results is 2.30% and the average recovery rate is 99.74%.Conclusion: The morphology identification is an intuitionistic and convenient method for identification of mineral TCMs,but it has a strong subjectivity,also has much difficulty in the identification of powder and some samples which is lack of specificity.The conventional physical and chemical identification has poor specificity,and is easy to be affected by impurities and associated minerals.The XRD technology can accurately identify and evaluate mineral TCMs and their powder,by parsing the phase composition and content.The NIRS and Raman spectroscopy technology,by establishing quantitative and qualitative modes with chemometrics methods,can achieve rapid identification and quality evaluation for mineral TCMs,which have their own strong points and weaknesses and is worthy to be popularized and as rapid detection technology of mineral TCMs.The above threemodern analysis techniques have good applicability,meanwhile their analysis speeds are quick and their analysis results are accurate and reliable.In this paper,the artificial intelligence methods of BP-ANN and SVM algorithm were also applied for the NIRS and Raman spectrum model in order to improve models.Based on artificial intelligence methods,the accuracy of qualitative and quantitative analysis of mineral TCMs were greatly improved,which can be applied to practical analysis.