Studies On Rapid Non-destructive Quantitative Analysis Of Fructus Cnidii Extract Using NIR Spectroscopy

Extensive application of the computer technology marks the coming of information era. It brings the new opportunity for the chemical discipline development. Medicine analyses as the important component of analytical chemistry, already has developed into the modern medicine analysis relying mainly on instrument analysis. Data and information that modern analytical instrument and means offer to us are a large amount of or even magnanimity, meanwhile give analyzing chemists put forward a question, i.e. how to deal with these primitive analysis data high-efficiently, draw the useful information regarding the material component and structure from it, become current analysis workers' main task gradually.With the rapid development of the optical technology and computer technology, the digitization of the spectrum signal becomes the typical characteristic of the modern spectrum analytical instrument, in addition method study of chemometrics constant deepening, making theory and technology of analyze science weak signal treatment riper and riper. It already developed into an independent analytical technology, and changed the pastappearance of analytical chemistry that lag behind swift and violent practice discipline progressively.The past ten years, with fibre-optical and the diffuse reflection technology apply in near-infrared analysis, online analysis and the real time supervisory control to procedure of production has come true, has got the considerable economic benefits. Near-infrared analysis method becomes revolutionary character in analytical chemistry technology, apply to broad fields such as petroleum, food, pharmacy industry, chemical industry, clinical medicine, textile industry, environmental science, bioscience and macromolecule etc., especially close in the field of medicine determining the nature and quantitative analysis, near-infrared spectra technology has had a large development over the past few years.The NIR spectra region is generally defined as the wavelength range from 780nm to 2500nm. It is customarily divided into two ranges, short-wave NIR spectroscopy (780nm～1100nm)and long-wave NIR spectra (1100nm～2500nm). This paper study long-wave NIR spectra information in the beginning, subsequently carry on research of short-wave NIR spectra of less data and analysis speed quicker, finally trial study on ultraviolet–visible spectra.Whom near infrared spectrum get O-H, N-H, C-H and S-H etc. flexible or crooked vibration is produced frequency doubling and hybrid vibration spectrum, most compounds are absorbed in this area, its result relates to the internal structure of the molecule, state of the molecule and functional group of molecule directly, we can receive information of determine the nature and quantitative analysis fromthe near infrared spectra. Compared with traditional analytical method, near infrared spectrum technology has a lot of advantages, it is a simple and convenient, fast, save time and high-efficient analytical method. The biggest advantage of near infrared spectrum technology is that the sample is not limited by the appearance, it can be used in the gas sample, the liquid sample and the solid sample. Because the depth of penetrating of near infrared light is heavy, near infrared spectrum technology enable to determine the sample directly with diffuse reflection technology. The solid sample can be any irregular forms such as powder form, paste and silk form etc..But the near infrared spectrum has certain limitation in using, it is the technology of indirect analysis, there are higher requirements in the treatment to spectrum data. The near infrared spectrum is a spectrum of absorbed weakly, overlapping seriously and without characteristic peak, it is unable to analyze by traditional spectrum analyse method. In the process of determine the nature and quantitative analysis obtain accurate reliable analysis results unless adopt certain data processing, so it only combine the chemometrics method can carry on accurate analysis.Among the chemometrics methods, partial least squares (PLS) are commonly used method, calculation simplely, high quality model and a large amount of business software now etc., so it already become a standard and commonly used chemometrics method. But PLS often produces greater prediction error while dealing with the non-linear system, artificial neural network (ANN) becoming focus in the fields of information processing and studying in recent years. The application of artificial neural network in chemistry, biotechnology andpharmacy field increases rapidly because of its enormous advantage on non-linear and anti-interference.In the quantitative analysis of medicines, especially the analysis of Chinese herbal medicine, was usually adopted is that the standard method of the pharmacopoeia mainly adopt the high performance liquid chromatography and gas phase chromatography (HPLC and GC). But these methods must dissolve, separate the sample to his effective composition, and then determine the content of the effective composition, time-consuming, strenuous. This paper successively used partial least squares method and artificial neural network method combine with near infrared spectrum information of fructus cnidii extracts and carry on predicting quantitative of the effective active composition among them, the result indicates this method is feasible, get comparatively good effect and realize rapid and non-destructive analysis of the active composition in fructus cnidii extracts.This paper has been optimized and discussed the choosing of spectrum range, the selection of the score and the choice of the best principal components in partial least squares models, furthermore, the optimal PLS models were obtained.In this paper, five main parameters in ANN models, namely, input neurons, hidden neurons, momentum, learning coefficient and number of iterations, were optimized and discussed; furthermore, the optimal ANN models were obtained. The present criterion of optimization is to make the error of the training set the smallest; however, it is very easy to choose an overfitting model, namely, the error of test set is not smallest. In order to avoid establishingoverfitting models, degree of approximation, was cited. The citation of degree of approximation avoids bringing overfitting phenomenon, training set and predicting set get optimization at last.This paper has adopted the pretreated technology of near infrared spectrum models. Besides including the information related to the fact that the sample makes up in the primitive spectrum that the instrument gathers, include the noise signal that produces from the factors of various fields at the same time. The deflection or drift phenomenon has often appeared in the near infrared spectrum, it causes greater interference to spectrum analyzed. It will influence the quality of models and the predict result accuracy of unknown sample unless deal with. This paper indagate and discuss on original spectrum and pretreatment spectrum separately, the result shows:The pretreatment to the original spectrum is very essential. This paper has set up best models of different spectra region by partial least squares and artificial neural network separately. The result indicates: set up models fast and choose the wave area of the spectrum easy by partial least squares, receive quantitative information feasible, but the precision of the whole result still remains to further improve; set up models need some time by artificial neural network, but after the model be set up the predict unknown sample very fast and there is better accuracy.Through the study on effective active compositions in fructus cnidii extracts indicates: combine partial least squares and artificial neural network with near infrared spectrum can predict the content of the active composition accurately and quantitatively. Partial least squares and artificial neural network must have been expansiveforeground and applied values in traditional Chinese medicine is analyzed.