Study On Rapid Detection Technology Of Quality Of Zanthoxylum Bungeagum Maxim By Infrared Spectroscopy

In this paper, the intrinsic relation between quality indicators of zanthoxylum bungeanm maxim and infrared spectra is researched by infrared spectroscopy, which provide a new method and tool for detecting quality. The regression equation between characteristics peaks of volatile oil and protein and content value of volatile oil and protein is established based on MLR,PCR and PLS, accessed to the regression model of detection quality. At the same time, the classification models of different source are achieved by stepwise discriminant analysis and BP neural network. It includes three parts:the first part:the infrared spectrogram of zanthoxylum bungeanm maxim is analyzed to set up the identification model for finding a best method to distinguish; the second part, based on the infrared spectroscopy method rapid detected the volatile oil in zanthoxylum bungeanm maxim; the third part:the predictive model is built to quick detect the protein content in zanthoxylum bungeanm maxim.The main experiment results as follows:(1)The characteristics peak of volatile oil:3303cm-1,2926cm-1,1730cm-1,1235cm-1,1019cm-1,916cm-1,770cm-1; the characteristics peak of protein:2926cm-1,1628cm-1,1550cm-1,1400cm-1,1371cm-1,770cm-1。(2)The stepwise discriminate analysis is used to establish five classification models of zanthoxylum bungeanm maxim from different areas. And the accurate rate of calibration sample is 96%, the accurate rate of validation sample is 92%.The result shows that it is feasible to use stepwise discriminate analysis to establish classification models.(3) BP neural network can be used to make predictive model of classification. Based on the tests, the accurate rate is highest, the fitting residual is lowest, the predictive ablity of predictive model is best when the network structure is 11-8-1.The accurate rate of calibrarion sample and validation sample equal to 100%.(4)The identification model by BP neural is better than stepwise discriminate analysis.(5) Multiple Linear Regression,Principal Component Regression,Partial Least Squares method are respectively applied to bulid forecasting model of volatile oil content. By comparison, MLR is suited the forecasting model of volatile oil content in zanthoxylum bungeanm maxim.(6) The forecasting model of protein content is built by Multiple Linear Regression,Principal Component Regression and Partial Least Squares. The result shows that MLR is better than others to predict the protein content in zanthoxylum bungeanm maxim.