Developmen Of Grain Near Infrared Instrument

Our country is a Great Power of grain. The accurate , fast test of the grain's quality is extremely important to the production and market transaction . Now in our country manygrain texture is assessed by sensory , however, grain texture is tending to equipment texture.This is a tendency of grain analysis.Then development of the analysis technique of NIR is the fastest and the most attractive spectrometry technology since the 90's. because of the uneaqualed convenient and universality,it more and more arouses the domestic and foreign experts'attention .It's known as analysis"giant"in the fields of analytical chemistry .It can be said that its appearance brings a revolution of analysis technique again . At the same time , it plays an important promoting role for the development of industry fields, such as the petrochemical industry, agriculture, food, biochemical, medical, papermaking, environmental protection and so on .Recent years NIR detection technique has already been generally accepted as a standard examination method.The study of this article is from the program---"NIR detection technique of the quality of agriculture products and the development of instrument"--- the development plan of JiLin province .The purpose is to develop a grain NIR detection instrument. It can detect multi composition of the grain at the same time .The mainly study job is as following : 1 On the base of analyzing NIR detection technique and the instruments'situation of domestic and foreign , and according to the principle of diffuse reflection spectroscope analysis ,we developed a NIR detection instrument. We determined that filter is the instruments'spectroscopic device, and the optical source is halogen lamp, and the sample pool is a rotary. We choose nine narrow-band filters which are sensitive to the component of grain. Their center wavelength are 2175.9nm, 1710.8nm, 1676.5nm, 1759.6nm, 2090.4nm, 2134.4nm, 1811nm, 2050nm, 2237.8nm.2 The design of grain NIR detection machine's structure and route . We have introduced the machine's whole structure and main components . Light will loss when their through the elements ,so we reduce the use of optical element in the design of optical path ,and simplize the structure of optical path; The angle of the halogen lamp's frame is adjustable to reduce the errors in the machine manufacture ,and this can insure that the light can irradiate the samples vertically; setting 45 degrees of the detector and the incident 4 light ,it can eliminate the influences of mirror reflection . We use the modular design , mainly including one-chip computer module , step-motor driving module, location control module, temperature collection circuit module, spectra signal regulated circuit module ,serial port communication module etc . Data acquisition doesn't use the traditional A/D conversion mode ,but directly collectes the data by using the PCI-6070E data acquisition card which is produced by the USA NI company .This method can directly observe the spectrum signal and it is easy to adjust for hardware .3 The exploitation of grain NIR detection machine's working software .Working software includes program of lower computer(one-chip computer) and program of upper computer(PC).The program of lower computer's exploitation adopts modular design and assemble language compiled to complete the design of program AT89C51.The function of temperature collection of one-chip computer1 is to monitor the temperature of detector's working environment .It is divided into main program module ,temperature collection module and serial port break communication module; The function of control instrument operation of one-chip computer2 is to control the NIR detector's ordered rotation, that is to complete controlling three step-motors. We found that the rotational speed of chop light tablet and sample pool influences the detector-collection's correctness in the instrument's adjustment .At last, we determined that the most suitable rotational speed of step-motor1 and step-motor2 is 30r per minute .At the same time ,in order to ensure the accurate location of filter, prevent the inaccurate location because of the filter disk's rotator inertia moment ,the filter disk's rotational speed should not too fast. But in order to ensure work efficiency, the filter disk's rotational speed should not too slow. According to the experiment ,the best speed is 10r per minute. Upper motor program adopts graphical programming—LabVIEW to exploitation .it is used to complete the function of acquisition , storing, analysis, indicating etc .It becomes a complete detection system together with lower motor. The program adopts sequential structure, divided into three frames, and they are data acquisition and primary treatment, data analysis and absorbency and content calculation.4 Building the model of quantitative analysis and micro-collection library of Grain NIR detector .On the base of chemometrics methods and computer technology, by building multiple linear regression analysis model and investigating the different wavelength compositional modeling ,we choose the best wavelengths from them and developed BP Neural Network model for the best wavelength composition. NIR is on the base of reference analysis method. Its precision and accuracy are rely on the precision and accuracy of reference analysis method, so the preparation samples could not be used to build model directly .We measure the content of protein of the 26 samples by using the kjeldahl determination.20 of them are calibration sets, and the others are test set .Scanning 26 sample sets, we can get absorption spectrum on this instrument. Then we make multiple linear regression and build BT neural network modeling to the samples by using DPS. By building multiple linear regression model of the different wavelength combination, after analyzing, we found that the quantitative analysis model to the calibration sets'predictor and chemical ratio is increasing in whole as the increasing wavelength of bibliography modeling ,but the tendency of increasing is decreasing . The correlation coefficient of using 7 wavelength prediction and 8 wavelength predictor and chemical ratio are 0.886 and 0.887, and the prediction results are almost the same ; In the prediction of test set ,we found that the model of 4 wavelength is the best. This model's predictor and chemical ratio's correlation coefficient is 0.707.Using BP neural network modeling to the optimum wavelength, the correlation coefficient of test sets'predictor and chemical ratio is 0.996. The prediction effect of BP neural network modeling is better than multiple linear regression modeling .5 Grain NIR detector's experimental study .Testing the instrument's system performance from noise variance, repeatability and accuracy, we found that if noise variance is low ,and repeatability is poor ,and adopts multiplicating loading samples, then take the average of sample contents can promote the accuracy of prediction effect .Above all, this article combines NIR analysis technique with virtual instrument technology to design a grain NIR detector. It can realize the detection of grain's multi composition at the same time . This detector'system performance is stable, and it is convenient to adjust, extension and maintain. Man-machine interface is friendly . And it also adds the system's agility .It can analyze and deal the testing data directly and real-timely.