Using Near-infrared Spectroscopy To Detect The Content Of Protein And Fat In Yoghourt

With the improvement of people's living standard,people have improved their demand on food health function more than food safety and nutrition.Yoghourt is called "long life food".It is not only good-tasted,but also has quite high nutritional and health promoting effect.Nowadays,with the increasing demand of yoghourt, yoghourt has been a mainstream of dairy products research and development. However,in the production process of yoghourt,it has become a problem to be solved,which mastering the transformation of material and the quality and performance of the final products,as it is directly relevant to the safety problem of food.Near infrared spectroscopy analysis(NIRS)is a new technology which can estimate the content of one or multiple chemical constituents of samples rapidly by using the optical properties in NIR spectrum area of organic chemicals.As there is no pretreatment during the analysis process,no consumption or destruction,and no instrument pollution,NIRS can be called a representative of "green analysis technology".Meanwhile,it is a "transient" analysis technology which enjoys a good name of "GIANT METHOD"Though the instrumental analysis has replaced the traditional analysis on dairy products analysis in recent ten years,it is still hard to break away the troditional mode of interventional devastating analysis and chemical analysis.The velocity of chemical and other instrumental analysis are all restricted by pretreatment of sample and complexity of analysis process.Besides,most of modernization instruments are complecated,whose operating rules and regulations are involved, and the using conditions are very rigorous.So the traditional experiment can only be carried out in a special laboratory.In order to solve the problem of slow inspection velocity,low efficiency and high cost of dairy products,both domestic and overseas scientists develop the research one rapid detection of the composition of dairy products actively.Protein and fat are tow main nutrient components in yoghourt,and also important indexes of yoghourt quality evaluation.Using Near-infrared Spectroscopy to detect the content of protein and fat in yoghourt is still infrequent in China.In this article,the author uses near-infrared reflectance spectroscopy to detect the content of protein and fat in yoghourt,for searching after the fast measure technique of yoghourt,which serves for food safety problem.There was 40 yoghourt samples of many different brands such as Yili, Mengniu,Guangming etc.were collected for this study.The contents of protein and fat,which the two main nutrition ingredients of yoghout were selected as detecting indexes.The chemidal values,which were determined by the conventional chemical analyze methods of Bradford and Rose-Gettlieb separately,were taken as the accomplished reference data for quantity prediction matrix.The near-infrared spectrum of yoghourt samples were scanned by WQF-400N Fourier transforms near-infrared spectrometer.The author questeded the best parameter enactments of spetctropmeter by means of choosing and revising scanning parameter of spetctropmeter,samples status,and process methods of samples.The near-infrared spectrum of youghourt samples were scanned accordingly.The samples' spectrum scanning parameters were established as fallows.The scope of scanning wave number:10 000cm^-1-3 500 cm^-1.The resolution:4 cm^-1.The signal gain:8.The background and the sample scanning times:32.The detector:PbS.The rotational sample receptacle was revolved to reduce the sample non-uniform influence.The effects to NIRS that brings by the chemical and physical properties of yoghourt were considered in this article.Yoghourt is a typical pseudoplastic fluid. The bulk of fat sphere and protein gel in yoghourt,as well as the texture and structure of yoghourt,all effect NIRS obviously.In addition,as the yoghourt is classified to the "milking system",the yoghourt granule in samples brings in Scatter Effect to the ray,that leads to the uncertainty of both the alignment and the extent of the ray.In view of the considerations above,a scanning technique which combined "transmission" and "reflection" was introduced.An apprutenance of improved receotacke which was designed by Jiangsu University voluntarily was adopted in the experiment.What the advantages of his technique were both of the extent and the signal of the ray was doubled,and the open design of apprutenance solved the loading and purification problem at the same time.The original NIR spectrum was optimized.With comparative the accuracy of PLS models,the optimal range of wavenumber was selected;the best spectrum pre-processing methods were chose,and the Outlier was eliminated,thus the prediction model's forecast ability is also effectively improved.In this research,the optimal range of wavenumber is 3 600-7 500cm^-1,and there were 1 011 wavelength data in all which was from 3 602cm^-1to 7 498cm^-1in this spectra for modeling.The best spectrum pre-processing methods which were determined by the comparison with the modeling results were as follows.Protein parameter: five-point second-order differentiation and eleven-point MAF smooth processing (differentiation after smoothing);Fat parameter:eleven-point first-order differentiation,fifteenth-point MAF smooth processing,normalization and MSC.Tow methods,of which one was partial least squares(PLS)regression and the other was PLS+ANN that combined with PLS and artificial neural networks (ANN),were both utilized on establishing quantitative models.PLS models and PLS+ANN models were building with using NIRSA2.2 which was a chemometrics data processing system software developed by NIR study group of Jiangsu University.The parameters of protein and fat PLS models were as follows.The correlation coefficients were 0.925 1 and 0.911 4.The Variances Estimations were 0.038 1 and 0.074 8.The forecast parameters of the test suite sample were:mean absolute error:0.034 5 and 0.061 5;average relative error:2.026%and 4.087%. The parameters of protein and fat PLS+ANN models were as follows.The correlation coefficients were 0.926 3 and 0.920 2.The Variances Estimations were 0.036 2 and 0.109 1.The forecast parameters of the test suite sample were:mean absolute error:0.033 5 and 0.003 1;average relative error:1.936%and 3.338%. The PLS models and PLS+ANN models that were built with two different methods in this research were all high precision.The protein parameters PLS+ANN models and PLS models comparison results were as follows.The correlation coefficients:0.001 2 higher.The mean absolute error:0.001 0 lower.The average relative error:0.09%lower.The correlation coefficients:0.001 9 lower.The fat parameters PLS+ANN models and PLS models comparison results were as follows. The correlation coefficients:0.008 8 higher.The mean absolute error:0.058 4 lower. The average relative error:0.649%lower.But the correlation coefficient is 0.034 higher.Generally,the prediction results of PLS+ANN method were better than those of PLS method slightly,that indicated it was likely to exist some stated nonlinear relationship between the NIR spectrum and the content of ingredients in samples.