Development And Application Of Near Infrared Spectroscopy For Determination In Food And Feedstuff

This paper presented the advantages of applications of near infrared spectroscopy (NIRS) with chemometrics methods used in rapid determination and pattern recognition of Food and Feedstuff. The work showed in this paper can also been used in rapid determination in food and feedstuff component and its quality controlling.There are four chapters in this paper. The research work has been emphasized on chemometrics and NIRS used in determination of contents of Food and Feedstuff, differentiating pasteurized milk and reconstituted milk.Chapter One reviewed the development of NIRS and several kinds of chemometric methods commonly used in NIRS, such as pretreatment of spectra data, quantitative/qualitative analysis. It summarized the development of NIRS used in Food and Feedstuff in the recent years both here and abroad, and showed a summary for this paper.Chapter Two studied some chemometrics methods that giving prediction for components of food and feedstuff. First, spectra of 70 feedstuff samples were collected. With the spectra, a PLS-BP prediction model for determination of moisture, ash content, protein and phosphorus in feedstuff powder was established while the other for aspartic acid (Asp), glutamic acid (Glu), serine(Ser), histidine (His). With 50 spectra of potato samples, the PLS-BP prediction model for determination of contents of fibre, starch and protein was established. Three PLS-GRNN prediction model were established. One for measuring chlorine, fibre, fat in feedstuff powder; another for contents of total sugar and acid in Nanfeng Orange; the third one for contents of fibre, starch and protein in potato. A PLS-Elman model for phenylalanine(Phe), lysine(Lys), tyrosine(Tyr), cystine(Cys) in feedstuff powder and one for contents of carotene in tomato was established with data of spectra. The original spectra data were compressed by PLS before put into networks that the models' training iteration times were greatly reduced with good veracity and efficiency.In Chapter Three the pattern recognition of differentiating pasteurized milk and reconstituted milk by Map Networks-near infrared spectroscopy model was established with good veracity and recurrence. In Chapter Four the comprehensive summary and precis for our research was mainly noticed.