The Determination About Uniformity Mixing And Moisture Content Of Feed Based On Near Infrared Spectroscopic Technology

The mixing uniformity and the moisture content of compound feed are two important indexed in feed production. At present the conventional determination method of the two indexes need to depend on the laboratory conditions, but the prosess is complex indeed, so it is not conducive to promote the efficiency of the feed production. My paper explores a new method of determination so as to detect the above two indexes conveniently, accurately.In the test1,1make three kinds of the most important components of a compound feed, corn, soybean meal and premix, as raw material, using near infrared spectroscopy equipment, feed mixing machine, simulated the process of mixed feed, collected the online data, measured the CV value of uniformity methyl viologen, then esTablelished the reference matrix of uniformity. The sample data pretreated with9point smoothing and MSC by using chemometrics software UN X, then make three kinds of pattern recognition method, respectively, SIMCA, LDA and SVM, the accurate rate of determination of feed mixing state were94.8%,98%and92.7%. In the test2,Ⅰ detected the moisture content of mixed feed by using near infrared spectroscopy. Specifically first I acquired the conventional detection value of moisture content by using hot air dryer, then I compared the effect of modeling with4kinds of pretreatments,at last I established the moisture rate determination of PLS model,the maximum deviation between the predictive value with the conventional detection value is2.1%.In the end he results show that, near infrared spectrum detection technique can be applied to detect the feed uniformity and moistrue rate, the main advantages include accurate measuring result, convenient operation, establele error; online detection; non-destructive to feed samples, environmental protection. The spectral information analysis results show that the suiTablele preconditioning methods can reduce the modeling error, and LDA has the higher accuracy than SIMCA and SVM for estimating the sample information when the information conform to the linear change。...