【摘要】：建立良好的蛋白飼料近紅外光譜定量分析模型及實現(xiàn)在不同儀器間的模型共享,能極大提高模型的利用效率,滿足飼料行業(yè)快速發(fā)展的需要。針對蛋白飼料原料粗蛋白含量近紅外分析模型適用性問題,首次采用光譜差值轉(zhuǎn)移、直接校正和分段直接校正法進行了三臺不同類型的近紅外光譜儀之間的模型轉(zhuǎn)移研究。實驗樣品為四種蛋白飼料原料:玉米蛋白粉、菜粕、酒糟和魚粉。實驗儀器包括MATRIX-Ⅰ傅里葉變換型近紅外光譜儀(主儀器),Spectrum 400傅里葉變換型近紅外光譜儀(從儀器1)和SupNIR-2750光柵掃描型近紅外光譜儀(從儀器2)。研究表明,同一樣品體系在主儀器和從儀器2上所得光譜數(shù)據(jù)的差異性相對較小,且均與從儀器1所得光譜數(shù)據(jù)的差異性相對較大。除分段直接校正法對玉米蛋白粉從儀器2的預測結(jié)果無促進作用之外,其他模型的預測均方根誤差和系統(tǒng)偏差均明顯低于轉(zhuǎn)移前。玉米蛋白粉、菜粕和酒糟樣品采用三種方法轉(zhuǎn)移后的模型預測相對分析誤差(RPD)均大于3.0,預測效果良好。魚粉樣品模型轉(zhuǎn)移后的預測RPD均大于2.5,預測效果較好。三種方法對于蛋白飼料原料不同儀器間的光譜差異進行了有效校正。該研究結(jié)果對于蛋白飼料品質(zhì)近紅外快速分析模型的廣泛應用具有重要意義。
[Abstract]:The establishment of a good quantitative analysis model of protein feed by near infrared spectroscopy and the sharing of the model among different instruments can greatly improve the utilization efficiency of the model and meet the needs of the rapid development of feed industry. Aiming at the applicability of near infrared analysis model of crude protein content of protein feed, the model transfer between three different types of near infrared spectrometers was studied by spectral difference transfer, direct correction and subsection direct correction for the first time. The samples were four kinds of protein feedstuff: corn protein meal, rapeseed meal, distiller's grains and fish meal. The experimental instruments include MATRIX- I Fourier transform near infrared spectrometer (main instrument), Spectrum 400 Fourier transform near infrared spectrometer from instrument 1) and SupNIR-2750 grating scanning near infrared spectrometer from instrument 2. The results show that the difference of spectral data obtained from the main instrument and from instrument 2 of the same sample system is relatively small, and the difference between the spectral data obtained from the same sample system and that from instrument 1 is relatively large. With the exception of direct correction method, the RMS error and systematic deviation of other models were significantly lower than those before transfer, except that the prediction results of corn protein powder from instrument 2 were not promoted. The relative analysis error (RPD) of the models after the transfer of corn protein meal, rapeseed meal and distiller's grains samples was higher than 3.0, and the prediction effect was good. The predicted RPD of fish meal model after transfer was more than 2.5, and the prediction effect was better. Three methods were used to correct the spectral differences among different instruments for protein feed materials. The results of this study are of great significance for the wide application of NIR fast analysis model of protein feed quality.
【作者單位】： 中國農(nóng)業(yè)大學工學院;
【基金】：國家重大科學儀器設備開發(fā)專項課題(2014YQ47037705) 國家科技支撐計劃資助(2014BAD08B11-2) 國家自然科學基金項目(31201827)資助
【分類號】：S816;O657.33
，

本文編號：2337041