【摘要】：對(duì)乳粉中蛋白質(zhì)和脂肪近紅外定量模型的優(yōu)化進(jìn)行了研究。結(jié)果表明:利用全波段光譜建立的模型,蛋白質(zhì)和脂肪的模型預(yù)測(cè)評(píng)價(jià)參數(shù)分別為:校正集預(yù)測(cè)偏差(RMSECV)為2.837、2.984,內(nèi)部交互驗(yàn)證標(biāo)準(zhǔn)偏差(RMSEP)為3.406、3.751,交互檢驗(yàn)相對(duì)偏差(RPD)為2.6、2.5。經(jīng)過波長(zhǎng)優(yōu)選后,蛋白質(zhì)和脂肪優(yōu)選的波數(shù)范圍分別為(9403.5~7498.1)cm-1、(9403.5~6098)cm-1,蛋白質(zhì)和脂肪的模型預(yù)測(cè)評(píng)價(jià)參數(shù)有一定的提高,分別為RMSECV 1.963、2.317,RMSEP 2.396、3.035,RPD 2.9、2.6。通過正交試驗(yàn)對(duì)定標(biāo)方法和光譜預(yù)處理方法進(jìn)行了優(yōu)化:蛋白質(zhì)的最佳參數(shù)組合為:定標(biāo)方法為改進(jìn)偏最小二乘法,多元離散校正,一階導(dǎo)數(shù)處理,導(dǎo)數(shù)處理間隔點(diǎn)為8,平滑處理間隔點(diǎn)數(shù)為4,二次平滑處理間隔點(diǎn)數(shù)為1,目標(biāo)函數(shù)達(dá)到98.25%;脂肪的最佳參數(shù)組合為:定標(biāo)方法為偏最小二乘法,散射校正為多元離散校正,導(dǎo)數(shù)處理為二階,導(dǎo)數(shù)處理間隔點(diǎn)為1,平滑處理間隔點(diǎn)數(shù)為4,二次平滑處理間隔點(diǎn)數(shù)為4,目標(biāo)函數(shù)達(dá)到95.26%。
[Abstract]:The optimization of the quantitative model of protein and fat in milk powder was studied. The results showed that the model, protein and fat model of the model, protein and fat were determined as follows: the corrected set prediction deviation (RMSECV) is 2.837, 2.984, the internal interaction verification standard deviation (RMSEP) is 3.406, 3.751, and the relative deviation (RPD) of the interaction test is 2.6, 2.5. After the optimized wavelength, the wavenumber range of protein and fat is (9403.5-749.1) cm-1, (9403.5-6098) cm-1, and the model of protein and fat has a certain improvement, which are RMSECV 1.963, 2.317, RMSEP 2.396, 3.035 and RPD 2.9, 2.6, respectively. The calibration method and the spectrum pretreatment method are optimized by the orthogonal test: the optimal parameters of the protein are as follows: the calibration method is to improve the partial least square method, the multivariate discrete correction, the first derivative processing, the derivative processing interval point is 8, the number of the smoothing processing intervals is 4, the number of the secondary smoothing processing intervals is 1, and the target function reaches 98.25 percent; the optimal parameter combination of the fat is that the scaling method is a partial least square method, the scattering correction is a multivariate discrete correction, the derivative processing is a second order, the derivative processing interval point is 1, the number of the smoothing processing intervals is 4, The number of secondary smoothing processing intervals is 4, and the target function reaches 95.26%.
【作者單位】： 江蘇食品藥品職業(yè)技術(shù)學(xué)院;
【基金】：淮安市應(yīng)用研究與科技攻關(guān)計(jì)劃項(xiàng)目(HAS2014024-4)
【分類號(hào)】：O657.33;TS252.7
，

本文編號(hào)：2502387