1～21日龄黄羽肉鸡棉籽粕傅里叶近红外及化学成分净能预测模型研究

Prediction Models for Net Energy Value of Cottonseed Meal for Yellow-feathered Broilers Aged from 1 to 21 Days Using Fourier Near Infrared Spectroscopy and Chemical Composition

本试验在用比较屠宰法实测25个棉籽粕样品净能(NE)值的基础上,旨在研究用傅里叶近红外(NIRS)和化学成分2种方法建立的NE预测模型的可行性,并比较2种预测模型的预测效果。1)棉籽粕NE值的测定采用维持NE(NEm)+沉积NE(NEp)的方法。其中NEm用回归法测定,设自由采食及限饲20%、40%、60%和80%5个采食梯度,NEp采用套算法测定;每个梯度和棉籽粕样品均设6个重复,每个重复2只鸡。试验动物为382只平均体重为(62.20±0.64)g的7日龄末空腹康达尔黄羽肉公鸡,试验期为7 d。2)分别建立自然状态和扩大水分背景的NIRS预测模型M1和M2。3)将25个棉籽粕样品的表观代谢能(AME)、粗蛋白质、粗脂肪、粗纤维、中性洗涤纤维、酸性洗涤纤维和灰分7种成分值与NE值进行一元和多元线性回归分析。结果如下:1)M1、M2的校正决定系数(R2cal)分别为0.999、0.985,校正标准差(RMSEE)分别为0.033、0.084 MJ/kg DM,交叉验证决定系数(R2cv)分别为0.966、0.967,交叉验证标准差(RMSECV)分别为0.120、0.117 MJ/kg DM,预测决定系数(R2val)分别为0.843、0.957,预测标准差(RMSEP)分别为0.260、0.136 MJ/kg DM,2个模型预测值与实测值配对t检验结果均不显著(P>0.05)。2)用化学成分结合AME建立的最佳预测方程的R2和RSD分别为0.985和0.093 MJ/kg DM。结果表明:1)应用NIRS和AME结合化学成分均能建立预测效果可靠的棉籽粕NE预测模型;2)NIRS所建M2模型的预测效果与AME结合化学成分所建模型相当。

This trial was to study the feasibility of establishing prediction models for the net energy（NE） values using Fourier near infrared spectroscopy（NIRS） and chemical composition on the basis of 25 cottonseed meal NE values measured by comparative slaughter experiment,and to compare the predictive results of them.1） NE was calculated as NE for maintenance（NEm） plus NE for deposition（NEp）.The NEm was measured by regression method with 5 feeding levels including ad libitum feeding and restricted feeding by 20%,40%,60% and 80%,respectively.NEp was measured by the method of substitution.A total of 382 Kangdaer fasting yellow-feathered broilers at 7 days of age with average body weight of（62.20±0.64） g were randomly allotted into every level of cottonseed meal sample with 6 replicates each and 2 chickens in each replicate.The experiment lasted for 7 days.2） NIRS calibration models（M1 and M2） of NE were established under the natural condition and a larger moisture background,respectively.3） Predictive equations for apparent metabolizable energy（AME）,crude protein（CP）,ether extract（EE）,crude fiber（CF）,neutral detergent fiber（NDF）,acid detergent fiber（ADF）,and ash with NE were derived from the methods of one-dimensional and multivariate linear regressions.The results showed as follows： 1） the R2cal and root mean square error of calibration（RMSEE） of 2 models（M1/M2） were 0.999/0.985 and 0.033/0.084 MJ/kg DM,the R2cv and root mean square error of cross validation（RMSECV） were 0.966/0.967 and 0.120/0.117 MJ/kg DM,the R2val and root mean square error of prediction（RMSEP） were 0.843/0.957 and 0.260/0.136 MJ/kg DM,respectively,and the results of paired-samples t test of NIRS predictive values and determined values were not significantly different（P0.05）.2） The R2 and the RSD of the optimum regression equations from chemical composition combined with AME were 0.985 and 0.093 MJ/kg DM,respectively.These results indicate as follows： 1） the two methods above can both establish NE predictive models of cottonseed meal with reliable results;2） the predictive accuracy of M2 is similar to the optimum equation from chemical composition combined with AME.