利用近红外(NIR)光谱技术结合偏小二乘(PLS)的方法,通过分别选择不同波长、不同光谱预处理方法,建立并优化棉籽饼粉水解液中氨基氮含量校正模型。波长选择为1 300~1 800 nm,采用光谱预处理阶导数+减条直线,得到校正模型的交叉验证均方根(RMSECV)为0.457 g/L,决定系数(R2)为0.927 2,剩余预测偏差(RPD)为3.71。并对校正模型进行外部检验,预测含量与实际含量进行对比,决定系数为0.944 5,平均相对误差为5.05%。结果证明预测模型能够快速、准确地对棉籽饼粉水解液氨基氮含量进行预测和监控,为建立棉籽饼粉水解液评价系统奠定定基础。

In this study,the model for determination of amino nitrogen concentration in cottonseed meal hydrolysate was developed by near infrared spectroscopy.The near infrared measurements of samples were analyzed by partial least-squares( PLS) regression with selecting different wavelength and spectral pre-processing methods.The parameters of optimized model as followed: wavelength was 1300-1800nm,spectral pre-processing methods was first derivative + straight line subtraction.The root mean square error of cross validation was 0.457 g/L; the determination coefficients were 0.927 2; the residual predictive deviation was 3.71.Compared with predict value and actual measured value of external validation set,determination coefficients was 0.944 5 and the average relative errors were 5.05%.These results showed that prediction model could predict amino nitrogen concentration in cottonseed meal hydrolysate accurately and quickly,and it would provide theoretical basis for the control the quality of cottonseed meal hydrolysate.