近红外光谱定量技术在方便面油份快速测定中的应用

探讨了用近红外漫反射光谱快速无损检测方便面含油率的数据处理方法，采用浸反射光谱、一阶导数光谱和二阶导数光谱，使用了四元线性向前逐步回归和BP人工神经网络的数学方法，对40个校正样本建立了的线性和非线性两种校正模型，用28个预测样要检验了校正模型的预测精度，其中线性校正模型中，采用二阶导数光谱的预测精度最好，预测平均误差为5．741％，预测误差的标准差为1．842；非线性校正模型中，采用一阶导数光谱，隐层单元数为2时，校正模型的预测精度最好，预测平均相对误差为5．149％，预测误差的标准差为1．675结果表明近红外漫反射光谱分析检测方便面的含油率能满足实际生产的要求。

Quantitative Technology of Near Infrared Spectroscopy Applied in Inspecting Oil Content of Instant Noodles

Near-infrared spectroscopy(NIRS),with the characters of quickness,non-destruction,long detection range,high precision and reliable detection data,etc.,has been a newly and popularly quantitative and qualitative analysis method.Whether or not can NIRS be used in practical production,it depends on its predicting precision.This paper discusses the initial diffused reflection spectra,the first and second derivative spectra,which are treated with 4-element step-forward regression and BP artificial neural network.Both linear and non-linear calibration models were developed from 40 calibrating samples,whose precision was detected with 28 predicting samples.The results show that the second derivative spectra have the highest predicting precision with 5.741% of relative error and 1.842 of standard error in the linear model.On the other hand,the first derivative spectra with 2 units in hidden layer have the highest predicting precision with 5.149% of relative error and 1.675 of standard error in the non-linear model.Therefore,it is concluded that NIRS for quantitativly inspecting oil content of instant noodles is feasible.