偏最小二乘-反向传播-近红外光谱法同时测定饲料中4种氨基酸

偏最小二乘与人工神经网络联用对70个饲料样品建立起天门冬氨酸(Asp)、谷氨酸(Glu)、丝氨酸(Ser)和组氨酸(His)4种氨基酸含量的预测校正模型,以样品平行扫描光谱验证校正模型预测的准确性和重现性。用偏最小二乘法将原始数据压缩为主成分,采用单隐层的反向传播网络建模。取前3个主成分的12个数据输入网络,以Kolmogorov定理为依据,经过实验确定中间层的神经元个数为25,初始训练迭代次数为1000。偏最小二乘-反向传播网络模型对样品4个组分含量的预测决定系数(R2)分别为:0.981、0.997、0.979、0.946;样品平行扫描光谱预测值的标准偏差分别为:0.020、0.029、0.017、0.023。本研究为近红外快速检测在组分含量较低的样品实现多组分同时测定提供了思路。

Determination of Four Amino Acid in Feedstuff Powder Using Near Infrared Spectroscopy and Partial Least Square-Back-Propagation Network Model

Partial least squares (PLS) and artificial neural networks (ANN) prediction model for the determination of aspartic acid(Asp), glutamic acid(Glu), serine(Ser), histidine(His) in feedstuff powder had been established with good veracity and recurrence. 12 peak value data from 3 principal components straight ahead compressed from original data by PLS were taken as inputs of Back-Propagation Network (BP) while 4 predictive targets as outputs, according to Kolmogorov theorem and experiment, 25 nerve cells were taken as hidden nodes. Its training iteration times supposed as 10 000. Predictive correlation coefficient by the model are 0.981, 0.997, 0.979 0.946 while the standard. deviation of an unknown sample scanned parallelly are 0.020, 0.029, 0.017, 0.023. The result shows an idea for multi-component quantitative analysis of other samples with lower contents.