摘要
饲料经过酸-硅胶沉淀除去基质,测定所得上清液的紫外吸收光谱。采用偏最小二乘法( PLS)、人工神经网络( ANN)、支持向量回归法( SVR)三种化学计量学方法建立紫外吸收光谱对三聚氰胺浓度的预测模型。结果:PLS模型的R2为0.9926-0.9940,均方根差为0.2346-0.2612;ANN模型的R2为0.9999,均方根差为0.0265-0.0408;SVR模型的R2为0.9997-0.9999,均方根差为0.0010-0.0024。 SVR模型的预测效果最好。研究表明,紫外吸收光谱-化学计量学建模用于快速、准确测定饲料中三聚氰胺是可行的,且设备要求低、操作简单,有望推广使用。
The precipitation processing of feed was carried out by acid-silica gel to remove those matrixes,and the obtained superna-tants were measured by UV spectroscopy. Three chemometric methods including partial least squares( PLS) ,artificial neural network ( ANN) and support vector regression( SVR) were used to establish models. As a result,the determination coefficients( R2 ) of PLS is 0. 9926-0. 9940 with RMSE is 0. 2346-0. 2612,the determination coefficients( R2 ) of ANN is 0. 9999 with RMSE is 0. 0265-0. 0408,the determination coefficients(R2)of SVR is 0. 9997-0. 9999 with RMSE is 0. 0010-0. 0024. Compared with PLS model and ANN model,the SVR model had best prediction performance. The results showed that UV-Vis spectroscopy coupled with SVR can be applied for rapid and accurate determination of melamine in feed. It is simple and low equipment requirements,is expected to be widely used.
出处
《化学研究与应用》
CAS
CSCD
北大核心
2014年第9期1422-1427,共6页
Chemical Research and Application