摘要
针对航空煤油中总酸值量较小,在近红外定量分析时有用信息易被干扰的问题,采用误差反向传播神经网络(BP-ANN)建立航空煤油总酸值近红外光谱分析模型。根据模型校正集预测偏差最小原则,确定了隐含层神经元个数、学习效率等参数。用建立的网络模型预测了验证集样品总酸值,预测的相关系数R2为0.9778,预测标准偏差(RMSEP)为0.00066。结果表明,该方法适于对航空煤油总酸值进行快速、准确的定量分析,并且与偏最小二乘法(PLS)预测模型进行对比预测精度更高。
Due to the fact that the amount of total acid number of aviation kerosene is small and useful information is likely to be interfered in the near-infrared spectroscopy（NIR） quantitative analysis,a multivariate calibration NIR model called as back propagation-artificial neural network（BP-ANN）was used to predict the total acid number in aviation kerosene.The net parameters such as learning rate and momentum were determined according to the minimum deviation principle.Total acid number of the test set was predicted by using the obtained models,the test set correlation coefficients（R2） was 0.9778,and the root mean square error of prediction（RMSEP） was 0.00066.The prediction accuracy was better than that of partial least squares（PLS） model.The obtained results showed that the method is fast and accurate and suitable for quantitative anlysis of total acid number in aviation kerosene.
出处
《分析科学学报》
CAS
CSCD
北大核心
2011年第6期751-754,共4页
Journal of Analytical Science
