人工神经网络在近红外分析方法中的应用及深色油品的分析——人工神经网络-近红外分析方法快速测定原油馏程

采用近红外分析方法测定原油馏程性质，比较了人工神经网络，主成分回归和偏最小二乘法的校正结果，认为对于组成变化比较复杂的原油，采用人工神经网络方法能够得到较好的建模效果，对验证集的测定结果表明，近红外分析方法与所用参考分析方法（气相色谱模拟蒸馏）测定的原油馏程结果之间没有显著性差别，二者测定结果的差值在参考方法ASTM D5307的再现性范围内，说明以人工神经网络为校正方法的近红外分析方法可以用于原油馏程性质的快速测定，其单次测定只需大约10min.

APPLICATION OF ARTIFICIAL NEURAL NETWORK IN NEAR-INFRARED ANALYSIS AND DARK-COLORED OIL ANALYSIS

The determination of distillation characterization for crude oils by near-infrared spectroscopy (NIR) was studied. The gas chromatography simulation distillation method was used as the reference method. Various calibration methods such as artificial neural network (ANN), principal component regression and partial least square were compared. It was found that ANN was the best calibration method for the crude oils with more complicate compo-nents. The paired test showed that the result of NIR method and that of gas chromatography method were coincident, the difference between the property tested by the former and that by the latter was fallen in the reproducibility limit of ASTM D5307. Determining the boiling range distribution of crude oil by NIR technique was proved to be practical and fast, each analysis only needs about 10min.