估算加氢处理润滑油基础油芳香烃含量的神经网络模型

AN ESTIMATION MODEL OF ANN FOR PREDICTING AROMATICS CONTENT IN HYDROTREATED LUBE BASE OIL

建立了用加氢处理润滑油基础油的密度 (2 0℃ )、粘度 (1 0 0℃ )、折光指数 (2 0℃ )和粘度指数估算其芳烃含量的神经网络模型 ,并将预测结果与主成分回归模型所得结果进行了比较。结果表明 ,神经网络模型预测数据的最大误差为 1 .6% ,绝对值平均误差为 0 .79% ;而主成分回归模型预测值的最大误差为 6.0 % ,绝对值平均误差为 2 .1 7%。可见 ,神经网络模型对润滑油基础油芳烃含量的估算精度优于主成分回归模型 。

An ANN (Neural network algorithm) model is developed for predicting aromatics content of hydrotreated lube base oil with density(20 ℃), viscosity(100 ℃), refractive index (20 ℃) and viscosity index For Comparison, a regression model by principle components analysis (PCA) technology is also presented With the same statistical data sets, the results of predicting practice show that the ANN model is better than PCA model for prediction and can be used for estimating the aromatics content of hydrotreated lube base oil