BP神经网络法结合红外光谱快速测定在用润滑油胺类抗氧剂含量  

Rapid determination of amino antioxidant content in in-service lubricating oil based on BP-ANN combined with FTIR

作　　者：王菊香[1] 邢志娜[1] 李伟[1] 刘洁[1] 

机构地区：[1]海军航空工程学院飞行器工程系,山东烟台264001

出　　处：《计算机与应用化学》2016年第2期197-199,共3页Computers and Applied Chemistry

摘　　要：润滑油的润滑性能与抗氧剂的含量密切相关,快速准确地检测在用润滑油的抗氧剂损耗程度是非常重要的。本文研究傅里叶红外光谱结合BP神经网络快速测定合成润滑油中胺类抗氧剂的方法。对神经网络参数和波长范围进行选择,当隐含层神经元个数为13,学习速率、动量因子均为0.2,选择光谱波长范围(3420~3330)cm^(-1)和(1650~1510)cm^(-1)时,获得最佳的抗氧剂定量分析模型。模型的校正偏差SEC为0.052,相关系数R^2达到0.986,预测集的标准偏差SEP为0.059,与校正集结果接近。研究结果表明,BP神经网络结合红外光谱可实现润滑油抗氧剂的快速、准确分析。Lubricating property of synthetic lubricating oil is closely related to the content of antioxidant, and it is important to test the loss of antioxidant in the in-service oil accurately and rapid. A method of Fourier transfer infrared （FTIR） spectra combined with BP-artificial neural network （BP-ANN） was studied to rapid measure the content of amino antioxidant in the synthetic lubricating oil. BP-ANN parameters and wavelength range were chosen. The best analysis model of antioxidant content was built with hidden layer nodes of 13 and learning velocity of 0.2 and appending momentum factors of 0.2 and wavelength range of （3420-3330） cm^-1 and （1650-1510） cm^-1. Calibration error （SEC） of the model reduced to 0.052 and the correlation coefficient reached to 0.986. Prediction error （SEP） is 0.059 which is close to SEC. The study indicates that the method of FTIR combined with BP-ANN can measured antioxidant content of n-service oil quickly and accurately,

关 键 词：傅里叶红外光谱 BP-神经网络 合成润滑油 胺类抗氧剂 

分 类 号：O675.33[理学—化学]