某型柴油机润滑油发射光谱数据挖掘研究  

Mining Study of Oil Atomic Emission Spectrum Data of Certain Diesel Engine

作　　者：张乔斌[1] 张春辉[1] 朱爱芳[1] 

机构地区：[1]海军工程大学,武汉430033

出　　处：《计算机与数字工程》2014年第11期2037-2040,共4页Computer & Digital Engineering

摘　　要：油料原子发射光谱仪是目前国内外广泛应用的油液分析技术之一。为了深入挖掘某型柴油机润滑油中磨损元素的浓度与柴油机负荷、气缸间隙和运行时间之间的对应关系,应用神经网络建立了某型六缸柴油机主要磨损元素Fe的浓度仿真模型和预测模型。柴油机设置了7种工况,测量了69个油样。仿真模型中,69个油样仿真值的相对误差均小于15%;预测模型中,19个油样预测值的绝对误差均小于光谱仪精确度值,且84%的油样预测值的相对误差小于15%。预测结果表明:神经网络算法能较好地预测Fe元素浓度。Atomic emission spectroscopy is one of the most widely used techniques for oil analysis in the world now. In order to deeply mine the relation between the concentration of wearing elements of diesel engine and its loads, cylinders＇ clearances and runtime after renewing oil, a simulation model and a prediction model of Fe concentration of a type of six cyl- inder diesel engine are established by applying neural network. The engine set up seven different working conditions and measured concentration of sixty-nine oil samples. The results show that the relative errors of the simulation value of the 69 samples are within less than 15 %. The absolute errors of prediction value of the 19 samples are lower than the acceptable ac- curacy indices and the relative errors of 84% samples are within 15%. It is proved that Fe concentration can be predicted ef- fectively by Neural Network algorithm.

关 键 词：神经网络 光谱分析 柴油机 磨损元素 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]