基于PCA-BP神经网络的设备润滑状态评价应用

Application of PCA-BP Neural Network in Equipment Lubrication State Evaluation

目前,油液监测技术已涉及到多数工业领域。其技术内容除包含润滑剂本身理化的性能指标外,还包含诊断人员对设备磨损、油液污染的综合评价。针对诊断内容对专家经验需求较大,且具有模糊现象的弊端,将BP神经网络模型引入到油液监测诊断评价的过程中。文章首先采用PCA提取累计贡献率大于90%的油液主成分,然后以处理后的数据集合训练BP神经网络,最后采用ROC曲线验证模型效果。结果表明,相比训练集未降维的BP神经网络识别率结果,PCA-BP神经网络的识别结果更优。

At present,oil monitoring technology has been applied to most of industrial fields.Its content not only contains physical and chemical properties of lubricant,but also contains the estimate of diagnostic engineer on equipment wear and polluted condition.The BP neural network model is introduced into the process of oil-liquid monitoring and diagnosis in view of the need for expert experience and the disadvantages of fuzzy phenomena.The paper first discusses that PCA is used to extract the main components of oil fluid with cumulative contribution rate of more than 90%.Then BP neural network is trained with the processed data set.Finally,ROC curve is used to verify the model effect.The results show that the recognition rate of PCA-BP neural network is better than that of unreduced-dimension BP neural network in training set.