摘要
基于滑油效能数据,通过构建容错率高、泛化能力强的BP神经网络模型诊断柴油机运转故障,实现对柴油机零件磨损状况的监测。经过比较光谱分析、自动磨粒分析等油液分析技术的优缺点,选择光谱分析法监测滑油效能数据;构建神经网络,经测试比较确定隐含层节点数及构建函数的最佳选择;利用已知的滑油效能数据训练构建的神经网络,经检测其所得误差在允许范围内,成功验证基于滑油效能数据、BP 神经网络,可以较精确的监测柴油机零件磨损状况,并通过神经网络结合 Simulink 和数据库,构建柴油机零件磨损状况模拟仿真系统,实现了对滑油效能数据的模拟。
Based on the lubricating oil efficiency data,this paper build a BP neural network model with high fault tolerance and strong generalization ability to diagnose the running status of diesel engine and realize the diagnosis of the wear degree of its parts.After comparing and analyzing the advantages and disadvantages of oil analysis techniques such as spectral analysis and automatic abrasive particle analysis,the spectral analysis method is selected to monitor the performance data of lubricating oil.The neural network is constructed,and the best number of hidden layer nodes and construction function are determined by testing and comparison.The measured oil performance data is used to train the neural network,the error obtained by the detection is within the allowable range.It verifies that based on the lubricating oil performance data,the BP neural network can more accurately monitor the degree of wear of diesel engine parts.Through the neural network combined with Simulink and the database,a simulation system for the wear degree of the diesel engine parts is constructed,and the simulation of the lubricating oil efficiency data is realized.
作者
胡浩帆
HU Haofan(COSCO Shipping Specialized Carriers Co.,Ltd.,Guangzhou 510623)
出处
《广东造船》
2022年第3期82-85,共4页
Guangdong shipbuilding
关键词
柴油机
滑油分析
磨损
神经网络
故障诊断
Diesel engine
Lubricating oil analysis
Wear
Neural network
Fault diagnosis
