燃气轮机润滑系统磨损趋势预测  

Wear Trend Forecasting of Gas Turbine Lubricating System

作　　者：孙佳斯 佟文伟 郎宏 刘宇佳 何山 

机构地区：[1]中国航发沈阳发动机研究所,辽宁沈阳110015

出　　处：《润滑与密封》2017年第6期113-118,共6页Lubrication Engineering

摘　　要：为预测燃气轮机润滑系统试验阶段潜在的磨损故障,应用神经网络及遗传算法(GA)对某型燃气轮机长时试验用润滑油中典型元素的光谱监控数据趋势进行预测。通过改变光谱分析数据归一化的范围及调整学习率的自适应性对标准误差反向传播(BP)神经网络进行改进,并利用遗传算法对改进的BP神经网络的权值和阈值进行优化,建立适合某型燃气轮机润滑系统试验阶段磨损趋势预测的模型。结果表明:所建模型具有很高的预测精度和很强的实用性,能有效地提高磨损故障的预测成功率。The oil spectral analysis data of the classic metal elements contained in the used oil during the long-time test was forecasted by genetic algorithm （GA） and neural network to predict the wear trend of the gas turbine lubricating sys- tem.The standard BP network was improved by changing the normalization range of the oil spectral analysis data and adjus- ting the adaptivity of learning rate,and the weights and thresholds of the improved BP network was optimized by GA.The wear trend forecasting model of a gas turbine lubricating system during its testing period was built, and the potential wear trend was predicted by the well-trained model.The result shows the proposed model is practical and promising in the wear trend prediction of gas turbine lubricating system during the testing period ,which can improve the forecasting accuracy of wear fault.

关 键 词：燃气轮机 润滑系统 磨损趋势预测 BP神经网络 遗传算法 

分 类 号：V263.6[航空宇航科学与技术—航空宇航制造工程] TH117.1[机械工程—机械设计及理论]