摘要
转子碰摩故障通常为不平衡、不对中以及油膜涡动等故障引发的二次故障,其信号通常具有周期、拟周期和混沌这3种复杂的非线性特征。本文针对油膜涡动下的转子碰摩故障诊断问题,建立了含不平衡、油膜涡动以及碰摩故障耦合动力学模型,利用数值仿真研究了转子系统在油膜涡动下的碰摩故障频谱特征,提取了反映耦合故障的特征信息。为了准确的对碰摩故障进行诊断,通过不断改变系统参数获取了包括各种状态下的耦合故障样本。最后构造了结构自适应神经网络,利用一半样本对神经网络进行训练,再用另一半样本对训练好的神经网络进行测试,识别率达到了94%以上。计算结果充分表明了本文方法对于识别油膜涡动和碰摩耦合故障的有效性。
Rub is usually the second fault which results from unbalance,misalignment and oil whiff and so on. Its signal output usually has three characteristics： periodic ,quasi-periodic, and chaos; and it has complicated nonlinear characteristics. In this paper, we first establish a dynamic model for rotor with coupling faults ,considering non-linear oil whiff force. The model includes three faults： mass unbalance, oil whiff and rubbing. Then we study the frequency domain characteristics of rotor response using numerical integration, and analyze the characteristics information of the coupling faults. In order to diagnose the rotor rubbing faults accurately, enough rubbing samples in various states are needed. Finally, we construct a structural self-adaptive neural network, and use half of the samples to train the net, and other samples to test the net. The recognition rate of over 94 percent demonstrates the effectiveness of this method for recognizing rub faults in oil whirling.
出处
《机械科学与技术》
CSCD
北大核心
2007年第11期1447-1453,共7页
Mechanical Science and Technology for Aerospace Engineering
基金
国家自然科学基金项目(50705042)
航空科学基金项目(2007ZB52022)资助
关键词
转子动力学
故障诊断
油膜涡动
碰摩
神经网络
rotor dynamics
fault diagnosis
oil whirl
rub fault recognition rate
neural network
