摘要
针对齿轮箱早期故障特征不明显,提出一种基于时序和径向基(Radial Basis Function,RBF)神经网络相结合的诊断方法。通过对齿轮箱正常和故障运行状态的振动信号进行时序分析,提取时序模型的自回归系数作为状态特征,并将其组成特征向量输入RBF神经网络分类器进行网络训练,从而实现了对齿轮正常、裂纹、断齿和局部点蚀的状态识别与诊断。结果表明,基于时序—RBF神经网络结合的方法对于早期或多故障分类是可行的。
Due to incipient fault features of gear being not obvious,a method based on time series analysis and radial basis function neural networks is proposed.First the vibratory signals in normal and fault states have been analyzed by time series analysis respectively,so state features can be extracted effectively by the time series model's autoregressive coefficients.Then the autoregressive coefficients make up the eigenvectors which are taken as inputs for neural networks training.Consequently the identification and diagnosis of gears in different working conditions,such as normal,crack,gear tooth broken,and partial pitting etc.have been accomplished.The diagnosis result shows that the method based on time series analysis and RBF neural network is feasible for multiple or early fault classification.
出处
《机械传动》
CSCD
北大核心
2008年第4期63-66,共4页
Journal of Mechanical Transmission
关键词
齿轮故障
诊断
时序分析
特征提取
RBF神经网络
Gear faults Diagnosis Time series analysis Feature extraction RBF Neural networks
