一种时序-BP网络在飞轮系统齿轮故障诊断中的应用研究

Study on Gear Fault Diagnosis in Flywheel Energy Storage System Based on Time Series Analysis and BP Networks Method

提出了一种基于时序和反向传播网络(Back-Propagation Network,BP)相结合的诊断方法。通过对飞轮储能系统(FESS)齿轮箱正常和故障运行状态的振动信号进行时序分析,利用时序模型提取齿轮不同状态的特征,并以其自回归参数组成特征向量作为BP网络分类器的输入进行网络训练,从而实现了对齿轮正常、裂纹和剥落状态的识别与诊断。实验结果表明,基于时序-BP网络结合的方法对于飞轮系统齿轮故障分类和检测是一种非常有效的诊断手段。

The time series analysis combined with Back-Propagation network method was presented for flywheel energy storage system （FESS）. The vibratory signals in normal and fault states were analyzed by time series analysis. The different state features were extracted by the time series model. The autoregressive coefficient eigenvectors were treated as inputs for BP network training. Different states such as normal, crack, and spalling etc. were identified and diagnosed effectively. The result shows that the method based on time series analysis and BP network is an effective means for classification and detection in FESS.