基于谐波小波包的旋转机械故障诊断新方法

摘要

For fault vibration signal of rotating machinery is complicated dynamic nonstationary, a novel fault diagnosis method based on harmonic wavelet packet and Elraan neural network is proposed. Fault signal is decomposed by harmonic wavelet packet and the characteristic vectors of energy are extracted. Then these vectors are input into Elman neural network to distinguish faults. The result of experiment and comparation to BP neural network show that this method has significant advantage in fault diagnosis of rotating machinery.%针对旋转机械的故障振动信号通常为复杂的动态非平稳信号,提出一种基于谐波小波包和Elman神经网络的故障诊断新方法.应用谐波小波包对信号进行分解,提取倍频能量特征向量,代入Elman神经网络,实现故障分类.通过试验分析及与BP网络的诊断结果对比,表明该方法在旋转机械的故障诊断方面具有显著优势.