双树复小波变换用于转子模拟故障诊断研究

Study on Dual-tree Complex Wavelet Transform Used in Rotor Simulation Fault Diagnosis

针对机械故障微弱特征提取的难题,提出结合双树复小波变换(DTCWT)及迭代奇异值分解(ISVD)降噪的分析方法,凸显分解后各频带信号的特征。该方法首先利用2个并行的、高度对称的Q-shift滤波器组对原始振动信号进行滤波处理,获得复小波系数和复尺度系数;然后对各层小波系数和末层尺度系数分别进行单枝重构,得到各层细节信号及末层近似信号;最后对各层细节信号和末层近似信号进行ISVD降噪,并全部叠加,即获得降噪后的振动信号。对转子故障模拟试验的信号分析表明,该方法具有很好的降噪效果,有效地提取出了隐藏在转子振动信号中的微弱周期性冲击成分,并清晰刻画了真实的碰摩故障特征。与传统小波相比,该方法具有运算效率高、平移不变性好、抗频率混叠和完全重构等优点。研究结果为现场设备转子故障诊断奠定了基础。

In order to dual-tree complex wavelet extract the weak feature transform （ DTCWT ） of mechanical vibration signal, a new method combining with and iterative singular value decomposition （ISVD） is presented. First, the original vibration signal is filtered by two parallel and symmetric Quarter-shift filter banks, and consequently the complex wavelet coefficients and complex scaling coefficients are acquired. Then the detail signal of each layer and approximation signal of last layer are constructed from the wavelet coefficients of each layer and scaling coefficients of last layer, respectively. Finally, the detail signal and approximation signal are de-noised by ISVD and superposed into expected signal. The rotor experiment results show that the DTCWT-ISVD method effectively extracts the weak impact signal hidden in rotor vibration signal, and clearly depicts the real rub-impact fault feature, due to the satisfied noise reduction effect. Compared with the traditional wavelet algorithm, the DTCWT method has the advantages of high computational efficiency, excellent translation invariant, anti-aliasing and complete reconstruction. The study results lay a foundation for rotor fault diagnosis of field equipment.