基于无参数经验小波变换的风电齿轮箱故障特征提取

Fault feature extraction of a wind turbine gearbox using adaptiveparameterless empirical wavelet transform

风电机组通常以集群规模化运行,机组结构复杂、振动测点多,所产生的振动数据量大,仅靠人工进行故障诊断具有较大挑战。提出基于无参数经验小波变换的风电齿轮箱故障特征提取方法,运用尺度空间方法和经验法则对振动信号的傅里叶谱进行自动分割,获得不同的滤波频带,据此设计一系列经验小波滤波器对信号进行分解和重构,获得不同频带下的经验模式,进一步采用裕度因子对分解后的经验模式进行排序,选取裕度因子最大的经验模式作为故障敏感模式;该方法能在无需预设任何参数的情况下对振动信号进行分解与故障特征提取,具有自适应性。风电试验台和实测风电齿轮箱故障案例验证了方法的有效性。

Wind turbines operate as an equipment cluster,bringing massive vibration signals due to their complex structures and multiply vibration measures.Only analysing the vibration signals to detect fault by human is challenging.In this paper,a fault feature extraction method for wind turbine gearboxes was proposed on the basis of the parameterless empirical wavelet transform.The scale space method and empirical law were utilized to automatically split the Fourier spectrum of the vibration signal,and different frequency bands were obtained.A series of empirical wavelet filters were designed based on the split frequency bands to decompose the signal into multiply empirical modes.The metric of margin factor was adopted to sort the empirical modes,and the empirical mode with maximum margin factor was recognized as the most sensitive one to fault.The proposed method is adaptive without any presented parameters.The fault signals from an experimental platform and a real wind turbine gearbox verified the proposed method.