自适应最大相关峭度反褶积方法诊断齿轮轴承复合故障

为了解决传统最大相关峭度反褶积(maximum correlated kurtosis deconvolution，MCKD)在故障诊断中容易出现因参数选择不当而影响诊断效果的问题，该文提出了一种基于量子遗传算法(quantum genetic algorithm，QGA)的自适应最大相关峭度反褶积方法(maximum correlated kurtosis deconvolution with quantum genetic algorithm，QMCKD)用于齿轮和轴承复合故障诊断。通过量子遗传算法自适应选择最大相关峭度反褶积的2个关键参数滤波器长度(L)和反褶积周期(T)。使用QMCKD处理原始振动信号，提取复合故障信号中的所有单个故障信号，分别对单个故障信号进行频谱分析从而识别故障特征。在对齿面磨损-滚动轴承外圈损伤复合故障诊断中，QMCKD能够识别齿轮故障频率及其2～4倍频，识别轴承故障频率及其2～6倍频，且主要频率成分周围干扰谱线很少，故障类型容易识别。与直接频谱分析和变分模态分解(variational mode decomposition，VMD)相比，该方法在诊断效果上具有优越性。在对齿根裂纹-轴承滚动体损伤复合故障诊断中，QMCKD能够突出齿轮故障频率及其2～5倍频，突出轴承故障频率及其2～8倍频，齿轮和轴承故障特征明显，验证了方法的稳定性。试验结果表明QMCKD能够有效识别复合故障中齿轮和轴承的故障特征，可用于齿轮箱的齿轮、轴承复合故障诊断。

Compound fault diagnosis method for gear bearing based on adaptive maximum correlated kurtosis deconvolution

Abstract: Gear boxes often operate in harsh working environment, the core components, gears and bearings, are prone to malfunction. Therefore, in order to find the fault as early as possible and prevent the occurrence of the accident, a reliable and effective fault diagnosis method is needed. Maximum correlated kurtosis deconvolution (MCKD) is an effective diagnosis method which can be applied to the processing of compound fault vibration signal of gear boxes. But the diagnosis performance of MCKD is directly affected by two key parameters (filter length and deconvolution period). In order to overcome the insufficiency of MCKD in parameter selection and improve the diagnosis quality, an adaptive maximum correlated kurtosis deconvolution gearbox fault diagnosis method method is proposed based on quantum genetic algorithm (QMCKD). The key parameters of MCKD are adaptively selected using quantum genetic algorithm (QGA), the raw signal is processed by QMCKD to extracte every single fault signal from the compound fault signal, and the single fault signals are analyzed by frequency spectrum method to identify the fault features. The superiority of QMCKD was verified by comparing with variational mode decompositio (VMD). The results show that QMCKD can develop the advantages of MCKD in signal processing and highlight the periodic components of interest, after processed by QMCKD, the periodic components of the signal are more obvious, the characteristic frequencies are much easier to identify, and the compound faults of the gearbox can be separated more effectively. QMCKD was applied to the compound fault diagnosis of planetary gear tooth surface wear-rolling bearing outer ring damage. In the frequency spectrum of the gear fault signal, the main frequency components were the gear meshing frequency and its 2 times and 3times, the 4 times meshing frequency also could be identified, and the amplitude of the harmonics of meshing frequency were significantly larger than that of health state. The frequency components reflected the fault features of planetary gear tooth surface wear. The fault frequency of the outer ring damage and its high-order harmonics could be completely highlighted in the envelop spectrum of the separated gearing fault signal which reflected the fault features of rolling bearing outer ring damage. In dealing with the experimental signal of compound fault of tooth root crack and bearing rolling element damage, QMCKD successfully identifies the gear fault frequency and its 2-5 times frequency, and the bearing fault frequency and its 2-8 times frequency. The fault characteristics of gear and bearing are obvious, which verifies the stability of the method. QMCKD can effectively identify the fault characteristics of gears and bearings in complex faults, and can be used in the fault diagnosis of gears and bearings in gearboxes.