基于收敛趋势变分模式分解的齿轮箱故障诊断方法

从中心频率的角度出发,深入分析变分模式分解算法中不同初始中心频率的分解特性;利用分解特性对变分模式分解中使用的初始中心频率进行合理更新,在没有先验知识的情况下自适应分解信号的整个分析频带;根据峭度准则,从分解的子信号中选取包含故障信息最丰富的故障分量;对选出的最佳故障分量进行平衡参数优化和稀疏编码收缩处理,并进行包络分...

Gearbox fault diagnosis method based on convergent trend-guided variational mode decomposition

From the perspective of the center frequency, the decomposition characteristics of different initial center frequencies in the variational mode decomposition algorithm were deeply analyzed. Making use of the decomposition characteristics, the initial center frequencies used in the variational mode decomposition were reasonably updated, without the prior knowledge, the entire analysis frequency band of the signal was adaptively decomposed. According to the kurtosis criterion, the fault component with the most abundant fault information was selected from the decomposed sub-signals. Envelope analysis was performed on the optimal fault component which has been processed by the balance parameter optimization and sparse code shrinkage. Based on the decomposition characteristics of variational mode, a complete gearbox fault diagnosis method was constructed based on the convergent trend-guided variational mode decomposition, and the diagnosis method was applied to the early local damage fault identification of gears in automobile transmission gearboxes and fault diagnosis of gearboxes in contact fatigue testing machines. Research results show that there is a convergent trend phenomenon in the variational mode decomposition algorithm. With the gradual increase of the initial center frequency, the convergent center frequency of the extracted mode has a specific convergent relationship with its corresponding initial center frequency. The proposed method can decompose the vibration signal adaptively without the prior knowledge of parameters. In experiment 1, the kurtosis of the fault component obtained by the proposed method is 3.056, and the kurtosis of the fault component after optimization is 24.812. The maximum kurtosis of the fault component in the traditional variational mode decomposition with two different ways for initializing the center frequency is 2.830 and 2.421, respectively. The fast spectral kurtosis analysis method fails to extract the fault component. In experiment 2, the kurtosis of fault component obtained by the proposed method is 3.467, and the kurtosis of the fault component after optimization is 19.780. The maximum kurtosis of the fault component in the traditional variational mode decomposition with two different ways for initializing the center frequency is 3.231 and 3.361, respectively. The fast spectral kurtosis analysis method fails to extract the fault components. The proposed method can enhance the transient characteristics and fault characteristic frequencies, and is more accurate and superior in the gearbox fault diagnosis. 22 figs, 30 refs.