一种自适应选取参数的改进变分模态分解方法

Improved variational mode decomposition method with adaptive parameter selection

针对传统的变分模态分解(VMD)方法中模态数和惩罚参数难以确定的问题,提出了一种自适应选取参数的改进变分模态分解方法。首先,综合考虑了故障的冲击性和周期性特点,以Gini指数和谱峰比指标为基础构建了加权谱峰比(WSPR)指标;然后,采用非洲秃鹫优化算法(AVOA)进行了寻优,得到了最佳的模态数和惩罚参数组合,克服了人为主观选择参数的弊端;最后,在VMD分解信号后,利用加权谱峰比最大原则自适应选取了敏感内涵模态分量,对最佳IMF进行了包络解调分析,提取了滚动轴承早期故障特征,利用仿真信号、单一故障滚动轴承试验信号及复合故障滚动轴承试验信号对所述方法进行了验证。实验结果表明:该方法可以准确地提取出仿真信号的故障频率(100 Hz)、单一故障信号的故障频率(236.4 Hz)和复合故障信号的故障频率(内圈故障频率149.14 Hz、外圈故障频率86.39 Hz),并且在与其他方法和指标的对比中,其最佳IMF的包络谱图中故障特征频率及其倍数频的谱峰更加明显,准确率更高且鲁棒性更强。研究结果表明:该方法能够有效提取轴承早期故障信号的微弱特征,实现故障类型准确识别的目的。

In response to the difficulty in determining the number of modes and penalty parameters in traditional variational mode decomposition(VMD) methods,an improved variational mode decomposition method with adaptive parameter selection was proposed.Firstly,the impact and periodicity characteristics of faults were comprehensively considered,and a weighted spectral peak ratio(WSPR) index was constructed based on the Gini index and spectral peak ratio index.Secondly,in order to obtain the optimal combination of mode number and penalty parameter,the African vulture optimization algorithm(AVOA) was used to iteratively optimize the mode number and penalty parameter of the variational mode decomposition method,which overcame the drawbacks of subjective parameter selection.The constructed weighted spectral peak to peak ratio index could not only serve as the objective function for optimizing the parameters of the African vulture optimization algorithm,but also adaptively select intrinsic modal functions after signal decomposition using variational modal methods.Finally,the selected optimal intrinsic modal functions were subjected to envelope demodulation analysis to extract early fault characteristics of rolling bearing faults.The method was validated using simulated signals,single-fault rolling bearing test signals,and composite-fault rolling bearing test signals.The experimental results show that the method can accurately extract the fault frequency(100 Hz) of the simulated signal,the fault frequency(236.4 Hz) of the single fault signal,and the fault frequency(inner ring fault frequency 149.14 Hz,outer ring fault frequency 86.39 Hz) of the composite fault signal.In comparison with other methods and indicators,the spectral peaks of the fault characteristic frequency and its multiple frequencies in the optimal IMF envelope spectrum are more prominent and obvious,with higher accuracy and stronger robustness.The research results show that the method can effectively extract the weak features of the early fault signal of the bearing and achieve accurate identification of the fault type.