改进MVMD方法及其在轴承故障诊断中应用

Improved MVMD Method and Its Application in Bearing Fault Diagnosis

随着多传感器采集系统的发展,多元信号的同步处理变得尤为重要。多变量变分模式分解(Multivariate Variational Mode Decomposition,MVMD)方法是一种在变分模式分解方法基础上发展起来的多元信号分析方法,具有物理意义清晰、抗干扰能力强等特点。然而,须预先设置分解模式个数和带宽平衡参数对其在实际工程中的应用产生限制。为此,提出一种改进MVMD方法。该方法构建带宽估计策略和MVMD单分量迭代分解策略,能够合理确定带宽平衡参数,并在固定带宽下基于峭度准则迭代分解出故障分量。多通道的轴承故障试验信号分析结果表明:提出的方法能够合理确定目标分量的带宽,且无须预设模式分量个数;该方法运行所需时间为8.3 s,相较于传统的固定参数MVMD、多维经验模态分解和快速谱峭度,可在没有损失时间效率的前提下更准确地识别出多通道故障特征。

With the development of multi-sensor acquisition system,the synchronous processing of multiple signals has become particularly important.Multi-variable variational mode decomposition(MVMD)is a multivariate signal analysis method developed on the basis of variational mode decomposition.It has the characteristics of clear physical meaning and strong anti-interference ability.However,bandwidth balance parameter and the number of decomposition modes need to be set in advance,which limits its application in practical engineering.Therefore,this paper proposes an improved MVMD method.This method constructs a bandwidth estimation strategy and an MVMD single-component iterative decomposition strategy,which can reasonably determine bandwidth balance parameter,and decompose fault components based on kurtosis criterion under fixed bandwidth.The analysis results of multi-channel bearing fault test signals show that the proposed method can reasonably determine the bandwidth of the target component and does not need preset the number of decomposition modes;The operation time of the proposed method is 8.3 s.Compared with the traditional fixed parameter MVMD,multivariate empirical mode decomposition and the fast kurtosis,the proposed method can identify the multichannel fault features more accurately without losing time efficiency.