基于VMD和AR模型的转子裂纹故障诊断方法

CRACK FAULT DIAGNOSIS OF ROTOR BASED ON VMD AND AR MODEL

针对裂纹转子振动位移信号不平稳、非线性,故障特征难提取,且在故障初期缺少对样本的收集和整理等问题,提出了一种基于VMD(Variational Mode Decomposition-变分模态分解)和AR(Auto Regressive-自回归)模型的转子裂纹故障诊断方法。采用VMD方法对转子位移信号处理,得到若干个平稳的本征模态函数(Intrinsic Mode Function),再分别对每个IMF分量建立AR模型,利用最小二乘法计算模型参数和残差的方差,将模型参数和残差的方差作为系统状态特征向量,建立马氏距离(Mahalanobis distance)判别函数,通过设置相应加权参数得到综合距离来实现裂纹故障诊断。最后采用VMD和AR方法进行了转子裂纹故障的诊断实验,实验结果表明,基于VMD和AR模型诊断转子裂纹故障是可行和有效的,克服了AR模型在诊断转子裂纹故障的不足。

Aiming at the vibration displacement signals of cracked rotors is unstable and non-linear,and the fault features are extracted difficultly,and the sample collection is lacking in the initial stage of fault,the crack fault diagnosis method based on Variational Mode Decomposition and AR model is proposed.A number of Intrinsic Mode Function are obtained by VMD as the preprocessing of the AR model.The AR model for each IMF component is established.The variance of the model parameters and residuals is calculated by the least square method.The model parameters and the variance of the residual are used as the characteristic vector of the system state.From the Mahalanobis distance and setting the corresponding weighted parameters,the synthetic distance is used to diagnose crack fault of rotor.Through the analysis of the experimental data,the feasibility of the fault diagnosis method based on the variational mode decomposition and the AR model are successfully verified,which provides a method for the fault diagnosis of the rotor,and the shortcomings of AR model in the diagnosis of rotor crack fault are overcome.