基于自适应VMD与GRNN的转子系统故障诊断方法研究

Fault Diagnosis Method of Rotor Systems Based on Adaptive VMD and GRNN

提出一种基于自适应变分模态分解(Variational Mode Decomposition,VMD)与广义回归神经网络(Generalized Regression Neural Network,GRNN)的故障诊断方法,有效解决转子系统振动信号特征提取与复合故障模式识别的问题。首先通过VMD将采集到的原始信号自适应分解为一系列的内涵模态分量(Intrinsic Mode Functions,IMF),然后根据相关系数-峭度准则选取IMF分量进行信号重构。最后获取重构信号的精细复合多尺度散布熵(Refined Composite Multiscale Dispersion Entropy,RCMDE)、均方根以及重心频率构成特征向量集,输入到GRNN神经网络进行训练和故障模式识别。数值仿真与故障模拟实验结果表明:采用基于自适应VMD与GRNN神经网络的方法可有效识别转子系统中的多故障模式。

A fault diagnosis method based on adaptive variational mode decomposition(VMD)and general regression neural network(GRNN)is proposed,which can effectively solve the problems of feature extraction and compound fault pattern recognition of rotor system vibration signals.Firstly,the original signal is adaptively decomposed into a series of intrinsic mode functions(IMF)components by VMD.Then,the IMF components are selected for signal reconstruction according to the correlation coefficient-kurtosis criterion.Finally,the refined composite multiscale dispersion entropy(RCMDE),mean square root and barycenter frequency of the reconstructed signal are calculated to form a feature vector set,which is then input to GRNN mode for training and fault pattern recognition.The results of numerical simulation and fault simulation experiments show that the method based on adaptive VMD and GRNN neural network can effectively identify multi-fault modes in rotor systems.