基于VMD-LSTM的滚动轴承退化状态识别

Degradation State Identification of Rolling Bearings Based on VMD-LSTM

针对滚动轴承退化信号的非平稳、非线性特点以及全寿命退化状态难以有效识别的问题,提出一种基于变分模态分解(VMD)和长短时记忆神经网络(LSTM)相结合的滚动轴承退化状态识别方法。该方法首先采用麻雀搜索算法(SSA)对VMD的两个参数(模态分量个数和惩罚因子)进行优化;然后将滚动轴承振动信号分解成若干个本征模态函数(IMF),再根据皮尔逊相关系数选择VMD分解得到的敏感IMF分量,对其重构后进行特征提取;最后将多维退化特征输入LSTM模型训练,建立退化状态模型。实验结果表明该方法能够准确识别轴承的退化状态,验证了该方法的优越性。

In order to solve the problems of non-stationarity,non-linearity of rolling bearing degradation signal and difficulty in effectively identifying the whole life degradation state,a rolling bearing degradation state identification method based on combination of variational mode decomposition(VMD)and long and short term memory neural network(LSTM)is proposed.First,the sparrow search algorithm(SSA)is used to optimize the two parameters(number of modal components and penalty factor)of VMD.Then,the rolling bearing vibration signal is decomposed into several intrinsic mode functions(IMF),and the VMD decomposed sensitive IMF components are selected according to Pearson correlation coefficient,and feature extraction is carried out after reconstruction.Finally,multi-dimensional degradation characteristics are input into LSTM model training to establish degradation state model.The experimental results show that this method can accurately identify the degradation state of bearings and verify the superiority of this method.