基于EMD降噪和LSTM网络的地铁风机轴承寿命预测

The Life Prediction of Bearing for Metro Fan Based on EMD Denoise and LSTM Network

针对地铁风机轴承剩余使用寿命预测问题,提出了一种基于滤波处理的经验模态分解(EMD)降噪与长短期记忆网络(LSTM)相结合的预测方法。对原始振动信号进行EMD分解,得到本征模态分量,结合互相关系数准则和峭度准则筛选有效分量,利用有效分量重构有效去除原始信号中混杂着的噪声。采用单调性和相关性评价指标从重构信号的时域特征和频域特征中选择轴承退化趋势的特征参数,使用特征参数对LSTM网络进行训练,从而实现地铁风机轴承的准确剩余使用寿命预测。为了验证提出方法的有效性,在轴承退化数据集上进行了实验,结果表明提出的方法可以有效的预测轴承退化趋势,对研究地铁风机轴承的健康监测和寿命预测具有重要意义。

Aiming at the problem of predicting the remaining useful life of metro fan bearing,a method based on filtering processing combined EMD noise elimination and LSTM network is proposed.EMD decomposition is performed on the original vibration signal to obtain the intrinsic mode function(IMF),the effective IMFs are selected by combining the correlation coefficient criterion and the kurtosis criterion,and the signal is reconstructed by effective IMFs.The feature parameters of bearing degradation trend is selected from the time-domain and frequency-domain features of the reconstructed signal by using monotonicity and correlation evaluation.By training LSTM network with feature parameters,the accurate remaining useful life of metro fan bearing can be implemented.Experiments on the degraded bearing dataset show the effectiveness of the proposed method.The proposed method is of great significance to the study of health monitoring and remaining useful life of Metro fan bearings.