基于注意力机制的CNN-LSTM剩余寿命预测研究

机械设备剩余寿命的准确预测可以降低昂贵的维护费用,提高机械设备的安全性。随着深度学习的发展以及注意力机制被广泛应用于各个领域,基于数据驱动的剩余寿命预测方法为机械设备寿命预测提供了众多的方法。文章提出了一种基于注意力机制的CNN-LSTM剩余寿命预测方法,该方法利用不同的注意力机制包括通道注意力、CBAM机制和自注意力等进行剩余寿命预测试验。注意力机制可以向CNN-LSTM提取的特征信息分配不同的权重,突出关键的特征信息,过滤无用信息,进而更准确地表示设备的退化特征信息,最终得到设备的剩余寿命。文章对NASA发动机数据集进行了剩余寿命预测试验,同时研究了不同注意力机制影响,试验结果表明,基于注意力机制的方法可以有效地进行剩余寿命预测,所提方法具有一定的应用价值。

Remaining Useful Life Prediction Based on Attention Mechanism CNN-LSTM

Accurate prediction of the remaining useful life of mechanical equipment can reduce the expensive maintenance costs and improve the safety of mechanical equipment.With the development of deep learning,and attention mechanism is widely used in various fields,data-driven residual life prediction method provides many methods for mechanical equipment life prediction.This paper proposes a CNN-LSTM remaining useful life prediction method based on attention mechanism,which uses different attention mechanisms,including SE,CBAM and self-attention,to carry out remaining useful life prediction experiments.The attention mechanism can assign different weights to the feature information extracted by CNN-LSTM,highlight the key feature information,filter the useless information,so as to more accurately represent the degradation feature information of the equipment,and finally obtain the remaining useful life of the equipment.The remaining useful life prediction experiments are carried out on NASA engine data sets,and the effects of different attention mechanisms are studied.The experimental results show that the method based on attention mechanism can effectively predict the remaining useful life,and the proposed method has a certain application value.