基于多注意力和变分编码时序网络的发动机剩余使用寿命预测方法

Prediction Method of Engine Remaining Useful Life Based on Multi-attention and Variational Coding Sequential Network

准确可靠的剩余使用寿命(RUL)预测结果可为决策者提供有价值的信息,以采取适当的维护策略,最大限度地利用设备,避免昂贵的故障维修费用;为了从高噪声的真实工况中对发动机故障进行有效诊断,提出了一种融合多注意力机制和变分编码的时序预测模型(MA-VBLSTM),通过嵌入多注意力机制获得所有特征在空间维度和通道维度的不同权重,以提高退化特征的提取能力;采用变分编码器进行退化信息编码并学习数据间深度隐藏的信息;利用双向长短时记忆网络的长短期时序数据双向处理能力实现发动机RUL的预测;实验结果表明,该模型在发动机CMAPSS数据集的FD001、FD002、FD003、FD004子数据集上,RMSE和Score值相比现有方法分别平均降低5.27%和10.70%、1.37%和1.68%、6.37%和26.94%、3.02%和2.06%。

The prediction results of accurate and reliable residual service life(RUL)can provide valuable information for decision makers,appropriate maintenance strategies are adopted to maximize the use of equipment,and avoid expensive maintenance costs.In order to effectively diagnose engine faults in real working condition with high noise,a sequential prediction model(MA-VBLSTM)integrating multiple attention mechanisms and variational coding is proposed.The different weights of all features in the spatial and channel dimensions are obtained by embedding the multiple attention mechanisms to improve the extraction ability of degraded features;The variational autoencoders are used to encode the degraded information and learn the deeply hidden information between the data;The prediction of the engine RUL is realized by using the bidirectional processing capability with the long and short time series data in the bidirectional long short-term memory network.Compared with existing methods,the RMSE and Score values of the proposed model on FD001,FD002,FD003 and FD004 sub-data sets are reduced by 5.27%and 10.70%,1.37%and 1.68%,6.37%and 26.94%,3.02%and 2.06%,respectively.