基于改进差分时域特征和深度学习优化的航空发动机剩余寿命预测算法

Optimized algorithm for aero-engine life prediction based on improved differential time-domain features and deep learning

实现航空发动机剩余寿命的准确预测对于保证飞行安全和提高维修效率具有重要意义,但现有的预测算法往往只是浅层结构,且对各传感器参数之间的相互关系缺乏关联性考虑,限制了对发动机参数信息的深度挖掘。在深度学习理论的基础上,着重考虑不同传感器之间的参数关系,引入差分时域特征扩充特征集,构建了基于长短时记忆网络的寿命预测模型DTF-LSTM。在C-MAPSS数据集上的实验结果表明,该算法相较于其他深度学习算法具有更低的均方根误差(RMSE)值,可以有效实现发动机剩余寿命预测。

It is of great significance to realize the accurate prediction of the remaining life of aero-engine to ensure flight safety and to improve maintenance efficiency.However,the existing algorithms only have shallow structure,the lack of correlation among the parameters of each sensor is considered,which limits the deep excavation of engine parameter information.Based on this issue,this paper focuses on the parameter relationship between different sensors,and expands the feature set by introducing differential time domain feature on the basis of the theory of deep learning.Therefore,the proposed method called DTF-LSTM constructs the life prediction model based on long short time memory network.Experimental results on the C-MAPSS dataset show that the algorithm has lower RMSE value than other deep learning algorithms,which can effectively realize the prediction of engine residual life.