[author_cn_name].[cn_title][J].空军工程大学学报:自然科学版,[year_id],[volume]([issue]):[start_page]-[end_page] 基于双向LSTM神经网络的航空发动机故障预测-Fault Prognostic of Aeroengine Using Bidirectional LSTM Neural Network
文章摘要
曾慧洁,郭建胜.基于双向LSTM神经网络的航空发动机故障预测[J].空军工程大学学报:自然科学版,2019,20(4):26-32
基于双向LSTM神经网络的航空发动机故障预测
Fault Prognostic of Aeroengine Using Bidirectional LSTM Neural Network
  
DOI:
中文关键词: 故障预测  时间序列  双向LSTM神经网络
英文关键词: fault prognostic  time sequence  bidirectional LSTM neural network
基金项目:
作者单位
曾慧洁,郭建胜 空军工程大学大学装备管理与无人机工程学院,西安,710051 
摘要点击次数: 16
全文下载次数: 20
中文摘要:
      准确的航空发动机故障预测能够为维修决策提供依据,提高装备完好率,避免灾难性故障并最小化经济损失。根据航空发动机传感器数据特点,提出一种基于双向长短期记忆(LSTM)神经网络的故障预测方法,建立故障预测模型,包括数据预处理、网络模型设计、训练与测试,得到在多种工作条件和故障下具有较强泛化能力的神经网络预测模型。使用C-MAPSS数据集对模型进行仿真验证,所提出的双向LSTM故障预测模型通过与RNN、GRU、LSTM时间序列模型对比,误差下降33.58%,得到更高的预测精度,非对称评分下降71.22%,具有更好的适应性。
英文摘要:
      Aeroengine fault prognostic can provide basis for maintenance decision making which can help to avoid catastrophic failures and minimize economic losses. According to the characteristics of aeroengine sensor data, a fault prognostic method based on bidirectional long short term memory (LSTM) neural network is proposed. A fault prognostic model is established, including data preprocessing, network design, training and testing. The model structure has strong generalization ability under various working conditions and faults is obtained. The model was validated using the C MAPSS data set. Compared with the RNN, GRU and LSTM time series models, the results show that the proposed Bidirectional LSTM fault prognostic model has an average error of 33.58%, which has better adaptability. The asymmetric score decreased by 71.22%, resulting in higher prediction accuracy.
查看全文   查看/发表评论  下载PDF阅读器
关闭