基于GWO-CNN-BiLSTM-attention的航空发动机故障诊断方法
DOI:
作者:
作者单位:

海军航空大学青岛校区

作者简介:

通讯作者:

中图分类号:

TN06; TP277

基金项目:

国家社科基金(SKJJ-2022-B-037);国家重点研发计划重点专项课题(2022YFC3102901); 山东省自然科学基金(ZR2020ME131)


Aero engine fault diagnosis method based on GWO-CNN-BiLSTM-attention
Author:
Affiliation:

1.Qingdao Campus of Naval Aviation University,Qingdao,266041;2.China

Fund Project:

National Social Science Foundation;National key research and development plan key special topics;Natural Science Foundation of Shandong Province

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为解决传统故障诊断方法在航空发动机故障诊断中识别精度较低的问题,提出一种基于灰狼优化算法-卷积神经网络-双向长短时记忆网络-注意力机制(GWO-CNN-BiLSTM-attention)的航空发动机故障诊断模型。首先,采用小波软阈值降噪和归一化处理高维航空发动机传感器数据,充分挖掘多维数据中的真实变化特征;其次,构建基于CNN-BiLSTM-attention的故障诊断模型,卷积神经网络(Convolutional Neural Network, CNN)用于特征提取和融合生成若干映射,将数据映射输入双向长短时神经网络(Bidirectional Long Short-Term Memory, BiLSTM)进行训练,捕捉时序数据中的时间依赖性,输出故障识别结果;最后,采用灰狼优化算法(Grey Wolf Optimization, GWO)选取超参数优解,提升模型性能。航空发动机故障诊断实验结果表明,GWO-CNN-BiLSTM-attention在的航空发动机故障诊断实验中取得了93.38%准确率和93.14% AUC值,在噪声条件下验证所提模型具有较高的性能和鲁棒性。

    Abstract:

    In order to solve the problem of low recognition accuracy of traditional fault diagnosis methods in aero engine fault diagnosis, a fault diagnosis model of aero engine fault diagnosis based on gray Wolf optimization algorithm, convolutional neural network, bidirectional long and short time memory network, and attention mechanism (GGO-CNN-BILSTM-attention) was proposed. Firstly, wavelet soft threshold is used to reduce noise and normalize high-dimensional aero engine sensor data to fully explore the real change characteristics of multidimensional data. Secondly, a fault diagnosis model based on CNN-Bilstm-attention is constructed, and a Convolutional Neural Network (CNN) is used for feature extraction and fusion to generate several mappings. The data mapping is input into Bidirectional Long Short-Term Memory (BiLSTM) for training, to capture the time dependence in time series data, and output fault identification results. Finally, Grey Wolf Optimization (GWO) was used to select hyperparameter optimization to improve the model performance. The experimental results of aero engine fault diagnosis show that the accuracy of the proposed model is 93.38% and the AUC value is 93.14% in the aero engine fault diagnosis experiment, which verifies that the proposed model has high performance and robustness under noisy conditions.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-06-27
  • 最后修改日期:2024-07-02
  • 录用日期:2024-07-10
  • 在线发布日期:
  • 出版日期: