航空学报 > 2022, Vol. 43 Issue (9): 625375-625375   doi: 10.7527/S1000-6893.2021.25375

基于智能神经网络的航空发动机运行安全分析

刘佳奇, 冯蕴雯, 路成, 薛小锋, 潘维煌   

  1. 西北工业大学 航空学院, 西安 710072
  • 收稿日期:2021-02-04 修回日期:2021-03-12 出版日期:2022-09-15 发布日期:2021-04-08
  • 通讯作者: 冯蕴雯,E-mail:fengyunwen@nwpu.edu.cn E-mail:fengyunwen@nwpu.edu.cn
  • 基金资助:
    国家自然科学基金(51875465); 民机专项(MJ-2020-Y-14)

Safety analysis of aero-engine operation based on intelligent neural network

LIU Jiaqi, FENG Yunwen, LU Cheng, XUE Xiaofeng, PAN Weihuang   

  1. School of Aeronautics, Northwestern Polytechnical University, Xi'an 710072, China
  • Received:2021-02-04 Revised:2021-03-12 Online:2022-09-15 Published:2021-04-08
  • Supported by:
    National Natural Science Foundation of China (51875465); Civil Aircraft Project (MJ-2020-Y-14)

摘要: 为提高航空发动机运行安全分析的精度和计算效率, 提出了一种航空发动机运行状态下的时变安全性分析方法。所提方法结合飞行任务特点和航空发动机工作特性, 以提取的快速存取记录器(QAR)信息为分析数据, 依托数据包络分析法, 考虑航空发动机运行过程中发动机工作状态、燃/滑油工作状态、飞机飞行状态、运行外界条件4类因素对运行安全性进行分析。针对航空发动机运行安全影响因素的高度非线性和强耦合性, 提出了一种PSO/BR-ANN智能神经网络模型。所提模型基于人工神经网络(ANN)算法通过改进粒子群优化(PSO)算法和贝叶斯正则化(BR)算法进行优化。通过对B737-800机型一次北京至乌鲁木齐飞行任务的航空发动机运行安全分析, 得到了时变的航空发动机安全裕度, 验证了方法的有效性。通过对PSO/BR-ANN、随机森林、ANN这3种算法进行比较, 说明PSO/BR-ANN智能神经网络模型提高了分析精度和计算效率。所提方法和模型可以为航空发动机的运行安全分析、特情处理、维修及设计提供有益参考。

关键词: 航空发动机, 运行安全, 深度学习, 智能算法, 人工神经网络, QAR运行数据

Abstract: To improve the accuracy and calculation efficiency of aero-engine operation safety analysis, this paper proposes a time-varying safety analysis method, combining the characteristics of flight missions and aero-engine operation, and relying on data envelopment analysis, with Quick Access Recorder (QAR) as the analysis data. The operation safety of the aero-engine is analyzed by considering four factors: engine operation state, fuel/oil operation state, aircraft flight state, and external operation conditions. To solve the high nonlinearity and strong coupling of the factors affecting aero-engine operation safety, an intelligent neural network model (PSO/BR-ANN) is proposed. The proposed model is based on the Artificial Neural Network (ANN) algorithm and optimized by the improved Particle Swarm Optimization (PSO) algorithm and Bayesian Regularization (BR) algorithm. The time-varying aero-engine safety margin is obtained by analyzing the aero-engine operation safety of a B737-800 flight mission from Beijing to Urumqi, verifying the effectiveness of the method. The comparison of PSO/BR-ANN, random forest and ANN shows that PSO/BR-ANN improves the analysis accuracy and calculation efficiency. The proposed method and model can provide useful reference for aero-engine operation safety analysis, special case treatment, maintenance and design.

Key words: aero-engine, operation safety, deep learning, intelligent algorithms, artificial neural networks, QAR operation data

中图分类号: