基于数据驱动的航空发动机性能趋势分析方法研究

DATA-DRIVEN RESEARCH ON JET-ENGINE PERFORMANCE TREND ANALYSIS

胡阳修

11649182

硕士

航天工程

韩品连

2018-06-05

2018-06-26

哈尔滨工业大学

深圳

现代航空发动机系统越来越复杂，非线性强，涉及到的热力学、动力学过程众多。而发动机的健康状态对飞机的安全飞行至关重要，因此对航空发动机的健康管理和故障监测也越来越受到重视。通过对发动机的状态进行监测分析，可以了解发动机的运行状态和故障情况，方便及时进行维护和修理，提高发动机的可靠性和运行效率。本文首先使用 MATLAB 平台的 Simulink 模块和 T-MATS，基于部件法建模建立了单转子航空发动机的 Simulink 动态模型，在此模型基础上分析了发动机的退化模型，使用得到的指数退化模型生成得到了发动机的多循环性能退化数据集。然后基于数据驱动的方法，使用人工神经网络（ANN）来进行航空发动机非线性系统的性能趋势分析。针对航空发动机的动态退化过程研究，传统的前向神经网络是一个常见的研究方向，本文利用一种改进的动态神经网络（DNN），通过时间延迟和内存机制改善了训练的结果。使用仿真数据将改进的动态神经网络（DNN）与传统前向神经网络相比，验证了改进动态神经网络（DNN）的准确性及鲁棒性。最后基于 JetCat P200-RX 微型涡喷发动机，搭建了试车试验台和数据采集平台，使用真实的微型涡喷多次启动运行数据，进行了实际实验验证。

Modern aero engine systems are more and more complex and nonlinear, which involve with many thermodynamic and kinetic processes. The health state of the engine is very important for the safe flight of the aircraft. Therefore, the health management and fault detect and monitoring of the aero engine is getting more and more attention. The running state and failure of the engine can be logged through monitoring and analysis of the state of the engine. Afterwards, the maintenance and repair work will be conducted timely, all these work improve the engine’s reliability and efficiency.In this study, we first use T-MATS and the Simulink block from MATLAB to build a dynamic model of a single spool aero engine on the basis of component modeling. The engine degradation model is analyzed and the degradation data sets are obtained with the exponential degradation model.Then, an artificial neural network (ANN) is applied to analysing the performance trend of an aeroengine system. The traditional forward neural network is a commonly used research method. In this study, An improved dynamic neural network (DNN) is proposed which greatly improves training results owing to time delay and memory of the former input. The improved dynamic neural network is then compared with the traditional forward neural network to verify the accuracy and robustness.Finally, we build a test bed and data acquisition platform for the JetCat P200-RX micro turbojet engine. With this test bed we made multiple experiments and acquire the micro turbojet starting and running data sets which are used to verify the results.

航空发动机 故障诊断 动态神经网络 主元分析

Aero engine Fault diagnosis Dynamic neural network Principal component analysis

中文

联合培养

南方科技大学

胡阳修. 基于数据驱动的航空发动机性能趋势分析方法研究[D]. 深圳. 哈尔滨工业大学,2018.