航空发动机状态预测与健康管理中的气路数据挖掘方法研究

发布时间：2018-08-02 12:54

【摘要】：事后维修与定期维修的航空发动机维修方式过于陈旧,存在效率低下、维修费用巨大、无法有效保证飞行安全性与可靠性等诸多弊端,而且这些弊端在实际工程应用中显露得越来越明显。与传统维修方式相比,航空发动机预测与健康管理(EPHM)技术实现了事后被动维修、定期维修向基于智能系统的视情维修转变,使工程技术人员在特定的时间准确定位发动机的潜在故障并展开主动的维修成为了可能,从而提高飞机维修效率、飞行安全性和飞机可靠性,降低维修成本。以Rolls-Royce公司研制的Trent700发动机气路系统为例,本文针对航空发动机预测与健康管理技术的一项核心问题即数据挖掘技术展开了深入研究。首先对该型号发动机气路系统状态参数的相关数据进行信息挖掘。以涡轮燃气温度TGT为例,对反映气路性能的状态参数的基线进行建模,并对模型进行了验证,基线精度达到了要求,为发动机气路状态的监测奠定了基础。之后以监测过程中计算得到的气路参数偏差值为基础,对其中蕴含的性能衰退信息进行挖掘,建立支持向量机(Support Vector Machines,SVM)算法的回归预测模型,对多个气路参数在未来的偏差量进行了单点趋势预测。此外,在此预测模型的基础上进行拓展,尝试融入模糊信息粒化理论,建立基于信息粒化的支持向量机(Granular Support Vector Machines, GrSVM)预测模型,对未来五个时间序列点进行范围性预测。最后通过仿真实验对模型的预测性能进行检验与分析,结果证明基于SVM的单点预测模型与基于GrSVM的范围预测模型的精度达到要求,为EPHM体系的趋势预测研究提供了参考。
[Abstract]:After maintenance and regular maintenance of aero-engine maintenance mode is too old, there are many shortcomings such as low efficiency, huge maintenance costs, can not effectively ensure flight safety and reliability, and so on. And these malpractices are more and more obvious in practical engineering application. Compared with the traditional maintenance method, the aero-engine prediction and health management (EPHM) technology realizes the passive maintenance after the event, and the periodic maintenance changes to the intelligent system-based maintenance. It is possible for engineers and technicians to accurately locate the potential faults of the engine and carry out active maintenance at a specific time, thus improving the aircraft maintenance efficiency, flight safety and aircraft reliability, and reducing the maintenance cost. Taking the gas path system of Trent700 engine developed by Rolls-Royce Company as an example, this paper makes a deep research on the data mining technology, which is one of the core problems of aero-engine prediction and health management technology. Firstly, the relevant data of the gas path system state parameters of the engine are mined. Taking the turbine gas temperature TGT as an example, the baseline of the state parameters reflecting the performance of the gas path is modeled, and the model is verified. The baseline accuracy meets the requirements, which lays a foundation for the monitoring of the gas path state of the engine. Then, based on the deviation values of gas path parameters calculated in the monitoring process, the performance degradation information contained therein is mined, and the regression prediction model of support vector machine (Support Vector machines) algorithm is established. A single point trend prediction is made for the deviation of multiple gas path parameters in the future. In addition, on the basis of this prediction model, we try to incorporate the fuzzy information granulation theory, establish the support vector machine (Granular Support Vector Machines, GrSVM) prediction model based on information granulation, and predict the range of the next five time series points. Finally, the prediction performance of the model is tested and analyzed by simulation experiments. The results show that the precision of the single point prediction model based on SVM and the range prediction model based on GrSVM meets the requirements, which provides a reference for the trend prediction research of EPHM system.
【学位授予单位】：中国民用航空飞行学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V263.6

【参考文献】