为了研究分析飞机的动力装置在执行飞行任务过程中的运行可靠性，针对运行可靠性影响因素的多维、耦合的特点，采用机器学习方法对动力装置运行可靠性的时变规律及其相关影响因素进行分析。提出了考虑动力装置的工作状态、飞机的运行外界条件、飞机的飞行状态3类因素分析动力装置实时运行状态下的时变可靠性方法；并基于飞机实际运行的快速存取记录器（QAR）数据，梳理了动力装置运行可靠性分析相关的3类因素、16个主要特征。结合飞机运行的时空关系，采用数据包络分析（DEA）方法对飞机动力装置的工作状态特性与性能裕度进行非参数分析，基于提取的QAR数据特征，采用随机森林、多变量神经网络回归算法，建立2种基于机器学习的动力装置运行可靠性分析模型。以B737-800机型为例，对一次北京至珠海的飞行任务的动力装置相关运行数据进行分析，对2种机器学习分析模型进行训练与测试研究。分析结果表明：对动力装置工作状态特性贡献度最大的特征依次为计算空速、飞行时间与飞行高度；对动力装置性能裕度贡献度最大的特征依次为动力装置工作状态特性、雷达气象与飞行时间。所采用的2种机器学习方法能较好反映动力装置运行过程的时变可靠性规律，可为动力装置的运行与特情处理提供参考。
To study the operational reliability of aircraft power plants during flight missions, we analyze the time-varying law and related influencing factors of power plant operational reliability using the machine learning method, meanwhile considering the multi-dimensional and coupling characteristics influencing the operational reliability. An operational reliability analysis method is proposed for power plants considering three factors: the operating state of the power plant, the operating state of the aircraft, and the operating environment of the power plant. Based on the QAR (Quick Access Recorder) data of the actual operation of the aircraft, this method identifies three kinds of factors and 16 main characteristics related to the operational reliability analysis of the power plant. Combined with the space-time relationship of aircraft operation, non-parametric analysis of the working state characteristics and the performance margin of aircraft power plants is conducted using DEA (Data Envelopment Analysis). According to the proposed QAR data characteristics, the random forest and multivariable neural network regression algorithm is used to establish two kinds of operational reliability analysis models of power plants based on machine learning. Taking B737-800 aircraft as an example, this paper analyzes the power plant operation data of a flight mission from Beijing to Zhuhai, and studies the training and testing of two machine learning analysis models. The analysis results show that the features contributing most to the power plant operating state characteristics are calculated airspeed, flight time, and flight altitude; those to the power plant performance margin are power plant operating state characteristics, radar weather, and flight time. The two types of machine learning methods proposed can well reflect the time-varying reliability law of the power plant operation process, providing reference for power plant operation and special situation handling.
[1] 叶博嘉, 鲍序, 刘博, 等. 基于机器学习的航空器进近飞行时间预测[J]. 航空学报, 2020, 41(10):324136. YE B J,BAO X,LIU B,et a1. A machine learning method to aircraft approach time prediction[J]. Acta Aeronautica et Astronautica Sinica, 2020, 41(10):324136(in Chinese).
[2] SUN J Z, LI C Y, LIU C, et al. A data-driven health indicator extraction method for aircraft air conditioning system health monitoring[J]. Chinese Journal of Aeronautics, 2019, 32(2):409-416.
[3] BRYAN M, SANTANU D, KANISHKA B, et al. Discovering anomalous aviation safety events using scalable data mining algorithms[J]. Journal of Aerospace Information Systems,2013,10(10):467-475.
[4] OEHLING J L, BARRY D J. Using machine learning methods in airline flight data monitoring to generate new operational safety knowledge from existing data[J]. Safety Science, 2019,114(4):89-104.
[5] ABRAR O A, RASHID M. An ensemble machine and deep learning model for risk prediction in aviation systems[C]//2020 6th Conference on Data Science and Machine Learning Applications (CDMA), 2020:54-59.
[6] BEULEN M, SCHERP L, SANTOS B F. Dynamic evaluation of airline Crew's flight requests using a neural network[J]. EURO Journal on Transportation and Logistics, 2020, 8:1-12.
[7] LHÉRITIER A, BOCAMAZO M, DELAHAYE T, et al. Airline itinerary choice modeling using machine learning[J]. Journal of Choice Modelling, 2019, 7(31):198-209.
[8] ZHOU D, ZHUANG X, ZUO H F, et al. Deep learning-based approach for civil aircraft hazard identification and prediction[J]. IEEE, 2020,8:103665-103683
[9] 曾声奎. 可靠性设计分析基础[M]. 北京:北京航天航空大学出版社, 2015, 1-14. ZENG S K. Basis of reliability design and analysis[M]. Beijing:Beihang University Press, 2015, 1-14(in Chinese).
[10] 孙元章, 程林, 何剑.电力系统运行可靠性理论[M]. 北京:清华大学出版社, 2012:1-15. SUN Y Z, CHENG L, HE J. Power system operational reliability theory[M]. Beijing:Tsinghua University Press, 2012:1-15(in Chinese).
[11] 康锐. 确信可靠性理论与方法[M]. 北京:国防工业出版社, 2020:3-136. KANG R. Belief reliability theory and methodology[M]. Beijing:National Defense Industry Press, 2020:3-136(in Chinese).
[12] 熊贝贝. 基于数据包络分析的环境绩效评价方法及其应用研究[D]. 合肥:中国科学技术大学, 2019. XIONG B B. Environmental efficiency evaluation and its applications via data envelopment analysis[D].Hefei:University of Science and Technology of China, 2019(in Chinese).
[13] NAHANGI M, CHEN Y T, MCCABE B. Safety-based efficiency evaluation of construction sites using data envelopment analysis (DEA)[J]. Safety Science, 2019, 113:382-388.
[14] TELLES E S, LACERDA D P, MORANDI M I W M, et al. Drum-buffer-rope in an engineering-to-order system:An analysis of an aerospace manufacturer using data envelopment analysis (DEA)[J]. International Journal of Production Economics, 2020, 222:107500.
[15] CHARNES A, COOPER W W, RHODES E. Measuring the efficiency of decision making units[J]. European Journal of Operational Research, 1978, 2:429-444.
[16] 杨国梁, 刘文斌, 郑海军. 数据包络分析方法(DEA)综述[J]. 系统工程学报, 2013, 28(6):840-860. YANG G D, LIU W B, ZHENG H J. Review of data envelopment analysis[J]. Journal of Systems Engineering, 2013, 28(6):840-860(in Chinese).
[17] 《运筹学》教材编写组. 运筹学.[M]第4版. 北京:清华大学出版社, 2012. Textbook Compiling Group of Operations Research. Operations research.[M]4th Edition. Beijing:Tsinghua University Press, 2012(in Chinese).
[18] THANASSOULIS E, KORTELAINEN M, JOHNES G, et al. Costs and efficiency of higher education institutions in England:a DEA analysis[J]. Journal of the Operational Research Society, 2011, 62(7):1282-1297.
[19] 马立杰. DEA理论及应用研究[D].济南:山东大学, 2007. MA L J. Study on DEA theory and its applications[D].Jinan:Shandong University, 2007(in Chinese).
[20] LEO B. Random forests[J]. Machine Learning, 2001,45:5-32.
[21] 刘玥. 基于集成学习的工业过程监测[D].杭州:浙江大学, 2019. LIU Y. Ensemble learning based industrial process monitoring[D]. Hangzhou:Zhejiang University, 2019(in Chinese).
[22] 邱锡鹏. 神经网络与深度学习[M]. 北京:机械工业出版社, 2020:82-98. QIU X P. Neural networks and deep learning[M]. Beijing:Machinery Industry Press, 2020:82-98(in Chinese).
CJA