基于XGBoost的民航飞机发动机性能参数预测模型
2023,31(6):46-52
摘要：为提高民航飞机发动机性能参数的预测精度，本文提出一种基于模糊推理和XGBoost算法的发动机性能参数预测方法。对发动机进行总体性能分析，确定油门杆位置、气压高度、总温、全重、马赫数及飞行阶段为影响发动机性能参数的主要因素。其次采用模糊推理对快速存取记录器（QAR）数据进行纵向飞行阶段划分，消除人为划分训练数据对预测精度的主观影响。最后，建立各发动机性能参数的XGBoost预测模型，并与多种预测模型进行对比实验。实验结果表明：对发动机N1、燃油流量参数的预测，XGBoost预测模型相比支持向量回归（SVM）、线性回归模型和BP神经网络，其精度更高且不需要对训练数据进行缩放。
关键词：航空发动机;模糊推理;XGBoost;QAR数据;性能参数预测
Aircraft engine performance parameters prediction Model based on XGBoost
Abstract：In order to improve the prediction accuracy of aircraft engine performance parameters, a new aeroengine performance prediction method based on fuzzy theory and XGBoost algorithm was proposed. Through the overall performance analysis of aeroengine, the angle of throttle, altitude, total temperature, gross weight, mach number and flight phase were identified as the main factors affecting aeroengine performance; Secondly, the fuzzy theory was used to divide the QAR data into vertical flight phase data, eliminating the subjective influence on prediction accuracy, which caused by artificially dividing the training data. Finally, XGBoost prediction model of aeroengine parameters was established, and compared with various prediction models. For the prediction of aeroengine N1 and fuel flow parameters, the experimental results show that the XGBoost prediction model which does not require scaling of training data has higher accuracy than support vector regression (SVM), liner regression models and BP neural network.
Key words：aeroengine; fuzzy theory ;XGBoost; QAR data; performance parameter prediction
收稿日期：2022-10-18
基金项目：天津市自然科学基金,多元投入青年项目,21JCQNJC00710,面向复杂航电运行安全的分布式融合仿真场景生成及边界自适应测试。
