摘要
为了有效开展民机襟翼偏角可靠性监测,结合快速存取记录器(quick access recorder, QAR)数据,基于Kriging模型,引入分解协调策略、平衡器优化(equilibrium optimizer, EO)算法和移动最小二乘(moving least square, MLS),提出了一种基于协同智能移动Kriging(collaborative intelligent moving Kriging, CIMK)方法。其中,分解协调策略用于处理襟翼左右偏角之间的关系,MLS用于选取有效建模样本并实现Kriging模型待定系数求解,EO算法用于确定MLS最优的紧支撑域半径。针对襟翼左右不对称进行故障原因分析,明确QAR数据中影响襟翼偏角的主要特征参数;结合相关影响参数的QAR数据,运用CIMK实现民机襟翼偏角模型(极限状态函数)的构建;基于构建的CIMK模型,通过Monte Carlo抽样方法进行民机襟翼偏角可靠性及影响性分析;以某型国产民机襟翼偏角为例,对所提出的方法进行可行性分析。研究结果表明:当襟翼偏角为告警许用值3°时,可靠度为0.450 2,影响襟翼偏角的因素重要程度依次为马赫数、左攻角、右攻角等。与RSM、Kriging、SVM和BP-ANN方法对比发现:在建模特性方面,所提方法平均绝对误差精度相对提高了53.02%,51.43%,49.03%和44.04%,平均相对误差精度相对提高了68.36%,66.76%,64.41%和62.64%;建模效率相对于Kriging、SVM和BP-ANN分别提高了50.62%,26.35%和43.01%;在仿真性能方面,当仿真次数为103次时,分析精度分别提高了8.82%,7.25%,6.22%和3.98%。
To effectively monitor the reliability of civil aircraft flap deflection angle,combined with the quick access recorder(QAR),the collaborative intelligent moving Kriging(CIMK)method is proposed by absorbing the Kriging model,decomposition and co-ordination strategy,equilibrium optimizer(EO),and moving least square(MLS).Among them,the decomposition coordination strategy is used to deal with the relationship between the flaps left and right deflection angles.MLS is employed to select effective modeling samples and solve the undetermined coefficients of Kriging model.EO method is applied to determine optimizing the local compact support region radius of MLS.Firstly,the fault reason for flap left-right asymmetry is analyzed to clarify the main characteristic parameters in QAR data.Secondly,combined with the QAR data of relevant influencing parameters,the civil aircraft flap deflection model(limit state function)is constructed by using CIMK.Then,the reliability and influence of civil aircraft flap deflection angle are analyzed by Monte Carlo(MC)sampling method.The results show that when the flap deflection angle is 3°,the reliability is 0.4502,and the important factors affecting the flap deflection angle are Mach number,left angle of attack,right angle of attack,etc.Compared with the response surface method(RSM),Kriging,support vector machine(SVM),and back-propagation-artificial neural network(BP-ANN),the average absolute error accuracy of the proposed method is relative improved by 53.02%,51.43%,49.03%,and 44.04%,the average relative error accuracy is relative improved by 68.36%,66.76%,64.41%,and 62.64%,and the modeling efficiency is relative improved by 50.62%,26.35%,and 43.01%respectively compared with Kriging,SVM and BP-ANN.When the number of simulations is 103,the analysis accuracy is relative improved by 8.82%,7.25%,6.22%,and 3.98%respectively.
作者
刘磊
腾达
冯蕴雯
LIU Lei;TENG Da;FENG Yunwen(The First Aircraft Research Institute of AVIC,Xi′an 710089,China;School of Aeronautics,Northwestern Polytechnical University,Xi′an 710072,China;Institute of Reliability and Operation Support Engineering,Northwestern Polytechnical University,Xi′an 710072,China)
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2023年第2期253-263,共11页
Journal of Northwestern Polytechnical University
