基于VPNRS-RF的飞机液压系统故障诊断模型  

Fault Diagnosis Model for Aircraft Hydraulic System Based on VPNRS-RF

作　　者：李耀华[1] 王签签 LI Yao-hua;WANG Qian-qian(College of Aeronautical Engineering a.Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学航空工程学院,天津300300

出　　处：《机械设计与制造》2024年第2期255-260,共6页Machinery Design & Manufacture

基　　金：国家自然科学基金委员会—中国民航局民航联合研究基金重点支持项目(U2033209);天津市研究生科研创新项目(2019YJSS073)。

摘　　要：针对飞机液压系统故障诊断特征值冗余难以精准获取和准确率不高的问题,提出一种基于变精度邻域粗糙集(Variable Precision Neighborhood Rough Set,VPNRS)和随机森林(Random Forest,RF)相结合的特征选择算法,并以此为基础建立飞机液压系统故障诊断模型。VPNRS-RF算法主要是利用随机森林算法分别对变精度邻域和模糊熵粗糙集约简后的特征进行重要度排序,再次筛选后确定最优特征子集,使用最优特征子集对在线贯序极限学习机(OSELM)分类模型进行训练,从而提高故障信息获取准确率。最后以飞机起落架收放系统为例进行仿真研究,验证了VPNRS-RFOSELM模型的优越性。In order to solve the problems of redundancy and low accuracy in fault diagnosis of aircraft hydraulic system,a feature selection algorithm based on the combination of Variable Precision Neighborhood Rough Set(VPNRS)and Random Forest(RF)was proposed,and a fault diagnosis model of aircraft hydraulic system was established based on this algorithm.The VPNRS-RF algorithm mainly uses the random forest algorithm to sort the importance of the features after reducing the rough set of variable precision neighborhood and fuzzy entropy,again after the screening to determine the optimal feature subset,using the optimal feature subset for online sequential extreme learning machine(OSELM)classification model for training,to improve fault information retrieval accuracy.Finally,the VPNRS-RF-OSELM model is verified by the simulation of aircraft landing gear retraction system.

关 键 词：VPNRS-RF OSELM 液压系统 特征选择 最优特征子集 

分 类 号：TH16[机械工程—机械制造及自动化] TP590.65[自动化与计算机技术]