摘要
为精准研判飞行模拟器故障问题,提升飞行模拟器故障应急处理能力,提出了1种基于Stacking集成学习的设备故障预判方法。首先,针对设备故障数据具有高维度和时序性的特点,采用正态分布法,对采集的原始故障数据进行缺失值和异常值处理,获取归一化数据,并基于核主成分分析法,提取数据的特征并降低数据维度;然后,构建以CNN、RF、BP神经网络为基分类器,XGBoost为元分类器的Stacking集成学习模型,采用交叉验证训练策略,对设备的故障类型作出预判;最后,构架典型案例。仿真结果表明,所提方法可正确有效地识别设备的多种故障状态。
In order to predict fault states of flight simulator accurately and effectively,and improve the ability to deal with emergencies of the flight simulator equipment fault.The equipment fault prejudgment method based on Stacking ensemble learning is proposed.The normal distribution method is used to process the missing values and outliers of the collected original fault data to obtain the normalized data for the characteristics of equipment fault data with high dimensionlity and time-series.Based on kernel principal component analysis,the features of data are extracted and the dimensions of data are reduced.A Stacking ensemble learning model with CNN,RF and BP neural networks as the base classifier and XGBoost as the meta-classifier is constructed,and a cross-valiclation training strategy is used to predict the fault types of equipment.Typical cases are constructed and simulation results show that the proposed method can accurately and effectively identify various fault states of equipment.
作者
王伟
周晓光
卓靖升
王刚
WANGWei;ZHOU Xiaoguang;ZHUO Jingsheng;WANG Gang(The 91475th Unit of PLA,Huludao Liaoning 125001,China;The 92635th Unit of PLA,Qingdao Shandong 266200,China)
出处
《海军航空大学学报》
2023年第5期413-418,共6页
Journal of Naval Aviation University
基金
国家社会科学基金(2020-SKJJ-C-030)。
