摘要
为提高航空发动机在试飞过程中的安全性,提升试飞安全监测的智能化程度,以某国产大推力军用涡扇发动机历史故障试飞数据为基础,建立了不同故障类型特征库,考虑航空发动机故障诊断小样本分类问题,提出了一种变尺度窗口切片的数据增强方法,基于扩充的数据样本库,构建了双向长短时记忆网络故障识别模型。测试数据验证表明,故障的识别和分类正确率能达到95%以上,满足了工程使用要求。
In order to improve the safety of aeroengine during flight test and increase the intelligence level of flight test safety monitoring,a feature library of different fault types is established based on the historical fault flight test data of a domestic high thrust military turbofan engine.Considering the classification of small samples for aeroengine fault diagnosis,a data enhancement method with variable scale window slice is proposed,Based on the extended data sample library,a two-way long short memory network fault identification model is constructed.The test data validation shows that the correct rate of fault identification and classification can reach more than 95%,which can meet the engineering requirements.
作者
刘振刚
LIU Zhengang(China Flight Test and Research Institute,Xi’an,Shaanxi 710089,China)
出处
《自动化应用》
2023年第10期179-181,共3页
Automation Application
关键词
试飞数据
变尺度窗口切片
数字增强
双向长短时记忆网络
特征提取与分类
flight test data
variable scale window slice
digital enhanced
bidirectional long short memory network
feature extraction and classification
