北京航空航天大学学报 ›› 2019, Vol. 45 ›› Issue (9): 1831-1837.doi: 10.13700/j.bh.1001-5965.2018.0769

• 论文 • 上一篇    下一篇

基于DWNN的机电作动器渐变性故障诊断

王剑, 王新民, 谢蓉, 曹宇燕, 李婷   

  1. 西北工业大学 自动化学院, 西安 710072
  • 收稿日期:2018-12-29 出版日期:2019-09-20 发布日期:2019-09-29
  • 通讯作者: 谢蓉 E-mail:xierong@nwpu.edu.cn
  • 作者简介:王剑,男,博士研究生。主要研究方向:飞行器机电系统故障诊断技术;王新民,男,教授,博士生导师。主要研究方向:飞行器控制技术、飞行器机电系统故障诊断技术;谢蓉,女,博士,副教授。主要研究方向:飞行器控制技术、飞行器健康管理技术;曹宇燕,女,博士研究生。主要研究方向:飞行器机电系统故障诊断技术;李婷,女,博士研究生。主要研究方向:飞行器机电系统故障诊断技术。
  • 基金资助:
    国家自然科学基金(61703341)

Gradual fault diagnosis for electromechanical actuator based on DWNN

WANG Jian, WANG Xinmin, XIE Rong, CAO Yuyan, LI Ting   

  1. College of Automation, Northwestern Polytechnical University, Xi'an 710072, China
  • Received:2018-12-29 Online:2019-09-20 Published:2019-09-29
  • Supported by:
    National Natural Science Foundation of China (61703341)

摘要: 飞机飞行控制系统机电作动器(EMA)的渐变性故障很难准确预判,若不能及早发现而任其发展就会影响到飞机的飞行安全性。针对EMA的渐变性故障,提出一种基于动态小波神经网络(DWNN)的故障诊断方法。首先,利用EMA在电机电枢绕组匝间短路、传动装置丝杆和滚珠磨损等多种渐变性故障状态下的运行数据来训练DWNN故障诊断模型;然后,利用训练好的DWNN模型对EMA渐变性故障进行诊断。创新之处在于DWNN模型利用小波分解算法去除了传感器测量信号中高频分量的影响,利用反馈神经网络的记忆能力融合了过去输入的信息和过去预测的信息,提高了对EMA渐变性故障诊断的准确性。通过对某型EMA进行故障诊断实验,仿真结果表明所提出的DWNN方法可以实现对EMA部件渐变性故障的准确诊断。

关键词: 机电作动器(EMA), 渐变性, 故障, 诊断, 动态小波神经网络(DWNN)

Abstract: It is difficult to accurately predict the gradual failure of electromechanical actuator (EMA) in an aircraft flight control system. The flight safety of an aircraft will be affected by these faults if they are not detected in early stage. A fault diagnosis method based on dynamic wavelet neural network (DWNN) is proposed to diagnose the gradual fault of EMA. This method trains the DWNN fault diagnosis model in offline step by using EMA's operation data of gradual faults, such as interturn short circuit of armature winding, screw and ball wear of transmission device, and then the trained DWNN model is used to diagnose the gradual faults of EMA online. The innovations of the research are as follows:First, the influence of the high-frequency components in the sensor measurement signals is removed using wavelet decomposition algorithm in the DWNN model; Second, the information input in the past and the information predicted in the past are integrated using the memory ability of the feedback neural network, so the accuracy of EMA gradual fault diagnosis is improved. The simulation results obtained from tests on a specific EMA show that the proposed DWNN method can accurately diagnose the gradual fault of EMA components.

Key words: electromechanical actuator (EMA), gradual, fault, diagnosis, dynamic wavelet neural network (DWNN)

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发