引用本文
  •    [点击复制]
  •    [点击复制]
【打印本页】 【在线阅读全文】【下载PDF全文】 查看/发表评论下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 21次   下载 11 本文二维码信息
码上扫一扫!
广义回归神经网络在冗余捷联惯导故障诊断中的应用研究
乔鹏超,孙湘钰,罗广地
0
(哈尔滨工程大学自动化学院,哈尔滨 150001)
摘要:
针对冗余捷联惯导的故障诊断问题,研究提出了一种基于广义回归神经网络的故障诊断方法。该方法在传感器输出数学模型未知的情况下,仅通过传感器之间的冗余关系,利用传感器正常工作时的测量值和改进的神经网络估计输出值生成残差进行故障诊断。仿真试验表明,利用神经网络补偿产生的残差可以检测到未补偿时的故障。该方法不仅可以检测到单故障,还对多传感器同时发生故障具有一定的检测能力。
关键词:  冗余捷联惯导  广义回归神经网络  传感器  故障诊断  故障检测
DOI:
基金项目:国家自然科学基金(61633008, 61773132);黑龙江省杰出青年基金(JJ2018JQ0059);中央高校基本科研业务费专项基金(HEUCFP201768)
Fault Diagnosis of Redundant Strapdown Inertial Navigation System Based on Generalized Regression Neural Network
QIAO Peng-chao,SUN Xiang-yu,LUO Guang-di
(College of Automation, Harbin Engineering University, Harbin 150001, China)
Abstract:
Aiming at the fault diagnosis of redundant strapdown inertial navigation system, a fault diagnosis method based on the generalized regression neural network is proposed without the prior knowledge of the sensor output mathematical model. The proposed method only uses the redundancy relationship between sensors, and generates residual errors to diagnose faults by using the sensors measurements in normal operation and the output values estimated by the improved neural network. The simulation results show that the residual generated by neural network compensation can detect faults without compensation. This method can detect not only single faults but also multiple sensors faults simultaneously.
Key words:  Redundant SINS  Generalized regression neural network  Sensors  Fault diagnosis  Fault detection

用微信扫一扫

用微信扫一扫