基于在线增量小波LS-SVM的垂直陀螺仪残差故障检测研究  

Fault Detection Research of online incremental Wavelet LS-SVM Based on Free Gyroscope Model Residual

作　　者：史岩[1,2] 李小民[1,2] 连光耀 

机构地区：[1]军械工程学院光学与电子工程系,石家庄050003 [2]军械技术研究所,石家庄050003

出　　处：《计算机测量与控制》2013年第1期14-17,26,共5页Computer Measurement &Control

摘　　要：垂直陀螺仪是无人机重要的飞行姿态传感器,其在飞行过程中实时获取无人机的飞行姿态信息,因而其故障检测对在线性有着很高的要求;最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)相比于支持向量机的具有训练速度快、计算复杂度和需要内存少的特点,且能够扩展为自回归的形式来处理动态问题;因此文章采用基于在线增量小波LS-SVM建立无人机垂直陀螺仪动态模型,实时获得实际值与模型预测值之间的残差,并依据残差对陀螺仪进行在线故障检测;实验结果表明,该方法能够对陀螺仪实现快速精确的在线检测。Free Gyroscope is an important flight attitude sensor, and it obtains the real--time flight attitude information which dynami cally varies in the flight, so it is exigent of online failure detection. Least Squares Support Vector Machine （LS--SVM） has the characters of fast training speed, less calculation complexity and less required memory with comparison to Support Vector Machine （SVM）, and it can ex tend auto regression format to handle the dynamic problem. In this paper, the online incremental LS--SVM is utilized to build the dynamic model of UAV free gyroscope in order to obtain the real time residual between real value and pre estimating value, and then fault detection can be done according to the residuals. The experimental results testify that this method can accomplish fast and accuracy online diagnosis.

关 键 词：无人机 垂直陀螺仪 动态建模 最小二乘支持向量机 故障检测 在线增量学习 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]