基于LSSVM的惯性/卫星紧组合导航系统缓变故障检测方法

Approach for detection of slowly growing errors in INS/GNSS tightly-coupled system based on LS-SVM

针对传统缓变故障检测方法——自主完好性检测外推法在惯性/卫星紧组合导航系统缓变故障检测中检测延时较大的缺点,设计了一种基于最小二乘支持向量机的缓变故障检测方法。该方法采用最小二乘支持向量机回归预测滤波器新息构成新的检验统计量,结合自主完好性检测外推法进行缓变故障检测。理论分析表明,该方法能够有效降低导航系统滤波器故障跟踪的影响,减小故障检测延迟时间。仿真结果表明,在检测缓变故障时,基于最小二乘支持向量机的故障检测算法比自主完好性外推法检测延时缩短了1.3倍以上。

Aiming at improving the instantaneity of detecting slowly growing errors with the autonomous integrity monitored extrapolation（AIME） method in the INS/GNSS tightly-coupled navigation system,this paper presented a new method based on least squares support vector machine（LS-SVM） combined with AIME.The method replaces the innovation of Kalman filter by the forecasted innovation of the Least squares support vector machine regression,which is used for the test statistics.The new method presented in this paper could reduce the influence of fault tracking caused by Kalman filter,so the real-time ability of slowly growing errors detection can be improved.Simulation results show that,the delay time of slowly growing errors detection is reduced by 1.3 times.The method based on Least squares support vector machine has better detection ability than traditional autonomous integrity monitored extrapolation method when detecting slowly growing errors.