基于局部离群因子的列车卫星定位故障检测方法

A local outlier factor based fault detection method for train satellite-based positioning

基于卫星导航系统的定位技术已成为我国下一代列车运行控制系统中自主定位的重要方式。卫星信号在传播过程中易受到环境多源噪声的影响而导致定位解算性能下降,需要对潜在发生的故障进行检测以保证定位性能。针对一致性假设检验中观测新息不再服从先验高斯分布问题,提出一种基于局部离群因子的卫星定位故障检测方法。首先,基于正常运行环境中的滤波新息构建历史数据集,采用核密度估计方法获取检验阈值。在此基础上,根据特定的邻域值计算当前时刻观测新息的局部离群因子,通过度量其与历史数据集中邻域数据之间的局部密度进而判别是否发生故障。最后,采用西部铁路实测数据对所提算法进行实验验证。研究结果表明,在不同偏差阶跃故障和不同速率斜坡故障场景下,所提出方法的故障检测性能优于滤波新息故障检测和自主完好性监测外推法。在15m阶跃故障场景中,所提出的算法故障检测率分别提高了100%和62%,故障期间水平位置均方根误差降低了36.1%和18.5%。在0.5m/s斜坡故障场景中,故障检测时延分别缩短了20s和11s,故障检测率提高了40%和20%,水平位置均方根误差降低了28.6%和9%。基于局部离群因子的故障检测方法具有高检测、低时延的显著优势,打破了先验特定分布假设对于故障检测性能的约束,有效提高了定位系统的定位精度和可靠性。

Satellite navigation-based positioning technology has become an important way of achieving autonomous positioning in Chinese next-generation train operation control systems.Satellite signals affected by environmental multi-source noise during the propagation process would lead to the degradation of positioning calculation performance.It is necessary to detect potential faults to ensure positioning performance.Since the observation innovations no longer obey the prior Gaussian distribution in the consistency hypothesis test,this paper proposed a train satellite positioning fault detection method with the local outlier factor.First,a historical data set was constructed based on the filtered innovation in the normal operating environment,and the test threshold was obtained by kernel density estimation.In addition,the local outlier factor of the filter innovation at the current moment was calculated by the specific neighbor value and used to judge the fault by measuring the local density between the current data and the neighbor data of the historical data set.Finally,the proposed algorithm was verified by using the field data from Western Railway.The results show that the proposed method has a better fault detection capability than the filter innovation-based fault detection and autonomous integrity monitoring extrapolation methods in the scenarios of different step faults and ramp faults with different rates.In the 15 m step fault scenario,the fault detection rate of the proposed method is increased by 100%and 62%,and the root-mean-square error of horizontal position is reduced by 36.1%and 18.5%,respectively.In the 0.5 m/s ramp fault scenario,the time-delay of fault detection is shortened by 20 s and 11s,the fault detection rate is increased by 40%and 20%,and the root-mean-square error is reduced by 28.6%and 9%,respectively.The local outlier factor-based fault detection method has the obvious advantages of a lower detection delay and a higher detection rate,making the fault detection performance not regulated by the prior specific distribution assumption and improving positioning accuracy and reliability.