传感器故障下的航天器状态监测与故障识别

State Monitoring and Fault Identification for Spacecraft under Sensor Fault

针对航天器上传感器故障可能导致状态估计错误的问题,提出了一种基于证据理论的多传感器融合与故障识别方法。该方法采用证据距离与冲突质量的乘积来度量证据间的冲突大小。在证据的融合中先根据冲突大小计算每个证据的可信度,并根据可信度大小修正证据,再用D-S规则进行融合。该冲突度量方法综合了证据间目标一致性与整体差异性的影响,在具有较强的不确定性和不可靠性的情况下,比仅采用证据距离或冲突质量来度量证据间冲突的大小更加准确。将其应用于微小卫星电源系统的状态监测,实验结果表明该方法可以有效地监测系统状态并正确地识别故障传感器,准确度优于其它方法。

As sensor fault on spacecraft may lead to state estimation error, a new method based on evidence theory is proposed for multi-sensor fusion and fault sensor identification. A new metric defined as the product of two evidences＇ conflict mass and distance is introduced to measure two evidences＇ conflict. The evidences are firstly modified according to the reliable coefficients calculated from the conflict, and fused by use of the D-S rule. The new conflict metric integrates the influence of both goal congruence and overall discrepancy. When the evidences have strong uncertainty and unreliability, it is more appropriate than that only using corLflict mass or distance. The method is applied in state monitoring of a micro-satellite＇ s power subsystem, approving that it could identify the system state and fault sensor more accurately than other methods.