基于回归分析的航天器异常检测方法

Spacecraft anomaly detection method based on regression analysis

针对某在轨航天器电源温控系统典型故障案例,提出基于Pearson相关系数回归分析的航天器多元遥测数据异常检测方法。首先,对相关遥测数据进行特征提取;其次,对正常状态下的遥测数据进行拟合建模,并用该模型检测后续遥测数据;最后,采用基于Pearson相关系数的回归分析方法对实测数据进行仿真分析。结果表明,针对航天器首次发生的异常,该方法能从多维遥测数据中快速提取与异常相关的特征遥测参数,可有效建立遥测参数预警模型及检测遥测数据的异变趋势,并能够推广应用于其他航天器。

Aiming at a typical fault case of a spacecraft power supply temperature control system,an anomaly detection method based on Pearson correlation coefficient regression analysis is proposed.Firstly,the feature extraction of relevant telemetry data is carried out.Secondly,the normal telemetry data are fitted and modeled,and the subsequent telemetry data are detected by the model.Finally,the regression analysis method based on Pearson correlation coefficient is used to simulate the measured data.The results show that the method can quickly extract the abnormal phase from the multi-dimensional telemetry data for unknown spacecraft anomalies,it can be applied to other spacecraft.