TELEMETRY PARAMETER PERIOD-BASED ANOMALY DETECTION

It is popular for a satellite control center to use Limit-Check algorithm to analyses telemetry parameters down from on-orbit satellites to detect any anomaly. Limit-check algorithm is simple and effective to deal with static telemetry parameter but can't find an anomaly whose telemetry parameter's value is still within the given limit. This paper presents a novel method which takes advantage of the dynamic and periodic characteristic of telemetry parameters to solve this problem. Using periodogram spectral estimation to get the cycle of a telemetry parameter from an on-board unit, we find a regularity that a parameter's value of each cycle is almost same or very close when the unit is in good condition. If the regularity is broken it means something wrong. Therefore, we establish an auto-regressive moving average (ARMA) model for the data sampled periodically from a raw telemetry parameter. We use the model to predicate what the next parameter's value should be and use it to compare with the actual measured value to find any anomaly. To verify the method's validation, we use it to analyze the telemetry parameters downloaded from an China on-orbit XX-1 satellite in 2012 and immediately find an anomaly happened in its solar panel rotational unit. Any anomaly like this whose telemetry parameter's value is within the given limit can't be detected before only by Limit-check algorithm. The novel method is more sensitive to find any subtle anomaly and can be used as a complement to Limit-check algorithm to guarantee on-orbit satellite's health.