Method for predictive maintenance of satellites

1. A method comprising:

during a first time period:

accessing a first set of telemetry data for a first population of satellites;

training a first model to predict failures for the first population of satellites within a first time window based on the first set of telemetry data; and

training a second model to predict failures for the first population of satellites within a second time window, shorter than the first time window, based on the first set of telemetry data; and

during a second time period:

accessing a second set of telemetry data broadcast by a first satellite;

predicting a first probability of failure of the first satellite within the first time window based on the first model and the second set of telemetry data;

predicting a second probability of failure of the first satellite within the second time window based on the second model and the second set of telemetry data; and

in response to the first probability of failure exceeding a first threshold probability and the second probability of failure falling below the first threshold probability, triggering prioritizing of telemetry data transmission from the first satellite.

2. The method of claim 1 , wherein triggering prioritizing of telemetry data transmission from the first satellite comprises:

in response to the first probability of failure exceeding a first threshold probability and the second probability of failure falling below the first threshold probability:

generating a notification for an operator to alter a set of operations to prioritize transmission of telemetry data from the first satellite; and

transmitting the notification to the operator.

3. The method of claim 2 :

further comprising, during the first time period:

training a third model to generate a first anomaly score characterizing the first set of telemetry data within the first time window; and

further comprising, during the second time period:

in response to the first anomaly score exceeding a first threshold anomaly score, characterizing the second set of telemetry data for the first time window as anomalous; and

wherein generating the notification comprises generating the notification in response to characterizing the second set of telemetry data for the first time window as anomalous.

4. The method of claim 3 , wherein generating the notification in response to characterizing the second set of telemetry data for the first time window as anomalous comprises generating the notification to prompt the operator to initiate investigation mode conditions.

5. The method of claim 3 :

further comprising, during the first time period:

training a fourth model to generate a second anomaly score characterizing the first timeseries telemetry data within the second time window; and

further comprising, during a third time period:

assigning a first influence value to a first feature associated with a first subsystem of the first satellite contributing to a failure of the first satellite based on the first and third model;

assigning a second influence value to a second feature associated with a second subsystem of the first satellite contributing to the failure of the first satellite based on the first and third model, the second influence value lower than the first influence value based on the second and fourth model; and

ranking the first and second feature according to the first influence value and the second influence value.

6. The method of claim 5 , wherein ranking the first and second features comprises ranking the first and second features based on values associated with each of the first and second features, the values assigned by a feature model.

7. The method of claim 6 , wherein ranking the first and second features comprises ranking the first and second features for investigation based on Shapley values associated with each of the first and second features, the Shapley values assigned by a feature model.

8. The method of claim 5 :

wherein generating the notification for the operator to alter the set of operations comprises generating the notification to respond to the first subsystem of the first satellite; and

further comprising generating a second notification to respond to the second subsystem of the first satellite.

9. The method of claim 2 , further comprising, during the second time period:

in response to the first probability of failure falling below the first threshold probability and the second probability of failure falling below the first threshold probability:

generating the notification for the operator to alter the set of operations to deprioritize manual monitoring of the first satellite; and

transmitting the notification to the operator.

10. The method of claim 2 , further comprising, during the second time period for the first satellite:

in response to the first probability of failure exceeding the first threshold probability and the second probability of failure exceeding the first threshold probability:

generating the notification for the operator to alter the set of operations to:

retrieve a script for transferring the first satellite to a safe state; and

prompt upload of the script to the first satellite; and

transmitting the notification to an operator.

11. The method of claim 1 :

further comprising, defining a reference time specifying a start point for the first and second time windows based on a probability of failure exceeding the threshold probability for a duration of time; and

wherein accessing the second set of telemetry data broadcast by the first satellite comprises accessing the second set of telemetry data broadcast by the first satellite based on the reference time.

12. The method of claim 1 :

during the first time period:

wherein accessing the first set of telemetry data comprises accessing the first set of telemetry data and a first set of failure data for a first population of satellites;

wherein training the first model comprises training the first model to predict failures for the first population of satellites within the first time window based on the first set of telemetry data and the first set of failure data; and

wherein training the second model comprises training the second model to predict failures for the first population of satellites within the second time window, shorter than the first time window, based on the first set of telemetry data and the first set of failure data.

13. A method comprising:

during a first time period:

accessing a first set of telemetry data and a first set of failure data for a first population of satellites;

training a first model to predict a failure of a first satellite of the first population of satellites within a first time window based on the first telemetry data and the first failure data; and

training a second model to predict the failure within a second time window, shorter than the first time window, based on the first telemetry data and first failure data; and

during a second time period:

accessing a second set of telemetry data broadcast by a first satellite;

predicting a first probability of failure of the first satellite within the first time window based on the first model and the second set of telemetry data;

predicting a second probability of failure of the first satellite within the second time window based on the second model and the second set of telemetry data; and

in response to the first probability of failure falling below a first threshold probability and the second probability of failure exceeding the first threshold probability, triggering alteration of a set of operations to:

detect an external source of failure of the first satellite;

retrieve a script for transferring the first satellite to a safe state; and

upload the script to the first satellite.

14. The method of claim 13 :

during the first time period:

wherein accessing the first set of telemetry data and a first set of failure data comprises accessing a first set of historical timeseries telemetry data and a first set of historical timeseries failure data for a first population of satellites, the first set of historical timeseries telemetry data including labeled times stamps specifying instances when the first satellite entered into investigation mode conditions; and

during the second time period:

wherein accessing the second set of telemetry data comprises accessing a first set of timeseries telemetry data broadcast by the first satellite.

15. The method of claim 13 :

further comprising, during the first time period:

training a third model to predict a first anomaly score characterizing the first telemetry data and the first failure data within the first time window; and

further comprising, during the second time period for the first satellite:

in response to the first anomaly score falling below a first threshold anomaly score, characterizing the second set of telemetry data for the first time window as nominal.

16. The method of claim 13 , wherein triggering alteration of the set of operations comprises:

in response to the first probability of failure falling below a first threshold probability and the second probability of failure exceeding the first threshold probability:

generating a notification for an operator to alter a set of operations to:

detect an external source of failure of the first satellite;

retrieve a script for transferring the first satellite to a safe state; and

upload the script to the first satellite; and

transmitting the notification to the operator.

17. The method of claim 16 :

further comprising, during the first time period:

training a fourth model to predict a second anomaly score characterizing the first telemetry data and the first failure data within the second time window;

further comprising, during the second time period:

in response to the second anomaly score exceeding the first threshold anomaly score, characterizing the second set of telemetry data for the second time window as anomalous; and

wherein generating the notification for the operator to alter the set of operations comprises generating the notification characterizing the second set of telemetry data for the second time window as anomalous.

18. The method of claim 13 :

during the second time period:

wherein accessing the second set of telemetry data comprises accessing the second set of telemetry data broadcast by a first satellite at a first frequency; and

wherein in response to the first probability of failure falling below the first threshold probability and the second probability of failure exceeding the first threshold probability comprises accessing a second set of telemetry data broadcast by the first satellite at a second frequency higher than the first frequency.

19. A method comprising:

during a first time period:

accessing a first set of telemetry data and a first set of failure data for a first population of satellites;

training a first model to predict a failure of a first satellite of the first population of satellites within a first time window based on the first set of telemetry data and the first set of failure data; and

training a second model to predict the failure within a second time window, shorter than the first time window, based on the first set of telemetry data and the first set of failure data; and

during a second time period:

accessing a second set of telemetry data broadcast by a first satellite;

predicting a first probability of failure of the first satellite within the first time window based on the first model and the second set of telemetry data;

predicting a second probability of failure of the first satellite within the second time window based on the second model and the second set of telemetry data; and

in response to the first probability of failure exceeding a threshold probability and the second probability of failure exceeding the threshold probability, triggering alteration of a set of operations to:

retrieve a script for entering a safe state for the first satellite; and

upload the script to the first satellite.

20. The method of claim 19 , wherein triggering alteration of a set of operations comprises:

in response to the first probability of failure exceeding a threshold probability and the second probability of failure exceeding the threshold probability:

generating a notification for an operator to alter a set of operations to:

retrieve a script for entering a safe state for the first satellite; and

upload the script to the first satellite; and

transmitting the notification to the operator.