Science Coffee - Algorithms to Monitor Telemetry for Subtle Indications of Debris Impacts
1. More Info, Same Data:
Algorithms to Monitor Telemetry for
Subtle Indications of Debris Impacts
Anne Aryadne Bennett
Graduate Research Assistant, CCAR
Systems Engineer, Northrop Grumman
CU Advisor: Dr. Hanspeter Schaub
NASA Research Collaborator: Dr. Russell Carpenter
1
This work is supported by a
NASA Space Technology Research Fellowship
2. Agenda
• Background: hazardous non-trackable orbital debris
• Debris to small to track, large enough to harm spacecraft
• What is known, and what isn’t
• Limitations of model verification measurements
• Research: Use spacecraft as in situ debris sensor
• Effects of strike on spacecraft
• Algorithms to detect subtle features
• Architecture for exploring model output vs. results
• Applications: Space environment management
• Local-scale: anomaly detection, state of health monitoring
• Global-scale: contribute to knowledge of debris environment
• Q&A / Discussion
Source: Maclay & McKnight,
https://doi.org/10.1016/j.jsse.2020.11.002
3. Hazardous Non-Trackable Orbital Debris
• Fragmentation events produce clouds of small debris
as well as larger trackable debris pieces
• Space Surveillance Network tracks down to ~10 cm
in LEO, ~70 cm in GEO (<10% of hazardous debris)
• Debris 1 cm or smaller can cause mission-ending damage
• Sample return missions indicate 100s-1,000s of impacts
• LDEF, Hubble solar arrays, etc.
• Recent events
• Strikes causing anomalies / shed debris, continued operations:
Sentinel-1A, WorldView-2, MMS constellation
• Satellites lost in abrupt anomalies, debris plausible:
Intelsat-29e, Telkom-1, AMOS-5
NASA test of 2
mm Aluminum
sphere impacting
tank at 7.2 km/s
Source: NASA
Engineering
and Safety
Center, 2019
presentation
Untrackable strikes can be benign, or can be catastrophic
Debris Strikes on Sentinel-1A
and Hubble Solar Arrays
Hubble clear hole: 2.5 mm
Source: ESA website Source: ESA website
Tracked Previously
Tracked with
Space Fence
Potentially
Hazardous
( > 1 cm)
Source: Aerospace Corp. STM paper, September 2020 (Sorge et. al)
4. Challenges in Modeling Untrackable Debris
• Model validation measurements are limited
• Radar in LEO: measures radar cross section
Sample returns: low orbits only, occasional data, small impacts
Optical in GEO: brightness, >30 cm
• 2017 NESC report on predicted vs. reported anomalies
• Found significant inconsistencies (vs. ORDEM3.0)
• Far fewer anomalies reported than predicted across multiple studies
• One study: 7 perturbations reported, 24-160 predicted
• Recommendation: collect additional data to validate models
• Typical spacecraft operations:
• Anomalous behavior => full review (Significant time + $$$)
• No anomalous behavior => not assessed/investigated/cataloged?
• Indications of debris strike often subtle
• New space: more satellites, more autonomy
Can subtle strike effects be sensed in spacecraft telemetry?
‘The data gap’
Another serious
data gap
?????
Source: NASA website
5. Identify Strikes in Spacecraft Telemetry
5
Telemetry from Sentinel-1A
strike shows spacecraft rate
increase and correction
Develop techniques to identify minor debris
strikes in spacecraft telemetry
Debris impact on
solar array imparts
rotation
5
Source: Acta Astronautica 137 (2017) 434-443
∆𝐻
• Debris strike imparts rotation
• Corrected by attitude control for most satellites
• Abrupt change in angular momentum of satellite
• Research Goal:
• Outcomes
• Use any active spacecraft as an in situ debris sensor
• Acquire data to validate models and improve risk assessments
• Monitor spacecraft state-of-health
Spacecraft Dynamics
Model
6. 0 10 20 30 40 50 60
datapoint
-5
0
5
10
15
Rate
Wavelet
10-5
0 50 100 150
Sim Time [s]
-1
0
1
2
Rate
Telemetry
10
-4
0 50 100 150
Sim Time [s]
0
1
2
3
4
5
Filter
Output
10-8
0 10 20 30 40 50 60
datapoint
-5
0
5
10
15
Rate
Wavelet
10-5
0 50 100 150
Sim Time [s]
-1
0
1
2
Rate
Telemetry
10-4
0 50 100 150
Sim Time [s]
0
1
2
3
4
5
Filter
Output
10
-8
0 10 20 30 40 50 60
datapoint
-5
0
5
10
15
Rate
Wavelet
10
-5
0 50 100 150
Sim Time [s]
-1
0
1
2
Rate
Telemetry
10-4
0 50 100 150
Sim Time [s]
0
1
2
3
4
5
Filter
Output
10-8
Strike Detection Algorithms
Strike at 50 s
Find ‘signal’
produced by S/C
response to strike
0 10 20 30 40 50 60
datapoint
-5
0
5
10
15
Rate
Wavelet
10
-5
0 50 100 150
Sim Time [s]
-1
0
1
2
Rate
Telemetry
10-4
0 50 100 150
Sim Time [s]
0
1
2
3
4
5
Filter
Output
10-8
• Digital signal processing: Matched filter
applied to rate and attitude telemetry
• Change detection algorithms:
applied to angular momentum
telemetry
Wavelet
Noisy data
obfuscates
signal
Filter output
shows clear
spike
0 200 400 600 800
Time [s]
13.2
13.4
13.6
13.8
S/C
Momentum
[N.m.s]
Momentum Data, 10 mg strike at t = 300 s
Raw Data
0 200 400 600 800
Time [s]
13.2
13.4
13.6
13.8
S/C
Momentum
[N.m.s]
Momentum Data, 10 mg strike at t = 300 s
Raw Data
0 200 400 600 800
Time [s]
13.2
13.4
13.6
13.8
S/C
Momentum
[N.m.s]
Momentum Data Overlaid with Distribution Parameters
Data
Expected
Changed
+2
+3
200 400 600 800 1000
Time [s]
0
5
10
15
20
Sum
x-axis
y-axis
z-axis
ts t rik e
200 400 600 800 1000
Time [s]
0
20
40
60
80
100
CUSUM
x-axis
y-axis
z-axis
ts t rik e
10 mg strike
15 mg
20 mg
Inertial angular momentum
changes abruptly when
strike occurs
Sequential probability ratio test
applied in sliding window filter
accentuates change dramatically
7. Preliminary Results: SDO
• NASA Solar Dynamics Observatory
• GEO spacecraft, stares at sun
• Stubby solar arrays, gimballed HGAs track ground
• Run CUSUM SPRT on body-frame angular momentum
• Filter to remove periodic signatures (adds uncertainty)
• Blank thruster firings, slews
• Frequent small blips, some larger features
Example
small blips
Example abrupt momentum transfer
(sources besides strike plausible)
Day of Year in 2016
Day of Year in 2018
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
0 50 100 150 200 250 300 350
0 50 100 150 200 250 300 350
0
10
20
30
40
0
10
20
30
0.0
1.5
1.0
0.5
Number
of
Strikes
Number
of
Strikes
Scaled
Flux
Measurements of sporadic annual
variations in micrometeoroid flux
Tallying small blips per day
Some correlation? Still skeptical,
models predict far fewer strikes
than this (i.e, a few per year)
Source: NASA website
8. Preliminary Results: MMS
• Magnetospheric Multiscale (MMS)
• 4 spacecraft in formation, highly elliptical orbit
• Spin-stabilized, four 60 m wirebooms
• Angular momentum oscillates after thruster firing,
changes at perigee, occasional ‘mystery torques’
near apogee, known strike is detected
• Demonstrates additional satellite-as-a-sensor use case
Telemetry from
debris strike
shows
wireboom
oscillation
9. Comparing Results to Models
• NESC report identified contributors to uncertainties
• ORDEM: Limited measurement data, uncertainty in particle shape/mass
• Bumper: Failure criteria, shield properties
• Anomalies: Reluctance/inability to share data, misattribution, small data set
Source: NESC
presentation Oct 2017
Compare
Prediction
via MINDSS
Data from telemetry
• Ongoing development: Model for In-orbit
Debris Strikes on Satellites (MINDSS)
• Compare perturbations instead of anomalies
• Larger data set, less reluctance to share
• No damage/failure prediction required
• Agile architecture capable of trading model
parameters
10. MINDSS (In Development)
• Architecture to predict perturbations
• Agile to trade parameters and leverage various
models
• Examine variability in perturbation prediction as
function of several variables
Pull flux data from
environmental models
100
102
104
Change in Angular Momentum [mNms]
10-2
100
102
104
Change
in
Semi-Major
Axis
[m]
Change in Angular Momentum vs Change in Orbit
10-2
Characteristic Length [m]
100
102
104
Change
in
Angular
Momentum
[mNms]
Change in Angular Momentum vs Characteristic Length
Not
detected
Detected
by both
Detected
by DH
only
Detected
by orbit
change
only
Developed by Libby Moore,
undergraduate researcher
Monte-Carlo strikes vs.
simple spacecraft model
Compare perturbations to
detectable thresholds
Source: ORDEM
user guide
Incorporate results from
ongoing/recent research
Source: Williamsen,
Procedia Engineering
204 (2017) 500-507
Source: Murray et al,
IOC 2019 paper 6135
11. Potential Applications: Anomaly Screening and Resolution
• State of health monitoring, anomaly attribution/response
• Quicker response/resolution to immediate strike-induced anomalies
• Monitor for anomalies that may not manifest immediately
• Example: startracker baffle strike leads to straylight issues later
• Improve operator knowledge of local debris environment
• Especially constellations
• Concrete method for implementing ‘Satellite as a Sensor’ concepts
• Great fit for constellations and product line satellites
• i.e., commercial GEO bus product lines
• More satellites=more sensors, more autonomy=better monitoring required
Source: DARPA website
12. Potential Applications: Wider Adoption, Broader Impact
• Use debris strike data to inform debris models
• Returns data correlated to debris mass
• Data collection in GEO, ‘data gap’
• Reporting perturbations may be more
palatable than reporting anomalies
• More events produce larger, more useful dataset
• Collect data in repository
• US Department of Commerce has shown interest
• Open architecture data repository for SSA data
• Ought to be trans-national effort
• Validated and/or improved debris risk assessments
• Assess risks more accurately
• Motivate appropriate behavior from New Space actors
• Data promotes improved space environment management
Source: Williamsen et al
https://doi.org/10.2514/1.A34765
more data,
fewer uncertainties
Source: NESC
presentation Oct 2017
Source: NASA website
Measure
critical regions
13. Key Takeaways
• Hazardous non-trackable debris is an ongoing challenge to safe space operations
• Modeling risk is challenging, models suffer from limited data sources
• Strikes can range from benign to mission-ending
• Change detection algorithms can highlight unexpected perturbations
in quiescent telemetry
• Inertial angular momentum is theoretically quiescent in absence of external torques
• Running algorithms on spacecraft data turns up unexpected features
• Con: obfuscate debris strikes
• Pro: may be useful for monitoring general state of health
• Techniques have utility on both local and global scales
• Monitor individual spacecraft for anomaly resolution
• Contribute perturbations to database to validate and/or improve debris risk
assessment methods
Strikes too small for
measurable effect
Strikes which do not
affect nominal operation,
but do affect spacecraft
Strikes which disrupt
nominal operation, but
are recoverable
Strikes which disable
spacecraft
Strikes causing severe
fragmentation events
Classes of Debris Events
