[go: up one dir, main page]
Scribd
Upload
43 views14 pages

A Data-Driven Approach To Cubesat Health Monitoring: Presented By: Serbinder Singh

This document presents a data-driven approach to health monitoring for CubeSats called Inductive Monitoring System (IMS). IMS uses machine learning models trained on archived telemetry data to monitor spacecraft systems in real-time. It captures relationships between multiple sensors to identify anomalies without requiring deep system knowledge. The approach was implemented in the health monitoring system for the PolySat CubeSat program. Tests on archived and real-time data successfully identified temperature and power anomalies. The system performed efficiently with low CPU usage. Future flight tests are planned to further evaluate the approach.

Uploaded by

Anonymous mE6MEje0
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
43 views14 pages

A Data-Driven Approach To Cubesat Health Monitoring: Presented By: Serbinder Singh

This document presents a data-driven approach to health monitoring for CubeSats called Inductive Monitoring System (IMS). IMS uses machine learning models trained on archived telemetry data to monitor spacecraft systems in real-time. It captures relationships between multiple sensors to identify anomalies without requiring deep system knowledge. The approach was implemented in the health monitoring system for the PolySat CubeSat program. Tests on archived and real-time data successfully identified temperature and power anomalies. The system performed efficiently with low CPU usage. Future flight tests are planned to further evaluate the approach.

Uploaded by

Anonymous mE6MEje0
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

A DATA-DRIVEN APPROACH

TO CUBESAT HEALTH
MONITORING
Presented By: Serbinder Singh
INTRODUCTION
• Spacecraft health monitoring is very important
• Verify proper functionality
• Involves looking at sensor readings and telemetry
• Can be a challenging task
• Becomes more difficult as complexity of spacecraft increases
• Traditional techniques become impractical
• Even more challenging for CubeSat’s
DATA-DRIVEN TECHNIQUES
• More simple and manageable
• Uses data mining and machine learning
• Archived data used to create models
• Models compared against real time data
• Many benefits over traditional approaches
• Deep system knowledge not necessary
• Relationship between multiple sensors captured
• Faster and more autonomous
INDUCTIVE HEALTH MONITORING (IMS)
• Created for use in various programs and missions in NASA
• Models relationship between a set of sensors in time-series data
as clusters
• Vector used to hold values of several related system parameters
• Each cluster defines a unique nominal state
• Group of clusters create a knowledge base
• Knowledge base produces deviation value
Real Time
Archived Data Telemetry

Learning Monitoring
Learning Module Knowledge Deviation Score
Params Base Module
IMS DATA
• Nominal archived data required
• Free of errors
• Comprehensive
• Easily updatable
POLYSAT HEALTH MONITORING SYSTEM
OVERVIEW
• Designed around IMS
learning/monitoring phases
• Learning done on ground
• Monitoring on spacecraft

Learning Models:
• Archived data from IPEX mission
• 50,000 data points
• Over 150 features
• Two main models created
• Temperature, Power
• Monitoring implemented in flight
software
RESULTS
TESTS ON ARCHIVED DATA
• IPEX data with known anomalies used
• Goal to identify anomalous sensors
• Temperature and Power models tested
• Run separately and deviation scores analyzed
TEMPERATURE MODEL
 Temperature issues with
side panels
 Positive X panel got very
hot
 Thermal conductivity
improved
REAL-TIME CUBESAT TESTING
• Tested monitoring system on the flight software
• Ran on test unit for ISX mission
• Side panels not assembled
• Used same temperature and power models

Important Results:
• Not all parameters have similar behavior mission to mission
• Importance of updating model
OTHER TESTS
• Fault Injection
• Simulated error injected into ISX test unit
• System successfully detected error
• Resource Usage
• System tested with relatively large models
• Deviation value calculated in less than 60 ms
• Very low CPU usage, and acceptable memory usage
CONCLUSIONS
• Application of IMS has great potential
• Archived data can have a purpose
• As more data is collected, the better the models become
• Our health monitoring system provided great results
• Identified anomalies in archived data and real-time tests
• Performed efficiently
• Flight tests to be performed on future missions
QUESTIONS?

You might also like