GSGC ISS Engineering Outreach Projects

1
GSGC ISS Engineering Outreach Projects

• ISS Payload Rack Data Trending Analysis
• E. Armanios, PI
• Richard Cross, GRA

SE SG Regional Meeting, Lexington
Kentucky November 11, 2006
2
One of Two GSGC ISS Engineering outreach projects
funded

• Quantifying Uncertainty in ISS Thermal Model
• Dr. Andrew Makeev, PI
• In collaboration with JSFC

3

• Debbie V. Nguy February 2, 2005
• Johnson Space Center, Houston
• RELEASE J05-002
• NASA PLANS UNIQUE SPACE STATION PARTNERSHIP WITH
  SEVEN UNIVERSITIES

• The grant opportunity was announced by NASA
  through the 2004 Aerospace Workforce Development
  Competition
• Award to seven universities Georgia Institute
  of Technology, Massachusetts Institute of
  Technology, Montana State University, Purdue
  University, University of Alabama-Huntsville,
  University of Mississippi and University of
  Wyoming.

4
Challenge

• The five EXPRESS racks
• EXpedite the PRocessing of Experiments to ISS
• Provide onboard experiments with
• Power, data connectivity, temperature control
• Sensors on these racks return a large amount of
  real time data
• Could the Storage size be reduced?
• Could it be used to assess health of experimental
  packages and predict failure?

5
EXPRESS Rack
6
EXPRESS Rack
7
Objectives

• Reduce data storage requirement by summarizing
  mean trends and scatter
• Data summary must retain enough information for
  analysis purposes
• Analyze EXPRESS rack data
• Identify departures from nominal sensor readings
• Locate correlations between sensor channels

8
Summary Approach

• Summary Methods
• Mean time-series are summarized with multi-layer
  perceptron neural networks trained with Bayesian
  regularization.
• Bayesian formulation allows the calculation of
  confidence intervals for sensor output and
  prevents over-fitting of data
• Scatter information retained by storing Hessian
  matrix from training
• This matrix can be used to estimate the variance
  in the sensor data

9
Analysis Overview

• Analysis Methods
• Deviations from nominal operation detected
  through calculation of confidence bounds on mean
  time-series
• Feature identification by testing the
  significance of the rate of change
• Neural network model facilitates calculation of
  correlations between various sensor outputs

10
Sample Analysis Raw Data
EXPRESS Rack 1 Temperature data from sensor
TS13 Data covers day 350 of 2004 to day 20 of 2005
11
1st Derivative Significance Analysis
12
Simultaneous Significance Comparison

• Timing of significant events can be compared for
  various sensor readings
• Consider Lab Thermal System

Possibly significant system-wide event
13
Sensor Correlation Findings

• Strong correlations were found in the temperature
  readings
• This result is expected since the various parts
  of the rack are thermally connected through the
  cooling system
• Small correlations between the rack temperatures
  and lab temperatures were found
• Some small correlations were found between power
  system (ELC, EMU, AFC2, Space currents, RFCA
  flow) and lab and rack thermal systems

14
Useful Results from Summary

• Storing the mean trends requires negligible disk
  space
• 49.2K vs. 14.7MB
• 0.3 the original database size
• Storing scatter information (Hessian matrix from
  training) requires more space
• Still a big reduction, 851K vs. 14.7MB
• This is an upper bound on required storage space
• Networks could be iteratively trained to identify
  smallest possible network that correctly captures
  data

15
Conclusions

• EXPRESS Rack data can be effectively compressed
  and stored
• Numerous analyses are possible
• Identification of significant deviations from
  normal operation
• Identification of significant features
• System-wide event identification
• Correlation analyses

16
Acknowledgements

• Mr. Rick Cissom, Payload Operations Project
  Manager, NASA MSFC
• Dr. Craig Cruzen, ISS Payload Operations
  Director, NASA MSFC

17
(No Transcript)
18
Feedback
We found our discussions about the research that
you and your colleagues have been performing on
ISS Payload Rack Data Trending to be informative.
We were impressed with the research your staff
has provided in the NASA Engineering Outreach
Program We look forward to working with you and
your staff. In addition, if there any assistance
that our organization can provide regarding
educational outreach and fostering your students
interest in Americas Space Program, please do
not hesitate to contact us .
19
Spin-off

• Initiation of GSGC- MSFC Mentoring
• Senior Design Project in Space Systems
• Briefings on MSFC projects
• Review of student team design projects

20
WednesdayMarch 29, 20063-4p.m.Montgomery
Knight Bldg.Design Lab, Rm. 442The Robotic
Lunar Exploration Program (RLEP)Mr. Raymond
EcholsRobotic Lunar Exploration Program - 2,
Lander Mission Mission Operations LeadMarshall
Space Flight Center
21
International Space Station (ISS) Operations and
ResearchDr. Craig CruzenInternational Space
Station ProgramPayload Operations
DirectorMarshall Space Flight Center
22
(No Transcript)
23
Follow-up

• Dr. Craig Cruzen Mr. Raymond Echols
• Space Systems Senior Design Project review panel
• ESMD Senior design project advisors