using generic, telemetry prognostic algorithms for … · 2010-07-13 · launch vehicle failures...
TRANSCRIPT
USING GENERIC, TELEMETRY
PROGNOSTIC ALGORITHMS
FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Len Losik, Ph.D
History of Equipment Data
10
/24
/200
9
2
1812Schilling
1845
Takobi
Konstantinov and Pouli
1874Olland
1906Seismic Recording
1946Missile Technology
1947Upper Atmosphere Monitoring
1957Digital Telemetry
1930
Radiosonde
Data Collection Monitoring/Diagnostics
1954Atlas Missile
Titan Missile
1955
Thor/Delta Missile
1964
INTELSAT/COMSAT
PANAMSAT
Commercial Space
Prognostics
1958NASA/JPL/DSN
1969AFCSN
Marconi Wireless Communications
1907
Titan Rocket1961
1975Atlas Rocket
Delta Rocket1962
1985
AFCSOC
1986
TDRSS
1978Jet Aircraft
1953
GPS
1978
Strip Chart Recorder1958
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
What are Prognostic Algorithms?
Pro-active diagnostics
Active reasoning
Proprietary algorithms for illustrating the early
signs of premature aging/failure in normal
appearing data from fully functional equipment
Used in a forensic analysis to identifying the early signs of (premature)
aging/failure in equipment using time series data
Prognostics is a discipline that acknowledges equipment/product failures do not
have the Markov property (instantaneous and random) and are predictable and
preventable and includes:
3
Allows measuring reliability invasively rather than
estimating reliability to quantify random behavior
Its invasive and requires integrating/adding/using
embedded measurements in products (telemetry) for
diagnostic, prognostic and prednostic analysis
Allows the production of equipment/products with
near-perfect reliability
Prognostics
DiagnosticsPrednostics
Relationship Between Diagnostics,
Prognostics and Prednostics
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Generic Failure Analysis Service is Available from the Evolution in Reliability Engineering Analysis: Reliability engineering analysis confirms FA’s 28 years of
research concluding that piece-parts/assemblies do not fail quickly but fail as a result of long-term, identifiable deterministic behavior
Using data-driven algorithms to identify equipment that will fail in the future, equipment and vehicle producers can increase reliability to over 99%, stopping infant mortality failures on spacecraft, satellites, missiles & launch vehicles Eliminates (reduce the cost) the need for space
launch & first year insurance
100% accurate for equipment that has telemetry
available Will encourage the use of telemetry on all equipment including single
function such as EED’s, pressure switches etc. 4
Detect
Diagnostic Process
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Launch Vehicle Failures and Satellite Infant
Mortality Failure Liability
Liability is currently on:
Commercial space vehicle insurance providers who are asking
builders to accept some liability
American tax payers
Why no liability for vehicle builders that produce
catastrophic failures?
As long as they used their “best effort” liability ends at delivery
Using prognostic algorithms, the equipment that is going to
fail and has already failed can be identified during test
Vehicle builders can measure equipment reliability
10
/24
/200
9
5
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Prognostic Algorithms
Are the next logical step in advancing
equipment and vehicle factory testing
which is excellent for identifying
equipment failures that have occurred
Identifies equipment that has failed
and is going to fail in the near future
Used across many industries (Computer, aircraft, nuclear
power, HVAC)
Developed from obtaining an intimacy with equipment’s
operational behavior seldom obtainable for satellites and
launch vehicles
Are used post failure to analyze factory data that
identifies equipment that should have been repaired or
replaced during acceptance testing but were missed
10
/24
/200
9
6
Prognostic Algorithm
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Failure Analysis’ Telemetry Prognostic Technology
Allows the identification of electronic
circuits and mechanical systems future
failures by prognosticians (engineers
trained to illustrate and identify the
early signs of premature aging/failure
Created and used on the Boeing/Air
Force Global Positioning System (GPS)
Block I satellites and Atlas F launch vehicle telemetry
Uses existing telemetry
Over 30 years of use on many satellites,
missiles and launch vehicles at the factory, in-orbit and on launch
pad including: GOES, INTELSAT, GPS, DMSP, ATLAS, TITAN,
TRIDENT, SUPERBIRD
7
Prognosis Process
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Failure Analysis’ Telemetry Prognostic Algorithms
Provides the capability to identify space equipment circuits
and electro-mechanical systems that failed for up to one year
Post failure analysis includes review of factory production
information including test telemetry to identify the early
signs of premature aging/failure always present during
factory test but missed by factory test personnel who:
Didn’t know what to look for
Didn’t know how to look for it
Didn’t know when to look for it
8
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Prognostic Technology/Algorithms Incorporates:
Includes a 4-dimension (time/amplitude/
frequency/phase) analysis to illustrate
behavior that indicates a failure is in
process
Identify the start of the early signs of
premature aging of electrical and
mechanical parts
Identify where in the failure process
the equipment was received
Identify the day-of-failure/remaining usable life
After an equipment failure has occurred, prognosticians
identify from past test data, when and where the early signs
of aging/failure began identifying
supplier/person/organization whose is liable for shipping
equipment that is going to fail within one year of use
10
/24
/200
9
9
4-D Telemetry Analysis
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Prognostics Diagnostics
Identifies equipment that has failed and will fail Identifies equipment that has failed or changed
performance
Uses active reasoning Used for passively monitoring information
Illustrates the information used by prognosticians
to predict equipment that will fail
Developed from ground test environment which
is not real-time
Actively monitor data to provide knowledge of
whether a failure has occurred, is occurring or
when a failure is likely to occur
Record data and look at it later, after the fact
response
All events are failure information until ruled out Events are recognized but no action is taken
Prognostician doesn’t just stand by and watch
failures occur, they stop them from occurring
Diagnostician waits for predefined error message
before taking action
Requires high skilled personnel – with in-depth
knowledge of what is being actively monitored
Allows lower skilled personnel - since data is
evaluated post event
Identifies system readiness for successful
deployment
Inadequate for meeting today’s customer
expectations
10
/24
/200
9
10
Prognostic/Diagnostic Comparison
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Active Reasoning Passive Reasoning
Evaluate symptoms continuously Evaluates symptoms after the fact
Uses fault reasoning Spurious symptoms misleads fault localization analysis.
Uses fidelity evaluation “how accurate are we?” Spurious symptoms are regarded as observational noise
Uses action selection “what should we do next?” Depends on monitoring agents to detect and report
abnormality using alarms or symptom events
Takes passively observed symptoms as input and returns
fault hypothesis as output.
After the fact search for root faults based on the observed
symptoms
Searching for the best fault explanation of the observed
symptoms.
Event-driven fault reasoning technique improves the
robustness of fault localization system by analyzing lost,
positive and spurious symptoms.
Assumes each event is a failure until ruled out
10
/24
/200
9
11
Active/Passive Reasoning Comparison
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Prognostic Algorithms and their use
10
/24
/200
9
12
Tool Box of Prognostic Algorithms
Equipment Factory
Satellite Factory
LV Factory
Launch Pad
Mission Control
Mission
Control(CCSDS)
Data Integration X X X X X XBaseline Analysis X X X X X XChange Analysis X X X X X
Comparison Analysis XData Mining X X X X X
Day of Failure X X X X X XDigital Processing X
Discrimination Analysis X X X X X XMathematical Modeling X X X X X X
Multi-variant Limit Analysis X X X X X XRate Change Analysis X X X X X
Remaining Usable Life X X X X X XStatistical Sampling X X X X X
State Change Analysis X X X X XSuper Impositioning X X X X X
Data Integration X X X X X XSuper Precision X X
Telemetry Authentication X Virtual Telemetry X X X X X X
Data Base Creation X
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Results Using Prognostic Algorithms
10
/24
/200
9
13
Months 12 Months
May be present during factory test
Will be present during factory test
2-D Presentation
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
EUVE Orbital Satellite Rate Gyro Failure
10
/24
/200
9
14
Finish
Start
mA
ps
Months
Months
Measurement Sample rate = 1/second
Catastrophic FailureIndication from prognostic algorithms
that unit will fail shortly
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
Flight History for Prediction Failure and Failure
Analysis on space Equipment Completed
Air Force GPS
NAVSTAR 1
NAVSTAR 2
NAVSTAR 3
NAVSTAR 4
NAVSTAR 5
NAVSTAR 6
NASA EUVE
NASA GOES I
SCC SUPERBIRD B
10
/24
/200
9
15
Rubidium Atomic Clocks
Batteries
Reaction wheel Assembly
TT&C Equipment
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
FA’s Failure Analysis Service is:
Independent of vehicle and vehicle manufacturer
Insensitive to amount of data available for analysis
Insensitive to resolution /accuracy of telemetry
Insensitive to noise in telemetry
Available today
Any failure
Any place
Any time
16
4-D Telemetry Analysis
3-D Telemetry Analysis
USING GENERIC TELEMETRY PROGNOSTIC
ALGORITHMS FOR LAUNCH VEHICLE AND
SATELLITE FAILURE ANALYSIS
FAILURE ANALYSIS
Creek bridge Business Complex
12 Glen Falls Circle
Salinas, CA 93906
www.failureanalysisco.com
Ph/Fax: (831) 443-4502
17