metrology & the consequences of bad measurement decisions
TRANSCRIPT
Metrology and the Consequences of Bad Measurement Decisions
Presented by Scott Mimbs
Corporate Sponsor:
NCSLI Virginia Section Meeting 1213
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 2
What you need to know about metrology
1. How does metrology affect me?• Measurement data affect your decisions
2. What do I need to look for?• Three essentials elements of “good” measurement data
3. What are the consequences of “bad” measurement-based decisions?
All measurement decisions have consequences…
The more critical the decision, the more critical the data. The more critical the data, the more critical the measurement. NASA Reference Publication 1342
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 3
Metrology and Measurement Data
Metrology is the “science of measurement and its application” (JCGM 200:2008).• This includes all theoretical and practical aspects of measurement
Measurements = DecisionsMeasurement data support decisions to…
• Establish research or investigative fact
• Establish scientific or legal fact
• Accept or reject a product
• Rework or complete a design
• Take corrective action or withhold it
• Continue or stop a process (including a space launch)
The objective of the design and control of measurement processes is to manage the risks taken in making decisions based on measurement data.
NASA Reference Publication 1342
Metrology — Calibration and Measurement Processes Guidelines
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 4
An undesirable situation or circumstance that has both a likelihood of occurring and a potentially negative consequence.
AS9100C
Defining Risk
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 5
International System of Units (SI)Bureau International des Poids Mesures
(BIPM)
Working StandardsPrimary Standards Lab or other Labs
Measurements by End-Item users
Primary National StandardNational Metrology Institute
Secondary StandardNational Metrology Institute or Primary
Standards Lab
CalibrationsIn-house or Commercial Calibration Labs
Highest level of Accuracy (scientific metrology)
Lowest level of Accuracy (end-item usage)
Errors cause Lowest level of Risk to final product or service
Errors cause Highest level of Risk to final product or service
Calibration requirements must ensure measuring and test equipment are acceptable and ready for use by end-users
Quality requirements must ensure the measurement will support the decision
A perfect instrument does not guarantee a “good” measurement…
Measurement Accuracy, Risk, and the Metrology chain
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 6
Measurement Risk through a Project Lifecycle
Measurement-related Risk: Risk of making incorrect measurement-based decisions• Based on measurement process limitations or process
mistakes
Product or System Risk: The negative consequence of an incorrect measurement-based decision • Quality or performance of end-products• Increased cost of measurements without added value
Concept & Prelim Design
Feasibility Analyses, Mockups, Prototypes,
TestingMeasurement
Research & Technology Development
Design, Fabrication, Integration & Test
Detailed Designs, Product fabrication, Op
Ready SystemMeasurement
Final Design, Production & Integration
Checkout, Operation & Maintenance
Functionality, Performance, UtilityMeasurement
Operation & Sustainment
Product / System Risk
Measurement Risk
In metrology, risk occurs as:
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 7
Range of possible values at 95% confidence
A B C
+ L
- L
Nominal
Probability of Incorrect Measurement Decisions
Measurement Uncertainty: The doubt that exists about a measurement’s result Every measurement—even the most careful—always has a margin of doubt Uncertainty is the inherent limitation of a measurement process, due to instrumentation
and process variation Measurement uncertainty does not include mistakes or process escapes
The probability of an incorrect decision is determined by: The amount of uncertainty in the measurement process Where the measurement result lies with respect to the tolerance limit (e.g., ± L) Knowledge acquired from previous measurements of similar items (i.e., a priori
distribution)
Probability of being out-of-tolerance
Probability of being out-of-tolerance
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 8
• Over specification of requirements costs money…
• However, under specification can be hazardous
Measurement Risk and End-item Performance
xd-xd
Nominal Value
xd-xd
xxf-xfxf-xf
Nominal Value
Point of complete loss of utility
Tolerance Limits
Area of degraded performance
Double measurement uncertainty
Measurement uncertainty Measurement uncertainty
Utility Range with margins Utility Range without margins
Start of degraded performance
Measurement uncertainty is inherent to the process or instruments and does not include mistakes, or process escapes
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 9
Essential Components of “Good” Measurement Data
Three essential components are required for measurements to adequately support decisions in a cost-effective manner
1. “Good requirements” - Reasonable measurement tolerances that are based on system performance
2. “Good equipment” - Measuring and test equipment that is properly calibrated
3. “Good measurements” - End-user measurement processes/procedures that adequately support the end-product performance requirements
Like the legs of a three-legged stool, all three components are necessary.
If one leg is missing, the risk that the stool will fall over increases; likewise, the risk of an incorrect decision increases dramatically if one of these components is missing.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 10
Component 1 – “Good Requirements”
Establish reasonable measurement requirements based on system performance. There must be a realistic link between the functional and design requirements. Without this link, either:
– the cost of verifying the measurement requirements will increase without adding value, or worse,
– verification of the design performance may not be adequate
Functional requirements: – how the product is intended to perform
– For example, how high for a given set of conditions (“spacecraft will operate in an orbit between 400 and 650 kilometers”)
Design requirements: – provides the realization for the Functional requirements
– the physical (e.g., size, weight, etc.) and operational (e.g., pressure, RPM, etc.) requirements of the product
This component can have the largest impact on subsequent cost of metrology and the achievable end item quality.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 11
Component 1 – Standardizing “Good Requirements”
ISO 9001:2008 and SAE AS9100C provide a standardized approach to “Good Requirements.”
ISO 9001:2008, Quality management systems — Requirements
7.3.2 Design and Development Inputs
These inputs shall include
a) functional and performance requirements,
7.3.3 Design and Development Outputs
Design and development outputs shall
c) contain or reference product acceptance criteria,
SAE AS9100, Quality Management Systems — Requirements for Aviation, Space and Defense Organizations adds to ISO 9001 requirements
7.3.1 Design and Development Planning
Design and development planning shall consider the ability to produce, inspect, test and maintain the product.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 12
Case Study 1 – Component 1 - Passive Latch Torque
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 13
Case Study 1 – Component 1 - Passive Latch Torque
Over specification of original Shuttle measurement design requirements led to three flight waivers and an expensive redesign of a failed torque system
1. Measurement requirement for latch bolt torque: 8,000-8,500 inch-lbs• Only lower torque limit linked to design requirements (maximum flight load)
2. Permissible bolt torque range per NASA standards: 8,000-10,200 inch-lbs3. Acceptable latch bolt torque for typical flight loads: 6,580-10,200 inch-lbs
Application of existing NASA or Industry standards would have allowed the use of off-the-shelf torque systems that were readily available
4 Latches Secure Airlock to Shuttle Close up of Latch Bolt
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 14
Component 2 – “Good Equipment”
Managing and controlling the accuracy of measuring equipment is an essential component. Periodic calibration of measuring instruments:
Ensures instrument accuracy Links measurements to national or international units of measure Enables measurements made at different places and/or times to be meaningfully compared
ISO 9001, Quality management systems — Requirements7.6 Control of Monitoring and Measuring Equipment, requires
The determination of measurements and equipment that provides evidence of product conformity
Calibration or verification of measuring instruments to national or international standards Calibration status labeling Assessment of measurement results made with non-conforming calibrated equipment
SAE AS9100, Quality Management Systems — Requirements for Aviation, Space and Defense Organizations adds to ISO 9001 requirements
A register for measuring equipment Suitable environmental conditions for calibration, inspection, and testing A process for recalling equipment requiring calibration/verification
NOTE: Utilizing international calibration standards, such as ISO 17025 and NCSLI Z540.3, provides a level of assurance as to the adequacy of the calibration results.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 15
Case Study 2 - Component 2 - “Good Equipment”
CoxHealth of Springfield, MO inadvertently overdosed 152 cancer patients, 76 of which received up to 70% higher than prescribed dosages
The device, a BrainLAB stereotactic radiation system used to treat areas 1.1 centimeters or smaller, was initially incorrectly calibrated by the CoxHealth chief physicist in 2004
The error went undetected for five years, until September 2009 when another CoxHealth physicist received training on the BrainLAB system
Although the calibration error was corrected, as of February 2012, the CoxHealth BrainLAB program remains suspended while lawsuits are settled
Incorrectly calibrated radiation treatment system overdosed 152 cancer patients
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 16
Component 3 - “Good Measurements”
Adequate measurement processes/procedures to control errors that could lead to incorrect decisions based on measurement data A properly calibrated instrument does not guarantee a “good” measurement result
ISO 9001, Quality management systems — Requirements8.2.4 Monitoring and measurement of product requires Measurements to verify that product requirements have been met Records to prove conformity with the acceptance criteria
SAE AS9100, Quality Management Systems — Requirements for Aviation, Space and Defense Organizations adds to ISO 9001 requirements Documentation of measurement requirements for product acceptance, including
a) criteria for acceptance and/or rejection,b) where in the sequence measurement and testing operations are to be performed,c) required records of the measurement results (at a minimum, indication of acceptance
or rejection), andd) any specific measurement instruments required and any specific instructions associated
with their use.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 17
• Hubble Space Telescope (HST) launched April 24, 1990
• On-orbit checkout revealed telescope could not be properly focused
• Ensuing investigation indicated the primary mirror was not built to specifications
• The optical test used to manufacture mirror was set-up incorrectly
• According to the HST failure report, project decision makers did not understand the accuracy of critical measurements
Case Study 3 - Component 3 - Hubble Space Telescope
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 18
Case Study 3 - Component 3 - Hubble Space Telescope
• Error was ten times the specified tolerance
• Two independent measurement tests clearly showed the error, yet both were discounted
• The HST failure report noted a mindset to discount any independent measurements with less accuracy than the RNC
• Discounting the INC and RvNC data was, in essence, “shopping for the answer they wanted”
• A servicing mission to correct the error was flown in December 1993 at a cost of over $1 billion USD
+ L (0.04 rms)
- L (0.04 rms)
Mirror Specification
INC RvNC
RNC
Each set of error bars represent the measurement process uncertainty, with 95% confidence, for the method used
Difference of 0.4 rms
“The project manager must understand the accuracy of critical measurements.”
HST Failure report
Mirror As-built
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 19
Obtaining “Good” Measurement Data
When examining measurement-based failures, in many, if not most cases, multiple components are inadequate.
Good Requirements Good Measurements
Good Equipment
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 20
Case Study 4 – Air Force B-2A Crash
The proximate cause of the crash was moisture in the Air Data System (ADS) which introduced large errors during a field calibration of several Port Transducer Units onboard the aircraft
The measurement procedure did not account or mitigate for moisture Component 3
There was a lack of understanding regarding how the ADS affects the aircraft flight safety Component 1
Although the calibration procedure was followed, an incorrect measurement-based decision led to the loss of a $1.4 billion asset, fortunately without loss of life
Air Force B-2A crash – February 23, 2008
The loss of the U.S. Air Force B-2 bomber in February 2008 is a dramatic example of measurement-based decisions leading to catastrophe. Performance requirement: Altimeter ± 75 feet of field elevation. Actual reading: + 136 feet of field elevation at take-off.
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 21
Case Study 5 – BP’s Texas City Refinery - March 23, 2005
Moments before and immediately after the explosion
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 22
Case Study 5 – BP’s Texas City Refinery (continued)
The Accident:• Distillation tower and attached
blowdown drum overfilled
• ~7600 gallons flammable liquid released
• Liquid ignited by an idling diesel truck
Proximate cause:• High-level alarm malfunctioned
• Level transmitter miscalibrated– Outdated 1975 data sheet
– Level transmitter indicated liquid level falling
– Level actually rising rapidly
Component 2
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 23
The Aftermath
Case Study 5 – BP’s Texas City Refinery (continued)
Root causes:• Cost-cutting, production
pressures, and failure to invest
• Lack of preventative maintenance and safety training
• Procedural workarounds to compensate for the deteriorating equipment Component 3
The Cost:– 15 deaths, – 180 injured– Over $2 billion,
including lawsuits
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 24
• Texaco drillers probe for oil
• A 75 foot “miscalculation” caused drill to breach salt mine shaft ~1200 feet below the lake’s surface
• 14 inch drill hole widened from lake water washing away the salt
• Lake began to drain into salt mine
Case Study 6 - Lake Peigneur, LA - November 20, 1980
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 25
Ensuing whirlpool sucked down
• Driller’s oil platform,
• Second drilling rig,
• 11 barges,
• Tugboat (at full throttle)
• 70 acres of Jefferson Island
• A barge loading dock
• Many cars, trucks, & trees
Case Study 6 - Lake Peigneur, LA - November 20, 1980
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 26
Whirlpool
• Reversed flow of 12-mile-long canal from the Gulf of Mexico
• Replaced 3.5 billion gallons of lake water with saltwater
• Created largest waterfall in LA
• Lake depth increased from ~11 feet to ~1300 feet
Cost
•~$50M to Texaco
• Unknown cost to ecosystem & surrounding community
• No fatalities (except 3 dogs)
“Measure thrice, cut once…”
Case Study 6 - Lake Peigneur, LA - November 20, 1980
Metrology and the C
onsequences of Bad M
easurement D
ecisions
Page 27
Considerations and Cautions
Measurements support decisions – accept, reject, rework, scrap, or even launch a space vehicle. In conformance testing, if a measurement does not support a decision, it is unnecessary.
Essential Components:
1. Design specifications must be linked to functional requirementsa) Over specification is expensive, b) Under specification can be hazardous
2. Calibration ensures the accuracy of measuring equipment a) Links units of measurement to International standards (i.e., provides a pedigree)
3. Measurement processes must control errors that may lead to incorrect measurement-based decisionsa) Proper selection and utilization of measuring equipmentb) Control/mitigate other relevant sources of error in the measurement process
– After all, a perfect instrument does not guarantee good measurement data
The more critical the decision, the more critical the data. The more critical the data, the more critical the measurement. NASA Reference Publication 1342