Objective Read World Uncertainty Analysis

1
Objective Read World Uncertainty Analysis
CMSC 2003 July 21-25 2003
2
Introduction

• Confidently optimize production processes against
  their requirements
• Inputs vs. Outputs
• Need to simulate process performance to optimize
  accuracy, speed, and costs
• Need reliable (easy to understand) uncertainty
  estimates for complex 3D measurements on the
  factory floor
• Need to estimate the benefits of combining
  measurement systems
• Common Type Network (n Trackers)
• Hybrid Type Network (Scanners Trackers etc.)
• Need real-time (easy to understand) feedback on
  measurement system performance
• Need traceable measurement uncertainty for each
  assembly

3
Process Description

• Inputs
• Object Characteristics (e.g., Volume, Surface)
• Expected Tolerances
• Instrument Types and Number of Stations
• Cycle Time
• Measurement Constraints (e.g., line-of-sight,
  targeting the actual critical features)

• Outputs
• GUM Compliant Uncertainty Estimates of Feature
  Measurements
• Measurement Plan
• Number of Instruments (Stations)
• Types of Instrument
• Instrument Placement
• Targeting Requirements
• Network/Orientation Requirements
• Transform vs. Bundle
• Number of Common Pts
• Closure
• Analysis Dependencies

4
Background Uncertainty

• Guide to Uncertainty in Measurement (GUM)
• ISO way to express uncertainty in measurement
• Error and Uncertainty are not the same
• Quantify components of Uncertainty
• Type A vs. B depends on the estimation method
• A Statistical Methods (e.g., Monte Carlo,
  1st-order Partials)
• B Other means (e.g., measurements, experience,
  specs)
• Random vs. Systematic Effects (e.g., Noise vs.
  Scale)
• Both are components estimated with Type A or B
  methods
• Uncertainty Estimates can contain Type A B
  methods
• GUM mandates uncertainty statements in order to
  provide traceability for measurement results
• A measurement result is complete only when
  accompanied by a quantitative statement of its
  uncertainty 1

1 - Taylor and Kuyatt, 1994 NIST TN/1297
5
Background Uncertainty

• Specifications
• Instrument specifications are not representative
  of the results from actual use of the instrument
  in a network
• 3D Measurement Networks
• 1 Instrument References
• 1 Instrument in multiple locations References
• n Instruments (types) References
• Application of 3D Measurement Systems
• Real use ? multiple stations and different
  instruments in the same network
• Quantify coordinate data uncertainty fields in a
  network
• Practical methods to estimate the uncertainty of
  specific systems
• Combining measurement systems
• Combining measurement uncertainties
• Results need to in an easy to understand and
  meaningful format

6
Monte Carlo

• What Non-linear statistical technique
• Why Difficult problems and expensive to state or
  solve
• When Consequences are expensive
• How
• List of possible conditions (where the activity
  being studied is to large or complex to be easily
  stated)
• Random numbers (from estimates of each measured
  component)
• Model of Network interactions
• Large number of solution are run
• Statistical inferences are drawn

Monte Carlo technique was developed during World
War II in Los Alamos for the atom bomb project
7
Models

• Modeling
• Instruments
• Axes
• Angles
• Ranging
• Offsets
• Joins
• Measurements
• Angles ? ppm
• Ranges ? ppm offset
• Confidence

8
Wing to Body JoinApplication

• Inputs
• CAD Model includes Features, Relationships,
  Tolerances
• Sweep, Dihedral, Incidence
• Scanners, Trackers, Local GPS, Robotics, Gap
  Measurement Devices
• Production Measurement Analysis lt 3 minutes
• Aluminum Surface
• Targeted and Pre-measured Assembly Interface
  Features
• Transfer critical object control to continuously
  visible features

• Outputs
• Surface 0.080 _at_ 2?
• Features 0.004 _at_ 2?
• 2 Scanners Local GPS GAP Measurement Tool
• Optimized Instrument Location
• Bundle Local GPS and Transfer to (11) Common Pts
• Local GPS updates at 2 Hz
• Aerodynamically matched orientation within
  process uncertainty

9
Application
10
Outputs
11
Outputs
12
Results
13
Results
14
Conclusions
15
Acknowledgements

• John Palmateer (Boeing)
• Dr. Joe Calkins (New River Kinematics)

16
Summary