Transcript An Argument for Applying Objective Based Optimization Models in

An Argument for Applying Objective Based
Optimization Models in the Design of
Verification and Inspection Activities
Stephen J. Walsh and Claude F. Norman
January 24, 2012
IAEA
International Atomic Energy Agency
The Context
“As we (the IAEA) are not a research organization,
we don’t know what we don’t know (regarding the
state of technology and R&D developments that may
be useful for us)”
– SGCP member at last week’s long term R&D
seminar
IAEA
The Objective of the Paper
•
To attract the attention of technical
researchers who solve an organization’s
problems from a holistic perspective where
the efficient use of technology is quantified.
•
To make them aware that we may have an
interest in advancing the technical
(probabilistic and statistical) formulations
which underlie the design and
implementation of standard inspection
activities (Traditional Safeguards Activities).
•
To provide them with the context of how we
wish to see the SOA advanced in a manner
consistent with the overarching objectives of
Information Driven Safeguards.
IAEA
“Innovative cultures are learning cultures”
Motivation:
Inspection/Verification Algorithms SOI vs. SOT vs. SOA
State of
Implementation
•
Standard statistical
approaches – circa < 1990
•
Random inspections
garnering more respect and
implementation ~ 2005
•
Minimal treatment of
diversion strategy
Applied many
approximations which were
motivated by limited
computing technology
•
Poor definition of detection
vs. selection probability
<
•
Expand to Exact Formula
and take advantage of cheap
computing power
•
Thorough treatment of
diversion strategy –
Simulation
•
•
IAEA
State of
Technology
<
Advance standard statistical
assumptions that are now
know to be false.
State of the Art
•
Theory: some standard
statistical approaches have
been generalized to
operational approaches.
•
Rigorous treatment of
diversion strategy
•
Some objective based
approaches are mature –
conduct an inspection under
an objective as opposed to
criterion.
“A culture of innovation can be a company’s primary source of
competitive advantage and can pay off steadily over the years.”
The Target Audience (Researchers who
think/work like this!)
Operations Research:
(decision science, management science)
Formally referred to as Optimization Theory
• An interdisciplinary mathematical science that
focuses on the effective use of technology by
organizations contrasted with many other science
and engineering disciplines which focus mainly on
technology giving only secondary consideration to
its use.
• An application of the scientific method to solving
an organization's operational problems.
IAEA
“Operations Research: The world’s most important invisible
Profession.”
Operations Research: Mathematical Regimes
and Techniques
•
•
•
•
•
•
•
•
•
•
Data Mining/Analysis
Decision Analysis
Engineering
Forecasting
Game Theory
Industrial Engineering
Logistics
Mathematical Modeling
Financial Engineering
Fuzzy logic
IAEA
• Mathematical
•
•
•
•
•
Optimization (STR-261)
Probability and Statistics
(Standard Inspection
Design, STR-340, Material
Balance Evaluation,
Random Inspections)
(non-)Linear programing
Simulation
Statistical Decision Theory
Pattern Theory
“In mathematics you don’t understand things, you just get
used to them.”
Technical Aspects of Inspection/Verification
Design
•
•
•
•
•
•
Stratification
Assumptions on the existence of ‘defects’
Diversion Strategy
Random selection of items to verify
Selection vs. detection probability
Inference/How data gathered during inspection is
analyzed
• Effectiveness of the inspection as a deterrent to
diversion
IAEA
Contrast:
General Description
Probabilistic Approach
Criterion based –
Desired Detection Probability 
Number of items to Verify
Optimization Approach
Objective based – build an
inspection plan on a quantitatively
defined objective
•
Objective of Inspection
Achieve a Prescribed Detection
Probability
•
•
•
Definition of Inspection
A collection of samples sizes for
each strata
•
•
Diversion Strategy
Strength of Conclusions
IAEA
To guarantee a detection
probability over all diversion
strategies
To minimize the time to
detection of diversion
To quantify the quality of the
inspection as a deterrent
A collection of sample sizes for
each strata
A measure of cost to verify
items in specific strata
A total cost constraint
A single instance is assumed to
compute the inspection
The space of all possible diversion
strategies is considered to find the
optimal inspection
Confidence that the criterion were
met.
Deeper conclusions: confidence
that the objective of inspection was
met.
Conclusions (Summary of the Paper)
• The purpose of the paper is to attract technical researchers with an OR
philosophy to our problem space.
• Some historical background on the development of
probabilistic/statistical approaches to inspection/verification will be
presented.
• To communicate we are interested in advancements in inspection
design that are consistent with the goals of information driven
safeguards.
• Illustrate our points by providing heuristic contrast of probabilistic
approaches with optimization approaches.
IAEA