Growing the
Lean Community
An LAI Plenary Conference
Understanding Risk
and Uncertainty
April 10, 2001
Presented By:
John Deyst
MIT
Research Sponsored LAI
Lean
Aerospace
Initiative
Advanced DSM model example
Relative Probability
Joint PDF of Cost and Schedule for (real) PD process
180
170
160
150
140
130
Schedule (days)
2 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
120
550
600
650
700
750
Cost ($k)
800
850
900
From
Browning,
1998
web.mit.edu/lean
Lean
Aerospace
Initiative
Evolution of Performance
1050
1000
950
900
850
800
750
700
0
20
40
60
80
100
120
140
Project Time (t, days)
From Browning,
1998
3 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Value Creation
Graph Created by James Chase
4 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Hypotheses
➢Product development is an investment
➢Uncertainty and risk are inherent, to
varying degrees, in all product
developments
➢Risk is the probability that the product
requirements will not be achieved, thus
imperiling the investment
5 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Hypotheses (cont.)
➢Quantifying uncertainty and risk will enable
better management of product
developments
➢Value in product development is the
amount by which risk is reduced per unit of
resource expended
➢Balancing risk across all aspects of
product development is a significant lean
principle
6 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Design Parameters vs.
Performance Measures
Design
Design
Parameters
Parameters
(many)
(many)
(e.g.
- L / D, η , w / w
1
Functional
Functional
Relationships
Relationships
f
)
7 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
(e.g.- Breguet Range
Equation)
Performance
Performance
Measures
Measures
(not
(not so
so many)
many)
(e.g. – Range)
web.mit.edu/lean
Lean
Aerospace
Initiative
Large
Large Uncertainty
Uncertainty
In
In Many
Many Design
Design
Parameters,
Parameters, Hence
Hence
Large
Large Risk
Risk to
to
Performance
Performance Measures
Measures
(initial state)
8 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
Viewing Product Development
Uncertainty Reduction
Product
Product
Development
Development
Small
Small Uncertainty
Uncertainty
In
In Design
Design
Parameters
Parameters and
and
Performance
Performance
Requirements
Requirements Met
Met
(final state)
web.mit.edu/lean
Lean
Aerospace
Initiative
Evolution of Knowledge
of Performance
2.5
Requirement
2
Mature
Design
1.5
1
Immature
Design
Risk of
Failure!
0.5
0
0
5
10
15
20
Performance
9 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Basic Assumptions
➢Effective product developments have the
following characteristics
➢Processes are well calibrated so that error
biases and trends are negligible
➢As new information is generated it is combined
with old Information according to the relative
qualities of new and old
➢All productive activities serve to reduce
uncertainty
10 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Product Development Model
➢From these assumptions information
theory yields the following differential
equationdp
= −up2
dt
Rate of change
of uncertainty
variance
Effectiveness
of the
expended
effort
Square of the variance of
2
uncertainty ( p = σ )
➢Rate of variance reduction is proportional
to
➢ effectiveness
➢ square of variance
11 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Uncertainty Reduction Derived
from the Model
Standard Deviations of L/D Estimation Errors
Upper Curve-2 Wk Spending Rate Ramp Up
Lower Curve-Constant Spending Rate
1.6
Standard Deviation
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
1
2
3
4
5
6
7
8
9
10
Time (wks)
12 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Multi – Parameter /Multi
Activity Product Development
➢Parameters are an array (n dimensional
design vector)
➢Uncertainty is a matrix (n x n covariance)
l
T
P = -PH UHP
Rate of change
of uncertainty
13 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
Effectiveness
matrix
Activity matrix
Uncertainty matrix
web.mit.edu/lean
Lean
Aerospace
Initiative
Model of a Transport Aircraft
Preliminary Design Study
➢Design Parameters-(3 dimensional)
➢Lift – to – Drag Ratio
➢Overall Efficiency
L/D
(η )
➢Initial to Final Weight Ratio
(w / w )
1
f
➢Performance Measure-(1 dimensional)
➢Range > 13,000 km
➢Product Development Activities
➢Aerodynamic design
➢Propulsion System Design
➢Structural Design
14 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean
Lean
Aerospace
Initiative
Model of a Transport Aircraft
Preliminary Design Study (cont.)
➢ Initial design parameter estimates indicate
range>13,000km
➢ Uncertainty in these estimates produces a probability of
0.4 that the final product will not meet the requirements
(unacceptable risk)
➢ Preliminary PD effort will refine the design and reduce
uncertainty by a factor of 5
➢ Initial Plan-spend $500K over a 10 week period to refine
the design and reduce uncertainty by the required factor of
5
➢ Final Plan- for the same reduction in uncertainty, but by
balancing risk over the three activities, the cost is reduced
by 13% (saves $66K)
15 - PD Deyst - 041001 © 2001 Massachusetts Institute of Technology
web.mit.edu/lean