Using Dynamic Multi-Attribute Tradespace
Exploration to Develop Value Robust Systems
Dr. Adam M. Ross, MIT
adamross@mit.edu
Sikorsky IEEE CT AES Lunch
May 24, 2007
Meeting Customer Needs
• Goal of design is to create value (profits,
usefulness, voice of the customer, etc…)
• Requirements capture a mapping of needs
to specifications to guide design
seari.mit.edu
© 2007 Massachusetts Institute of Technology
2
Deploying a “Valuable”
System…
Contexts change…
seari.mit.edu
© 2007 Massachusetts Institute of Technology
3
Meeting Customer Needs
(cont.)
• Goal of design is to create value (profits,
usefulness, voice of the customer, etc…)
•• Requirements
a mapping of needs to
People changecapture
their minds…
guidevalue,
designsystems must
• specifications
To continue totodeliver
change as well…
seari.mit.edu
© 2007 Massachusetts Institute of Technology
4
What is System Success?
Success is defined across multiple perspectives and multiple time periods
System success, Ψ, across N decision makers at time t
N
[
Net “experience”
Ψ (t ) = ∑ X DMi (t ) + ε
i =1
Net “expectations”
X DMi
C
(t ) ≥ YDMi (t ) + ε
YDMi
C
(t )
0 ≤ Ψ (t ) ≤ N
X DMi (t )
Decision maker i unaffected
system “experience” at time t
YDMi (t )
Decision maker i unaffected
system “expectation” at time t
ε CX (t )
DMi
ε
YDMi
C
(t )
Context effect on decision
maker i “experience” at time t
Context effect on decision
maker i “expectation” at time t
System Success: Net “experience” must meet or exceed net “expectations”
seari.mit.edu
]
© 2007 Massachusetts Institute of Technology
5
Characterizing the System
Design Opportunity
Expectations
Expectations
<
>
<
<
X
Experience
? Success?
Time
Experience
Exogenous
Influences
Decision Maker
Value
…
Exogenous
Influences
Designs
Resources
Expectations
Designer
Needs
Value
Constraints
System
Context
Experience
seari.mit.edu
Experience
© 2007 Massachusetts Institute of Technology
6
Types of Changes
∆ Designs (including technology)
∆ Context (including operating environment, competition)
Physical (e.g. nature)
∆ Constraints (including “laws”)
Human-made (e.g. policy, schedule)
Resources (e.g. capital)
∆ Needs (including attributes)
Scoping (e.g. self-imposed)
∆ DMs (including individuals and groups)
∆ Resources (including dollars and time)
How can System Designers cope with these types of changes during design?
seari.mit.edu
© 2007 Massachusetts Institute of Technology
7
Aspects of Dynamic MATE
How can System Designers cope with these types of changes during design?
Experience(t)
• System Success criteria
N
[
Ψ (t ) = ∑ X DMi (t ) + ε CX DMi (t ) ≥ YDMi (t ) + ε CYDMi (t )
]
i =1
Utility
– Expanding scope of system “value”
Expectation(t)
Exceeding
• Tradespace exploration
– Understanding success possibilities
across a large number of designs
Cost
Change Agent
• Change taxonomy
– Specifying and identifying change types
Internal
(Adaptable)
None
(Rigid)
Change Effect
External
(Flexible)
– Analyzing changeability of designs
• System Epoch/Era analysis
– Quantifying effects of changing
contexts on system success
Parameter level Parameter set
(Scalable)
(Modifiable)
Utility
• Tradespace networks
None
(Robust)
Cost
T1
U
Epoch 1
T2
T3
Tn
Epoch 2
Epoch 3
Epoch n
…
0
S1,b
S1,e S2,b
S2,e S3,b
S3,e
S n,b
Sn,e
Time
seari.mit.edu
© 2007 Massachusetts Institute of Technology
8
Tradespace Exploration
Value-driven design…
Firm
Designer
Customer
User
Value
Attributes
Utility
Analysis
Concept
Design
Variables
“Cost”
Each point is
a specific
design
Tradespace: {Design,Attributes} {Cost,Utility}
DESIGN VARIABLES:
Design trade parameters
Orbital Parameters
–
–
–
Co
st
,U
tili
ty
Apogee Altitude (km)
Perigee Altitude (km)
Orbit Inclination (deg)
Total Lifecycle Cost
($M2002)
Spacecraft Parameters
ATTRIBUTES:
Design decision metrics
–
–
–
–
–
–
–
–
–
–
Antenna Gain
Communication Architecture
Propulsion Type
Power Type
Total Delta V
Data Lifespan (yrs)
Equatorial Time (hrs/day)
Latency (hrs)
Latitude Diversity (deg)
Sample Altitude (km)
Assessment of cost and utility of large space of possible system designs
seari.mit.edu
© 2007 Massachusetts Institute of Technology
9
Example “Real Systems”
Spacetug vs CX-OLEV
XTOS vs Streak
Total Lifecycle Cost
($M2002)
Electric Cruiser
(2002 study)
CX-OLEV
(2009 launch)
Wet Mass kg
1405
1400
Dry Mass kg
805
670*
Propellant kg
600
730*
Equipment kg
300
213*
DV m/s
12000 – 16500***
15900**
Utility
0.69
0.69
Cost
148
130*
seari.mit.edu
XTOS (2002 study)
Streak (Oct 2005 launch)
Wet Mass kg
325 - 450
420
Lifetime (yrs)
2.3 - 0.5
1
Orbit
300 -185 km @ 20°
321a-296p -> 200 @ 96°
LV
Minotaur
Minotaur
Utility
0.61 - 0.55
0.57 - 0.54*
Modified Utility**
0.56 - 0.50
0.59
Cost $M
75 - 72
75***
Instruments
Three (?)
Ion gauge and atomic
oxygen sensor
© 2007 Massachusetts Institute of Technology
10
Tradespace Analysis: Selecting
“best” designs
B
D
C
New
New “best”
“best” design
design
Utility2
Utility1
Classic
Classic “best”
“best” design
design
E
C
B
E
D
A
A
Cost
Cost
Time
If the “best” design changes over time, how does one select the “best” design?
seari.mit.edu
© 2007 Massachusetts Institute of Technology
11
Example: X-TOS Transition Rules
Rule
Description
Utility
Utility
Tradespace Networks
Change agent origin
R1: Plane Change
Increase/decrease inclination, decrease ∆V
R2: Apogee Burn
Increase/decrease apogee, decrease ∆V
R3: Perigee Burn
Increase/decrease perigee, decrease ∆V
Internal (Adaptable)
R4: Plane Tug
Increase/decrease inclination, requires “tugable”
External (Flexible)
R5: Apogee Tug
Increase/decrease apogee, requires “tugable”
External (Flexible)
R6: Perigee Tug
Increase/decrease perigee, requires “tugable”
External (Flexible)
R7: Space Refuel
Increase ∆V, requires “refuelable”
External (Flexible)
R8: Add Sat
Change all orbit, ∆V
External (Flexible)
Internal (Adaptable)
Internal (Adaptable)
Transition
rules
Cost
Tradespace designs
= nodes
Cost
Applied transition rules = arcs
Cost
1
2
1
3
2
4
Transition rules are mechanisms to change one design into another
The more outgoing arcs, the more potential change mechanisms
seari.mit.edu
© 2007 Massachusetts Institute of Technology
12
Tradespace Networks: Changing
designs over time
B
D
C
New
New “best”
“best” design
design
Utility2
Utility1
Classic
Classic “best”
“best” design
design
E
C
B
E
D
A
A
Cost
Cost
Time
Select changeable designs that can approximate “best” designs in new contexts
seari.mit.edu
© 2007 Massachusetts Institute of Technology
13
Changeability Metric: Filtered
Outdegree
objective
subjective
Outdegree
Filtered Outdegree
# outgoing arcs
from a given node
# outgoing arcs from design at acceptable “cost”
(measure of changeability)
OD(<C)
OD(<Ĉ)
RK
RK+1
RK+1
ODK
> Ĉ
< Ĉ
> Ĉ
OD(Ĉ)
Outdegree
C
Cost
Ĉ
Subjective Filter
Filtered outdegree is a measure of the apparent
changeability of a design
seari.mit.edu
© 2007 Massachusetts Institute of Technology
14
Ex: X-TOS Outdegree function
DV
2471
903
1687
2535
1909
3030
Inclination
90
30
70
90
70
90
90
Apogee
460
460
460
460
1075
2000
770
Perigee
150
150
150
290
150
150
350
TDRSS
Com Arch
7156
TDRSS
TDRSS
TDRSS
TDRSS
TDRSS
TDRSS
Delta V
1200
1200
1200
1200
1200
1200
1000
Prop Type
Chem
Chem
Chem
Chem
Elec
Elec
Chem
Pwr Type
Fuel Cell
Fuel Cell
Fuel Cell
Fuel Cell
Fuel Cell
Solar Array
Solar Array
Ant Gain
Low
Low
Low
Low
Low
Low
Low
Data Life
0.51
0.51
0.51
10.05
0.52
0.61
11
Lat Div
180
60
140
180
140
180
180
Eq Time
5
11
6
5
2
2
5
Latency
2.27
2.27
2.27
2.30
2.42
2.67
2.40
Sample Alt
150
150
150
290
150
150
350
Cost ($10M)
4.21
4.21
4.21
4.88
4.52
4.99
4.15
Pareto Set designs (903, 1687, 2535, 2471) are not the
most changeable
Design 7156 becomes relatively more changeable as cost
threshold increases
Outdegree functions reveal differential nature of apparent changeability
seari.mit.edu
© 2007 Massachusetts Institute of Technology
15
Tradespace Networks in the
System Era
Pareto Tracing across
Epochs
System Era
T1
U
Epoch 1
T2
T3
Tn
Epoch 2
Epoch 3
Epoch n
…
0
S1,b
S1,e S2,b
S2,e S3,b
S3,e
Sn,b
Sn,e
Time
Change Tradespace (N=81), Path: 81-->10, Goal Util: 0.97
1
Change Tradespace (notional), Goal Util: 0.97
1
0.9
0.8
Changeability Quantified
as Filtered Outdegree
0.7
≈Nk
U
U
0.6
0.5
0.4
0.3
0.2
0.1
Rk
0
0
0.5
1
1.5
Total Delta C
2
2.5
0
0
1
2
3
Total Transition Time
4
ODk
Temporal strategy can be developed across networked tradespace
seari.mit.edu
© 2007 Massachusetts Institute of Technology
16
Example System Timeline
Example system: Serviceable satellite
Value
degradation
New Context: new
value function
(objective fcn)
Major failure
Service to
Major failure
“upgrade”
T1
System BOL
U
Epoch 1
T2
Tn
Epoch 2
Epoch n
System EOL
…
0
S1,b
Value outage:
S2,e
Servicing time
Same system,
Service to
but perceived
“restore”
value decrease
S1,e S2,b
Service to
“restore”
Sn,b
Sn,e
System timeline with “serviceability”-enabled paths allow value delivery
seari.mit.edu
© 2007 Massachusetts Institute of Technology
17
Achieving Value Robustness
Research suggests two strategies for
“Value Robustness”
Active
Utility
T1
New Context Drivers
• External Constraints
• Design Technologies
• Value Expectations
Passive
T2
Epoch 1
Epoch 2
0 Time
• Choose “clever” designs that
remain high value
• Quantifiable: Pareto Trace number
S1,b
Utility
1. Passive
S1,e S2,b
State 1
DV2≠DV1
U
S2,e
State 2
DV2=DV1
2. Active
• Choose changeable designs that
can deliver high value when
needed
• Quantifiable: Filtered Outdegree
Cost
Cost
Value robust designs can deliver value in spite of inevitable context change
seari.mit.edu
© 2007 Massachusetts Institute of Technology
18
Thank you for your attention!
Any questions?
adamross@mit.edu
For further details on topic please see:
Ross, Adam M., Managing Unarticulated Value: Changeability in
Multi-Attribute Tradespace Exploration. Cambridge, MA: MIT. PhD in
Engineering Systems. 2006.