Académique Documents
Professionnel Documents
Culture Documents
33 Systems Engineering
Lecture 6
Requirements Driven Systems
Design
Qi Van Eikema Hommes
Course Layout
Lecture 4 ✔
Stakeholder Analysis and Lecture 2 ✔ Lecture 13
Design VerificaKon
Requirements DefiniKon Systems Engineering and
As Human AcKvity ValidaKon, Lifecycle
Management
Lecture 5 innovaKon in
✔
Systems Engineering
Lecture 6 AxiomaKc
Design and DM‐DSM
Method
Lecture 7 CPM and DFSS
Lecture 8
Trade Space ExploraKon
Concept SelecKon
Lecture 10: Experiments
Lecture 11: Robust Design I
Lecture 12: Robust Design II
6/24/10 Qi Van Eikema Hommes 2
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Constraints
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 3
The Founder of AxiomaKc Design Theory
• Nam Pyo Suh—MIT Professor Emeritus.
• B.S., Mechanical Engineering, 1959, M.S., Mechanical Engineering, 1961, MIT
.
• Ph.D, Mechanical Engineering, 1964, Carnegie Mellon University.
• From 1965‐1969, Suh served as a professor at the University of South Carolina. In 1970
he began his professional career at MIT‐‐ serving as director of the MIT‐Industry
Polymer Processing Program from 1973‐1984; director of the Laboratory for
Manufacturing and ProducKvity from 1977‐1984; and Mechanical Engineering
Department Head from 1991 to 2001. Although sKll keeping the Ktle of Ralph E.
Cross Professor of Mechanical Engineering at MIT, Suh is now president of KAIST.
Qi Van Eikema Hommes
6/24/10 4
The Goals of AxiomaKc Design
• Establish a scienGfic basis for design
• Improve design acKviKes by providing the
designer with a theoreKcal foundaKon based on
logical and raGonal thought processes and tools.
• Make human designers more creaGve
• Reduce the random search process
• Minimize the iteraKve trial and error process
• Determine the best designs among those
proposed
Suh, Axiomatic Design, 2000, page 5
Qi Van Eikema Hommes
6/24/10 5
• an interplay between what we want
achieve and how we will achieve it.
What
6/24/10
Qi Van Eikema Hommes 5
The Four Domains of Design
Qi Van Eikema Hommes
6/24/10 7
DefiniKons
• Customer AOribute (CA)—what customer desire
from a product
• FuncGonal Requirement (FR)—minimum set of
independent requirements that completely
characterize the funcKonal needs of the product in
the funcKonal domain.
• Design Parameter (DP)—Key physical variables in the
physical domain that characterize the design that
saKsfies the specified FRs.
• Process Variables (PV)—key variables in the process
domain that characterize the process that can
generate the specified DPs.
Qi Van Eikema Hommes
6/24/10 8
Benefits of the Domains
• Customer Needs are stated in the customer’s
language
• FuncKonal Requirements and Constraints are
determined to saKsfy Customer Needs
• “The FRs must be determined in a soluKon neutral
environment” (or, in other words, say “what” not
“how”)
– BAD = the adhesive should not peel
– BETTER = the amachment should hold under the
following loading condiKons
• Provide Requirements Traceability
Qi Van Eikema Hommes
6/24/10 9
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 10
Axiom
• Axioms are truths that cannot be derived but
for which there are no counter examples or
excepKons.
• Examples of Axioms:
– First and second law of thermodynamics
– Newton’s three law of mechanics
Qi Van Eikema Hommes
6/24/10 11
How were the Design Axioms
Created?
• IdenKfying the common elements that are
present in all good designs:
– How did I make such a big improvement in a
process?
– How did I create the process?
– What are the common elements in good designs?
• Use logical reasoning process to reduce the
observaKons to two Axioms.
Qi Van Eikema Hommes
6/24/10 12
The Two Axioms
• Axiom 1: Independence Axiom—maintain the
independence of funcKonal requirements
(FRs).
• Axiom 2: The InformaGon Axiom—minimize
the informaKon content of the design.
Qi Van Eikema Hommes
6/24/10 13
Qi Van Eikema Hommes
6/24/10 15
Design Matrix Example
• FR1 = Provide access to the items
stored in the refrigerator
• FR2 = Minimize energy loss
• DP1 = VerKcally hung door
• DP2 = Thermal insulaKon material in
the door
=
FR2 x x DP2
Suh, Axiomatic Design, 2000
Qi Van Eikema Hommes
6/24/10 16
A Different Design
• FR1 = Provide access to the items
stored in the refrigerator
• FR2 = Minimize energy loss
• DP1 = Horizontal door
• DP2 = Thermal insulaKon material
in the door
FR1 x 0 DP1
=
FR2 0 x DP2 Image by MIT OpenCourseWare.
Qi Van Eikema Hommes
6/24/10 17
FR DP
Uncoupled
A11 0 0 1 1
Design 0 A22 0 2 2
0 0 A33 3 3
Decoupled A11 0 0 1 1
2 2
Design A21 A22 0
3 3
A31 A32 A33
1 1
Coupled A11 A12 A13
Design A21 A22 A23 2 2
6/24/10 18
Qi Van Eikema Hommes
Axiom 1: Independence Axiom
• To saKsfy the Independence Axiom, the design
matrix must be either diagonal or triangular.
A11 0 0 A11 0 0
0 A22 0 A21 A22 0
0 0 A33 A31 A32 A33
Uncoupled Decoupled
Design Design
€ €
Qi Van Eikema Hommes
6/24/10 19
Water Faucet Example
• FuncKonal Requirements:
– FR1: Adjust the water temperature (T)
– FR2: Adjust the water volume (Q)
Qi Van Eikema Hommes
6/24/10 20
What is the Design Matrix?
Qi Van Eikema Hommes
6/24/10 21
What is the Design Matrix?
Qi Van Eikema Hommes
6/24/10 23
FuncKonal Coupling vs Physical
Coupling
# of parts ≠ # of DPs
Qi Van Eikema Hommes
6/24/10 24
Why MeeKng Axiom 1 is
Desirable?
Qi Van Eikema Hommes
6/24/10 25
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Constraints
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 26
Zig Zagging
Functional domain Physical domain
FR DP
Qi Van Eikema Hommes
6/24/10 27
Refrigerator Design Example
• FR1 = Freeze food for long‐term
preservaKon
• FR2 = Maintain food at cold temp for
short‐term preservaKon
• DP1 = the freezer secKon
• DP2 = the chiller (refrigerator) secKon
FR1 x 0 DP1 Image by MIT OpenCourseWare.
=
FR2 0 x DP2
Qi Van Eikema Hommes
6/24/10 28
Decompose the System
FR1 = Freeze food for long term preservaKon
FR11 = Control freezer temp
FR12 = Maintain uniform freezer temp
FR13 = Control freezer humidity
FR2 = Maintain food at cold temp for short term preservaKon
FR21 = Control chiller temp
FR22 = Maintain uniform chiller temp
DP1 = The freezer secKon
DP11 = Sensor/compressor system for freezer secKon
DP12 = Air circulaKon system for freezer secKon
DP13 = Condenser that condenses the moisture in the air when dew
point is exceeded
DP1 = The chiller secKon
DP21 = Sensor/compressor for chiller secKon
DP22 = Air circulaKon system for chiller secKon
Qi Van Eikema Hommes
6/24/10 29
What Does The Design Matrix Look Like?
Freezer fan Fridge fan
Two cooling fan type
Capillary F-Fan
tube Freezing room
Cold air
R-Fan Refrigerating
room
Compressor
New Cooling System Refrigerator
A schematic drawing of a new refrigerator design. A cold-air circulation system with many vents.
DP1 DP2
DP12 DP11 DP13 DP22 DP21
FR12 0 0 0 0
FR1 0
FR11 0 0
FR13 0 0 0
FR2 FR22 0 0 0 0
FR21 0 0 0
6/24/10
Qi Van Eikema Hommes 31
Can We Save the Cost of a Fan?
Capillary
tube Fan Freezing room
Cold air
Evaporator
Condenser Cold air flow
Damper
Refrigerating
room
Compressor
Conventional Refrigerator
Schematic drawing of a conventional refrigerator. Cold-air circulation in a conventional
refrigerator..
Image by MIT OpenCourseWare.
6/24/10
Qi Van Eikema Hommes
32
DP1 DP2
DP12 DP11 DP13 DP22 DP21
FR1 FR12 x 0 0 0
X 0
FR11 x x 0 0
X 0
FR13 x 0 x 0 0
FR2 FR22 0 0 0 x 0
FR21 0
X 0
X 0 x x
Coupled design!
Qi Van Eikema Hommes
6/24/10 33
Benefits So Far from Axiom 1
• Reduce system coupling early on.
• Start the design with requirements first.
• Think about the design concept first before
applying robust engineering or opKmizaKon
blindly.
• Zig‐zagging instead of staying in one domain.
• Requirements traceability and raKonale.
Qi Van Eikema Hommes
6/24/10 34
Class Discussions
• How does Zig‐zagging help design synthesis?
• How does your organizaKon decompose
systems and requirements?
• Does this help with requirements traceability
throughout the design?
• How does AxiomaKc Design differ from QFD?
Qi Van Eikema Hommes
6/24/10 35
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Constraints
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 36
Constraints in AxiomaKc Design
• Constrant (C)—are bounds on acceptable
soluKons. Input constraints are imposed as
part of the design specificaKon. System
constraints are constraints imposed by the
system in which the design soluKon must
funcKon.
Qi Van Eikema Hommes
6/24/10 37
Constraints
• Two types of constraints:
– Input constraints—specific to the overall design
goals (all design proposed must saKsfy these).
• Example: cost
– System constraints—specific to a given design
(they are the result of design decisions made).
• Example: Diesel engine tailpipe emission standards
for diesel engines
• What kind of constraint is Safety?
Qi Van Eikema Hommes
6/24/10 38
What AxiomaKc Design Says about
Constraints
• “Constraints provide bounds on the
acceptable design soluKons and differ from
the FRs in that they do not have to be
independent.”
Qi Van Eikema Hommes
6/24/10 39
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 40
InformaKon Content
• InformaKon Content Ii for a given FRi is defined in
terms of the probability Pi of saKsfying FRi:
Ii = log2(1/Pi)=‐ log2(Pi) Probability density
Target
• When there are m FRs, Bias
m
Isys = −log 2 (Pm ) = −∑ log 2 Pi Area within
common
i=1 Pi range (Acr)
System range FR
Qi Van Eikema Hommes
€ 6/24/10 41
Axiom 2 InformaKon Content
• The InformaKon Axiom—Minimize
informaKon content I.
• Maximize the probability of meeKng FRs.
m
Isys = −log 2 (Pm ) = −∑ log 2 Pi
i=1
Qi Van Eikema Hommes
6/24/10 42
Example of Buying a House
Suh, AxiomaKc Design, 2001
• FR1: Commute Kme 15 – 30 minutes
• FR2: Quality of School (65% or more highschool graduates go
to colleges)
• FR3: Quality of air is good over 340 days a year
• FR4: price of house (4 BR, 3000 x^2, less than 650K)
Town FR1 = FR2=Quality FR3=Quality FR4=Price($)
commute of schools (%) of air (days)
Kme (min)
A 20‐40 50‐70 300‐320 450‐550k
15 20 30 40
Qi Van Eikema Hommes
6/24/10 45
Comparison of AxiomaKc Design
with Other Methods (Suh, 2001)
• Robust design cannot be accomplished by
applying the Taguchi method if the design
violates the Independence Axiom.
• OpKmizaKon of a bad design may lead to an
opKmized bad design or minor improvements.
• How is AxiomaKc Design similar/different from
QFD?
Qi Van Eikema Hommes
6/24/10 46
QuesKons about the Axioms
• Too good to be true? What about constraints?
• Are interacKons so bad? That’s what makes a
system great!
– DefiniKon of System‐‐A combinaKon of interacKng
elements organized to achieve one more stated
purposes.
Qi Van Eikema Hommes
6/24/10 47
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Constraints
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 48
Matrix Representation of a Network
--The Design Structure Matrix (DSM)
DSM is the adjacency matrix of a network graph
K
F J H D
J Level 1
E
A H
F I C
B
B
E G Level 2
C D
G Level 3
K
I A Level 4
10 B pillar
Glass runs
3 2 Glass
A
A pillar 9 Belt seals joint with glass
1 Belt seals
Mirror sail Regulator arms
6 Below belt retainer
Sheet Metal
Access hole
7
8 Equalizer channel
Motor
Electrical System
Qi Van Eikema Hommes
6/24/10 51
Car Door System Engineering Process (Before
ParKKoning DSM)
Sheet Metal
Subsystem
Electrical
Subsystem
Qi Van Eikema Hommes
6/24/10 52
Car Door System Engineering Process
(Axer ParKKoning)
Electrical Packaging
Qi Van Eikema Hommes
6/24/10 53
The Control Soxware System
• 1 production-level software
• 117 software modules (red dots)
• 1423 binary interactions (black lines)
• 39 such production
software releases
per year
• <2 weeks
per release
• What can I do about this web of interacKons?
• How can I convince management that changes are needed?
• How do I know I actually improved the architecture?
Embedded Software
System B
Embedded
Software
System A
(Sosa, Eppinger, Rowles 2007, Borgatti, Everett, and Freeman, 2002, UCINET)
Freeman
Freeman Freeman
Freeman Freeman
Freeman Freeman
Freeman
Centrality Centrality Centrality Centrality
Matrix 4 Centrality
inindegree
Centrality
out
Centrality Centrality
degree outdegree
degree Overall
Overallinin Overall
Overallout
out
In degree—how many others pass AA
AA BB CC
11 11 11
DD EE FF GG HH
11 33 33
BB 11 11 11 11 33 33
information to the element of CC 11 11 11 11 33 33
DD 11 11 11 11 33 33 0% 0%
interest. EE 11 11 11 11 33 33
0% 0%
FF 11 11 11 11 33 33
Out degree—how many others GG 11 11 11 11 33 33
HH 11 11 11 11 33 33
depend on the element of interest Matrix 5
for information. AA
AA BB CC DD EE FF GG HH
11 11 11 11 11 11 11 11 71 71
BB 11 11 11 11
CC 11 11 11 11
the systems. AA BB CC DD EE FF GG HH
AA 11 11 11 22 22
BB 11 11 11 22 22
CC 11 11 11 22 22
However, the metrics values don’t DD 11 11 11 22 22 0%
0% 0%
0%
EE 11 11 11 22 22
correlate well with components FF 11 11 11 22 22
GG 11 11 11 22 22
modularity. HH 11 11 11 22 22
Matrix 8
(Sosa, Eppinger, Rowles 2007, Borgatti, Everett, and Freeman, 2002, UCINET)
• Network centrality metrics can idenKfy the few
elements that have the largest impact on the
system.
• If the network has central players, the network
may be bus‐modular.
• If the network does not have central player, the
network system is either not connected, or highly
integral.
• Central players can be the priority for system
complexity reducKon strategy.
4/8/2010 Copyright Qi D. Van Eikema Hommes 60
DSM Method
• Capture system interacKons
• Analyze and improve system architecture and
system interfaces.
Qi Van Eikema Hommes
6/24/10 61
Lecture Outline
IntroducKon to AxiomaKc Design
Four domains
Axiom 1—Independence Axiom
Design Matrix
Zigzagging
Constraints
Axiom 2—InformaKon Axiom
Design Structure Matrix for Technical
Systems
DM—DSM Method
Qi Van Eikema Hommes
6/24/10 62
ExisKng Methods Concerning System
InteracKons
Design Axiomatic Requireme What
Structu Design’s nts We
re Design Manageme Want
Matrix Matrix nt
(DSM) (DM)
Provide analytical Yes Yes
system analysis
Allow iterations and Yes Yes
feedback loops
Relate the Yes Yes
requirements to the
system design
Can be applied in the Yes Yes Yes
early design phases
Provide complete Yes Yes
understanding of all
requirements
6/24/10
Qi Van Eikema Hommes 63
Solving System of Linear EquaKons
Question: 3 * x1 + 5 * x2 = 6 (1)
2 * x1 - x2 = 4 (2)
What is x1 and x2?
Solving by substitution:
Select x1 as the output variable in (1):
x1 = (6 – 5 * x2 ) / 3
Select x2 as the output variable in (2):
x2 = 2 * x1 – 4 = 2 * (6-5*x2)/3 - 4
x1=2 x2=0
Qi Van Eikema Hommes
6/24/10 64
ConverKng a DM into a DSM
DP1 DP2 DP3
1. Construct an Axiomatic
FR1 X 0 X
Design’s Design Matrix.
FR2 X X 0
FR3 0 X X
2. Select Output Variables.
DP3 = f (FR1, DP1) DP1 DP2 DP3
DP1 = f (FR2, DP2)
DP2 = f (FR3, DP3) FR1 X 0 X
FR2 X X 0
FR3 0 X X
3. Permute the matrix by row
so that the output variables DP1 DP2 DP3
are on the diagonal. We
get a precedence matrix DP1 X X 0
Parameters. DP3 X 0 X
Qi Van Eikema Hommes
6/24/10 65
SelecKng Output Variables
DP2 DP3
DP4
DP2 DP3
DP4
Qi Van Eikema Hommes
6/24/10 66
CVC Cluster Machines
Central
Wafer
Handler
Wafer
Processing
Module
Qi Van Eikema Hommes
6/24/10 67
Courtesy of KDF electronics. Used with permission.
CVC Electro‐staKc Chuck (ESC)
Process Chamber
Wafer
Backside gas channel
Electro-statically charged
plate
Process Process
Module Module 3
2 ESC Wafer ESC
Transport
Robot
Process Process
Module Module
ESC 4 ESC
Control
and
Logistics
Qi Van Eikema Hommes
6/24/10 69
Design Structure Matrix Built from
Design Matrix
Heat Transfer
Heat Transfer
Package the ESC into the
Packaging
Existing Processinto the
Modules
Modules
Control
ControlCircuit
Circuit
Design
Qi Van Eikema Hommes
6/24/10 70
The SelecKon of Output Variables
Qi Van Eikema Hommes
6/24/10 71
The SelecKon of Output Variables
DSM
Qi Van Eikema Hommes
6/24/10 74
OASIS Major Subsystems
Primary Slide
Sample Aliquot Sample
Incubator Integrity
Frame and Metering Buffer Cuvette
(SLIN) (SAIN)
Cabinetry (SRME) (ALBU) Incubator
(STRU) (CUIN)
Secondary
Sample Sample
Handling Metering
(SAHA) (SRME)
Reagent
Supply
Power
(RGSU)
Distributio
n (POWR)
Machine Control
(MACO) Slide
Vitros Tips
Application Supply Photometer Microtip
(PHMT) Loader
Services (APPS) (SLSU) Loading (VTLD)
(MTLD)
Qi Van Eikema Hommes
6/24/10 75
Case Study ObjecKves
1. Build a DSM from requirements using the DM‐
DSM conversion method;
2. Compare the resulKng DSM with the DSM
experts built using tradiKonal DSM construcKon
method.
3. Understand which types of requirements can be
used to predict system interacKons. Judge
whether the predicKon DSM is complete.
4. Aid the system integraKon manager’s work on
planning and managing OASIS subsystem
interfaces.
Qi Van Eikema Hommes
6/24/10 76
DSM Constructed from Requirements
Qi Van Eikema Hommes
6/24/10 77
Compare Requirements DSM with Expert DSM
Qi Van Eikema Hommes
6/24/10 80
The Achievable PotenKal
JNJ experts miss DSM from
6 marks. requirements misses
215 26 marks.
Providing:
• JNJ engineers involve soxware engineers in the
DSM building exercise
• Chemists write assay requirements
• JNJ updates the trace‐ability between product
level requirements and subsystem level
requirements
Qi Van Eikema Hommes
6/24/10 81
LimitaKon of the Method
Requirements DM DSM
Qi Van Eikema Hommes
6/24/10 82
Requirements DecomposiKon
Can predict system Cannot predict
interactions system interactions
Qi Van Eikema Hommes
6/24/10 83
Comparison of the Three Methods
AxiomaKc Design DSM DM - DSM
Matrix (DM) J
H
J
J
H D F
H
E G C I B K A
coupling when a
x x
A x x A
Qi Van Eikema Hommes
6/24/10 86
MIT OpenCourseWare
http://ocw.mit.edu
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.