2882 Spring 2005

2.
882
System Design and Analysis based on
AD and Complexity Theories
References
(1) Nam Pyo Suh, Axiomatic Design: Advances
and Applications, Oxford University Press, New
York, 2001
(2)Nam Pyo Suh, Complexity: Theory and
Applications, Oxford University Press, New
York, 2005
(3) Nam P. Suh, The Principles of Design, Oxford
University Press, 1990
Your name
Your field
Why?
Format/Assumptions
1. Active Learning
2. Project execution
3. Will assume no prior knowledge of
Axiomatic Design and Complexity
Theory.
Lecture 1
Introduction to
Axiomatic Design
Major Topics to be covered

1. Axiomatic Design
Theory
Applications
Many industrial examples
Actual design exercise
2. Complexity Theory
Theory
Applications
Todays Lecture
1.
Introduction -- Read Chapter 1 of AD
2.
Will email Homework Problems
Why Axiomatic Design

1. Engineering deals with design and
manufacture of complex systems
2. Examples: Space Shuttle
Microsoft Operating Systems
Manufacturing Systems
Materials
Organizations
Demands in Industry
Industrial competitiveness demands that
1. Shorten the lead-time for the introduction
of new products,
2. Lower manufacturing cost,
3. Improve the quality and reliability of
products,
4. Satisfy the required functions most
effectively.
Hardware, software, and systems must be designed right to be
controllable, reliable, manufacturable, productive, and otherwise
achieve their goals. The performance of poorly designed hardware,
software and systems cannot be improved through subsequent
corrective actions.
Relationship between design and analysis

1.
Feedback loop between analysis and

synthesis
2.
Scientific paradigm -- reductionism
3.
Synthesis -- Many FRs
Relationship between design and analysis
Figure by MIT OCW.
Y
Y = G ~
~ G = H-1 for GH >> 1
X
1+GH GH
Typical Approach to Realization and

Implementation of New Products
%
C om p l
100 %
80 %
T im e
Poor Planning, purely experience-based

design decisions, and trial-and-error based
design practice may lead to the following
consequences:
1. Project failure
2. Missed schedule
3. Cost over-run
4. High warranty cost
5. Frequent maintenance
6. Me, too product
7. Unhappy customer
Product Development: Typical Approach
%
Compl
Cost of
D evelp.
100 %
80%
Tim e
Product Development:Axiomatic Approach

%
Compl
Cost of
Develp.
100 %
80%
Time
TMA Projection System
Several slides describing TMA projection

removed for copyright reasons.
The MIT CMP machine

Our attempt to teach systems design
4 S.M. students designed and manufactured the machine and the
control system (including software for system integration) in 2
years. The system operated -- as designed -- when turned on with
minimal modification.
1 Ph.D. student studied the CMP process.
Spent $2 million -- Funded by an industrial firm.
What we taught them was the principles of design, so no
debugging or testing of prototypes was needed.
Copper Damascene Process

Cu 6
Photo removed for

copyright reasons.
Cu 5
Cu 4
Cu 3
Cu 2
Cu 1
W1
Reference: D. Edelstein et al., Tech. Dig. IEEE Int. Elec. Dev. Mtg., Washington D.C.,
pp. 773-776 (1997).
History
Goal
To establish the science base for
areas such as design and
manufacturing
How do you establish science base in

design?
Axiomatic approach
Algorithmic approach
Axiomatic Design
Axiomatic Design applies to all

designs:
Hardware
Software
Materials
Manufacturing
Organizations
Axiomatic Design
Axiomatic Design helps the design
decision making process.
Correct decisions
Shorten lead time
Improves the quality of products
Deal with complex systems
Simplify service and maintenance
Enhances creativity
Axiomatic Design
Axioms
Corollaries
Theorems
Applications
--
manufacturing, materials, etc.
System design
Complexity
hardware,
software,
LCD Projector Design
Several slides removed for copyright reasons.

See Example 3.4 in Suh, Axiomatic Design (2001).
Introduction
Stack of modules
Track
Robot
Loading
Station
Stack of modules
Unloading
station
System integration
Stack of modules
S ta ck o f m o d u le s
Track
Robot
Loading
Station
Stack of modules
M a ch in e A
F ig u re 3
S ta ck o f m o d u le s
M a ch in e B
A C lu ste r o f tw o m a ch in e s th a t a re p h y sica lly co u p le d to

m a n u fa c t u r e a p a r t .
Engine Design
Consider spark-ignition internal
engine used in passenger cars.
1.
combustion
Is the IC engine a good design?
2.
What are the functional requirements (FRs)
of an IC engine?
3.
How would you improve the design?
Functional Requirements of
a Spark-Ignition IC Engine
1. Maximize fuel efficiency
2. Eliminate hydrocarbon emission
3. Minimize CO emission
4. Minimize NOx emission
Conventional Engine is highly coupled!
There is no way we can satisfy the EPA

regulation on emission without using
catalytic converter.
February 7, 2005 Lecture
Software -- Acclaro
Think functionally first !!
Review
of
problems.
special
homework
Is this knob a good design or a poor design?
Figure removed for copyright reasons.

See Figure 3.1 in Suh, Axiomatic Design (2001).
Is this knob a good design or a poor

design?
What are the functional requirements of

the knob ??
Which is a better design?

Milled Flat
End of the
shaft
Slot
Milled Flat
End of the
shaft
Metal
Shaft
Injection
molded
n y lon Knob
(b)
(a)
Section view AA
Solution: The one on the right. Why?

Milled Flat
End of the
shaft
Slot
Milled Flat
End of the
shaft
Metal
Shaft
Injection
molded
n y lon Knob
(b)
(a)
Section view AA
Typical Design Process

Marketplace
Product attributes
Functional requirements
and constraints
Societal
Need
Recognize and
Formalize
(code)
Analyze
and/or Test
Compare
Reformulate
Ideate and
Create
Shortcomings:
discrepancies,
failure to improve
Figure by MIT OCW.
Product,
prototype, process
Who are the Designers?

How do we design? What is design?
Is the mayor of Boston a designer?
Design Process
1. Know their "customers' needs".
2. Define the problem they must solve to satisfy the needs.
3. Conceptualize the solution through synthesis, which involves the task
of satisfying several different functional requirements using a set of
inputs such as product design parameters within given constraints.
4. Perform analysis to optimize the proposed solution.
5. Check the resulting design solution to see if it meets the original
customer needs.
Definition of Design
Design is an interplay between
what we want to achieve and
how we want to achieve them.
Definition of Design
"What
we want
to
achieve"
"How
we want
to
achieve
them"
Example: Refrigerator Door Design

See E1.1 in Suh, Axiomatic Design (2001).
Mapping from Customer Needs to Functional

Requirements
Example:
Arrow's Impossibility Theorem
Consider the case of having three choices, A, B and

C. Three people were asked to indicate their
preference among these three choices.
Based on the input from these individuals, can we
make a decision as to what the group as a whole
prefers?
Example - Solution
The answer is "No. The following table lists the preferences indicated
by Smith, Kim and Stein:
Individuals
Smith
Kim
Stein
Group preference
Preferences
A>B>C, A>C
B>C>A, B>A
C>A>B, C>B
Choices
A vs. B
B vs. C
A
B
B
B
A
C
A >B
B>C
A vs. C
A
C
C
C>A
The results show that the group is confused as to what it wants. It

prefers A over B, and B over C, but it prefers C over A rather than A
over C as one might have expected based on the first two choices.
Creativity and Axiomatic Design
Axiomatic design enhances creativity

by eliminating bad ideas early and
thus, helping to channel the effort of
designers .
Historical Perspective on Axiomatic Design

Axioms are truths that cannot be derived but for which
there are no counter-examples or exceptions.
Many fields of science and technology owe their
advances to the development and existence of axioms.
(1) Euclid's geometry
(2) The first and second laws of thermodynamics
are axioms
(3) Newtonian mechanics
Axiomatic Design Framework

The Concept of Domains
{CAs}
{FRs}
{DP}
Mapping
Customer domain
Mapping
Functional domain
{PVs}
Mapping
Physical domain
Process domain
Fig. 1.1 Four Domains of the Design World. {x} are characteristic vectors of
each domain
Figure by MIT OCW.
Characteristics of the four domains of the design

world
Domains Character
Vectors
Customer Domain
{CAs}
Functional Domain
{FRs}
Physical Domain {DPs} Process Domain {PVs}
Manufacturing
Attributes which
consumers desire
Materials
Desired performance
Functional
requirements
specified for the
product
Required Properties
Physical variables
which can satisfy the
functional
requirements
Micro-structure
Processes
Software
Attributes desired in
the software
Output Spec of
Program codes
Input Variables or
Algorithms Modules
Program codes
Sub-routines machine
codes compilers
modules
Organization
Customer satisfaction
Functions of the
organization
Programs or Offices
or Activities
People and other

resources that can
support the programs
Systems
Attribute desired of
the overall system
Functional
requirements of the
system
Machines or
components,
sub-components
Resources (human,
financial, materials,
etc.)
Business
ROI
Business goals
Business structure
Human and financial

resource
Table by MIT OCW. After Table 1.1 in [Suh 2001].
Process variables that

can control design
parameters (DPs)
Definitions
Axiom:
An axiom is a self-evident truth or fundamental
truth for which there is no counter examples or
exceptions. It cannot be derived from other laws of
nature or principles.
Corollary:
A corollary is an inference derived from axioms or
propositions that follow from axioms or other
proven propositions.
Definitions - contd
Functional Requirement:
Functional requirements (FRs) are a minimum set of
independent requirements that completely characterize the
functional needs of the product (or software, organizations,
systems, etc.) in the functional domain. By definition, each
FR is independent of every other FR at the time the FRs are
established.
Constraint:
Constraints (Cs) are bounds on acceptable solutions. There
are two kinds of constraints: input constraints and system
constraints. Input constraints are imposed as part of the
design specifications. System constraints are constraints
imposed by the system in which the design solution must
function.
Definitions - contd
Design parameter:
Design parameters (DPs) are the key physical (or
other equivalent terms in the case of software
design, etc.) variables in the physical domain that
characterize the design that satisfies the specified
FRs.
Process variable:
Process variables (PVs) are the key variables (or
other equivalent term in the case of software design,
etc.) in the process domain that characterizes the
process that can generate the specified DPs.
The Design Axioms

Axiom 1: The Independence Axiom
Maintain the independence of the functional
requirements (FRs).
Axiom 2:
The Information Axiom
Minimize the information content of the

design.
Example 1.3 Beverage Can Design

Consider an aluminum beverage can
that contains carbonated drinks.
How many functional requirements
must the can satisfy?
How many physical parts does it have?
What are the design parameters

(DPs)? How many DPs are there?
Design Matrix
The relationship between {FRs} and {DPs} can be
written as
{FRs}=[A] {DPs}
When the above equation is written in a differential form
as
{dFRs}=[A] {dDPs}
[A] is defined as the Design Matrix given by elements :
Aij = FRi/DPi
Example
For a matrix A:
A11 A12 A13
[ A] = A21 A22 A23
A31 A32 A33
Equation (1.1) may be written as

FR1 = A11 DP1 + A12 DP2 + A13 DP3
FR2 = A21 DP1 + A22 DP2 + A23 DP3
FR3 = A31 DP1 + A32 DP2 + A33 DP3
(1.3)
Uncoupled, Decoupled, and Coupled Design

Uncoupled Design
A11 0
[ A] = 0 A22
0
0
0
0
A33
(1.4)
Decoupled Design
0
A11 0
[A] = A21 A22 0
A31 A32 A33
Coupled Design
All other design matrices
(1.5)
Design of Processes
{DPs}=[B] {PVs}
[B] is the design matrix that defines the

characteristics of the process design and is
similar in form to [A].
Constraints
What are constraints?
Constraints provide the bounds on the acceptable
design solutions and differ from the FRs in that
they do not have to be independent.
There are two kinds of constraints:
input constraints
system constraints.
New Manufacturing Paradigm Robust Design
Theorem 4 -- Ideal design
Example: Shaping of Hydraulic Tubes
To design a machine and a process that can achieve

the task, the functional requirements can be
formally stated as:
FR1= bend a titanium tube to prescribed
curvatures
FR2= maintain the circular cross-section of the
bent tube
Tube Bending Machine Design (conts)
Given that we have two FRs,

how many DPs do we need?

See Figure E1.6 in Suh, Axiomatic Design (2001).

DP1= Differential rotation of the bending rollers to bend the tube
DP2= The profile of the grooves on the periphery of the bending
rollers
Fixed set of
counter-rotating
grooved rollers
1= 2
Tube between
the two rollers
Pivot
axis 1
1<2
Figure ex.1.4.a
Flexible set of
counter-rotating
grooved rollers
for bending
Tube bending apparatus
Example: Van Seat Assembly

(Adopted from Oh, 1997)
Figures removed for copyright reasons.


Traditional SPC Approach to Reliability and Quality
The traditional way of solving this kind of problem has been to do
the following:
(a) Analyze the linkage to determine the sensitivity of the
error.
Table a Length of linkages and sensitivity analysis
Links
L12
L14
L23
L24
L27
L37
L45
L46
L56
L67
Nominal Length (mm)

370.00
41.43
134.00
334.86
35.75
162.00
51.55
33.50
83.00
334.70
Sensitivity (mm/mm)
3.29
3.74
6.32
1.48
6.55
5.94
11.72
10.17
12.06
3.71

(b) Assess uncertainty
measurement.
through
prototyping
and
The manufacturer of this van measured the distance between

the front to rear leg span as shown in Fig. ex.2.5.d. The mean
value of FR is determined to be 339.5 mm with a standard
deviation of f. Then, we can fit the data to a distribution
function. If we assume that the distribution is Gaussian, then the
reliability is given by
Reliability =
346
334
1
2 F
___
(FR FR) / 2 F2
dFR
The data plotted in Fig. ex.2.5.d yields a reliability of 95%.
(a)

c) Develop fixtures and gages to make sure
(critical
dimensions are controlled carefully.
that the
(d) Hire inspectors to monitor and control the key

characteristics using statistical process control (SPC).
New Manufacturing Paradigm Robust Design

This design has one FR, i.e., F, the front to rear leg span. This is a function of 10
DPs, i.e., 10 linkages. This may be expressed mathematically as
F = f (DP , DP ,.... DP
1
10
10
f
f
i
x
DP
DP
+
F =
i
x
DP
DP
i =1, except i= x
What we want to do is to m ake F=0
Decomposition, Zigzagging and Hierarchy
DP
FR
FR1
DP1
FR2
FR11
FR12
FR121
FR122
FR123
FR1231
DP2
DP11
DP12
DP121
DP122
FR1232
Functional Domain
DP123
DP1231
DP1232
Physical Domain
Figure by MIT OCW.
Figure 1.2 Zigzagging to decompose in the functional and the physical domains and
create the FR- and DP hierarchies
Identical Design and Equivalent Design

Equivalent Design:
When two different designs satisfy the same set of the
highest-level FRs but have different hierarchical
architecture, the designs are defined to be equivalent
designs.
Identical Design:
When two different designs satisfy the same set of
FRs and have the identical design architecture, the
designs are defined to be identical designs.
Example: Refrigerator Design
FR1 = Freeze food for long-term preservation

FR2 = Maintain food at cold temperature for shortterm preservation
To satisfy these two FRs, a refrigerator with two
compartments is designed.
Two DPs for this
refrigerator may be stated as:
DP1 = The freezer section
DP2 = The chiller (i.e., refrigerator) section.
FR1 = Freeze food for long-term preservation

FR2 = Maintain food at cold temperature for short-term
preservation
DP1 = The freezer section
DP2 = The chiller (i.e., refrigerator) section.
FR1 X 0DP1
FR2 0X DP2

Having chosen the DP1, we can now decompose FR1 as:
FR11 = Control temperature of the freezer section in
the range of -18 C +/- 2 C
FR12 = Maintain the uniform temperature
throughout the freezer section at the preset
temperature
FR13 = Control humidity of the freezer section to
relative humidity of 50%

FR11 = Control temperature of the freezer section in the range of -18 C +/- 2 C
FR12 = Maintain the uniform temperature throughout the freezer section at the preset temperature
FR13 = Control humidity of the freezer section to relative humidity of 50%
DP11 = Sensor/compressor system that turn on and off

the compressor when the air temperature is
higher and lower than the set temperature in the
freezer section, respectively.
DP12 = Air circulation system that blows air into the
freezer section and circulate it uniformly
throughout the freezer section at all times
DP13 = Condenser that condenses the moisture in the
returned air when its dew point is exceeded
Similarly, based on the choice of DP2 made, FR2

may be decomposed as:
FR21 = Control the temperature of the chilled
section in the range of 2 to 3 C
FR22 = Maintain a uniform temperature
throughout the chilled section within 1 C of
a preset temperature

FR21 = Control the temperature of the chilled section in the range of 2 to 3 C
FR22 = Maintain a uniform temperature throughout the chilled section within 1 C of
a preset temperature
DP21 = Sensor/compressor system that turn on and

off the compressor when the air temperature
is higher and lower than the set temperature
in the chiller section, respectively.
DP22 = Air circulation system that blows air into the
freezer section and circulate it uniformly
throughout the freezer section at all times


The design equation may be written as:
FR12 XOO DP12

FR11 = XXO DP11
FR13 XOX DP13
Equation (a) indicates that the design is a decoupled design.
FR22
FR21
DP22
DP21
X
X
0
X
Full DM of Uncoupled Refrigerator Design
DP1
DP2
____________________________________________________
DP12 DP11 DP13 DP22 DP21
____________________________________________________
FR12 X
0
0
0
0
FR1
FR11 X
X
0
0
0
FR13 X
0
X
0
0
_____________________________________________________
FR2
FR22 0
0
0
X
0
FR21 0
0
0
X
X
_____________________________________________
Full DM of Uncoupled Refrigerator Design
DP1
DP2
____________________________________________________
DP12 DP11 DP13
DP22 DP21
____________________________________________________
FR12
X
0
0
0
0
FR1
FR11
X
X
0
0
0
FR13
X
0
X
0
0
_____________________________________________________
FR2
FR22
X
0
0
0
0
FR21
0
0
0
X
0/X
_____________________________________________
Crew survivability system for the Orbital

Space Plane
Design of Crew Survivability System for

OSP
The highest-levels of FRs were decomposed
to develop
the detailed design of TPS, Landing System,
and Sensing System for Meteorite Damage.
High-level Decomposition
Functional Requirements (FR)
Design Parameters (DP)
Ensure crews survive launch ascent into

Orbit
Crew survivability systems
[FR1] Ensure crews survives pre-launch

FR1.1 Determine system readiness
[DP1] Launch-pad survivability system

DP1.1 System interface testing and
initialization
DP1.2 Passive threat protection
systems
DP1.3Threat response systems
FR1.2 Provide passive protection

from threats
FR1.3 Respond to threat
[FR2] Ensure crews survive phase I of
ascent (from liftoff to CESP staging)
[DP2] Phase I survivability system
[FR3] Ensure crews survive phase II of

ascent (from CESP staging to OSP
separation)
[DP3] Phase II survivability system
High-level Decomposition
(Acclaro, Courtesy of ADSI)
Courtesy of Axiomatic Design Solutions, Inc. Used with permission.
Design Matrix
(Software - Acclaro, Courtesy of ADSI)
Design Outcome (selected examples)
Design of Low Friction Sliding Surfaces

without Lubricants
What are the FRs?
What are the constraints?

without Lubricants
FR1 = Support the normal load
FR2 = Prevent particle generation
FR3 = Prevent particle agglomeration
FR4 = Remove wear particles from the
interface
Constraint: No lubricant
Friction at Dry Sliding Interface

Undulated Surface for Elimination of Particles

See Figures 7.11 & 7.13 in Suh, Complexity (2005).

without Lubricants
The design equation:
FR1 X 000 DP1 X 000 A

FR
DP 0 X x0 R
0
0
X
x
2

2
=
=
FR3 00 X 0 DP3 00 X 0
FR4 000 X DP4 000 X V
Suggested Solution
Transform the system with time-dependent
combinatorial complexity
to
a system with time-dependent periodic complexity.
2.882 System Design and Analysis

February 14
Taesik Lee 2004
What well do today

Information content
Robustness
Taesik Lee 2004
Review
Design process / Domain
Functional Requirements
Design Parameters
Taesik Lee 2004
Review
Design Domain
How do you go about design?

How
does our
product/system
satisfy it?
What do
customers
want?
What do we
want to
achieve?
How
do we want to
achieve it?
What are the

solutions to
be
generated?
Taesik Lee 2004
How
do we want to
generate it?
Review
Functional Requirement (FR)
Functional requirements (FRs) are a minimum set of

independent requirements that completely characterize
the functional needs of the artifact (product, software,
organization, system, etc.) in the functional domain.
Independently achievable, in principle.

The determination of a good set of FRs requires skill, ,
and many iterations
Taesik Lee 2004
Review
What constitute a GOOD set of FRs?
MECE: Mutually Exclusive and Collectively Exhaustive

One FR carries only one requirement
Solution neutral
Cover the battery contact by a plastic sliding door
Clarity/specificity
Juxtaposing two requirements into one doesnt make them a

single requirement
Bad example: missing target range, time factor, etc.
Logical hierarchical structure
Taesik Lee 2004
Review
Design Parameter (DP)
Design parameters (DPs) are the key physical (or other

equivalent terms in the case of software design, etc.)
variables in the physical domain that characterize the
design that satisfies the specified FRs.
DPs of a aluminum beverage belonging to the
main body piece
- Body thickness
- Shape of the bottom
-
Taesik Lee 2004
Review
Searching for a DP
Nothing substitutes for Knowledge

Be open to wild idea
Analogies
Visualization
Benchmarking, Patents, Literatures, etc. in OTHER application

area
Sketch your idea
Stimuli
Circulating ideas
Get exposed to foreign situations
Taesik Lee 2004
Review
Design Hierarchy
Decomposition by zigzagging
Process of developing detailed requirements and concepts by

moving between functional and physical domain
Yields a hierarchical FR-DP structure
FR1
FR11
FR111
FR1111
DP1
FR12
FR112
FR1112
FR1211
FR121
FR1212
DP11
FR122
DP111
DP1111
DP12
DP112
DP1112
DP1211
:
Taesik Lee 2004
DP121
DP1212
DP122
Review
Decomposition
Zigzagging
Process of breaking down a large problem into a set of smaller

ones so that each of the sub-problems is manageable
Decomposition process must involve both functional and
physical domain by moving between the two domains
Upper-level choice of DP drives the FRs at the next level
Lower-level FRs are a complete description of

functional needs of the upper-level FR-DP pair
Parent-Child relationship
Taesik Lee 2004
Information content
Design range
System range
Probability of success
(Allowable) Tolerance
Taesik Lee 2004
Information
Contents
Design Range
Examples of range in FR statements

-
Maintain the speed of a vehicle at a designated mph in the

range of 0mph - 90mph
Maintain the speed of a vehicle at a x mph +/- 5mph
Ensure no leakage under pressure up to 100 bar
Ensure 99% successful ignition at the first attempt in the
temperature range of -30C ~ 80C
Generate nailing forces of up to 2,000 N
Taesik Lee 2004
Information
Contents
Information
Contents
Design Range
Specification for FR
Acceptable range of values of a chosen FR metric; Goal-post
Different from tolerance
Different from operating range
Target value (nominal), Upper bound, Lower bound
FR*
FR
Between x and y
FR
Greater than x
(Larger the better)
Taesik Lee 2004
- (or 0)
FR
Smaller than x
(Smaller the better)
Information
Contents
System Range
Response/performance in FR domain, resulting from the chosen

design
Here, design includes both a chosen set of DPs and the way they
deliver/affect FRs
Due to various factors such as the input (DP) variation,

internal/external noise, etc., FR takes a range of values, forming
a range
p.d.f.
f(FR1,FR2)
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
FR
|sr|
FR2
Taesik Lee 2004
FR1
Information
Contents
Information content
dr u
P ( FR ) =
f ( FR ) dFR
dr l
dru
I = log2 P = log2 P( FR) = log2
f (FR)dFR
drl
|dr|
Design
Range
p.d.f.
f(FR)
p.d.f.
f(FR)
Common
Range, AC
System
Range,
p.d.f. f(FR)
drl
dru
System Range at time t0,

f(FR, t0)
FR
drl
|sr|
dru
Design Range |dr|
Where do we get f(FR)?

Taesik Lee 2004
FR
Information content for multiple FRs
I ( FR1 , FR 2 , L , FR n ) = log 2 p1, 2,..., n
f ( FR , FR
p1, 2,..., n =
, L , FR n ) dFR1 dFR 2 L dFR n
design hyperspace
If Uncoupled,
p1, 2,..., n =
L f ( FR , FR
1
dr1dr 2
,..., FRn )dFR1 dFR2 ...dFRn
drn
f ( FR )dFR
1
dr1
( FR2 )dFR2 L
dr 2
( FR1n )dFRn
drn
= p1 p 2 ... p n
I ( FR1 , FR2 , L , FRn ) = I ( FR1 ) + I ( FR2 ) + ... + I ( FRn )

Taesik Lee 2004
Information
Contents
Information
Contents
If Decoupled,
p1, 2,..., n =
L f ( FR , FR
1
dr1dr 2
,..., FRn )dFR1 dFR2 ...dFRn
drn
L f ( FR
dr 1dr 2
FR1 , FR 2 ,..., FR n 1 ) f ( FR n 1 FR1 , FR 2 ,..., FR n 2 ) L
drn
L f ( FR 2 FR1 ) f ( FR1 ) dFR1 dFR 2 ...dFR n
FR1 a O DP1
=
DP
FR
b
c
2
2
For example,
p1, 2 =
f ( FR , FR
1
)dFR1 dFR2
dr1dr 2
f ( FR
FR1 ) f ( FR1 )dFR1 dFR2
dr1dr 2
f
(
FR
FR
)
dFR
2
1
2 f ( FR1 ) dFR1
dr1dr2
Taesik Lee 2004
Information
Contents
1
b
DP2 = FR2 DP1
c
c
1
b
= FR2 FR1
a
c
FR1 a O DP1
=
FR
2 b c DP2
Let g1 and g2 be the pdf of DP1 and DP2.
From
f ( FR ) =
1 FR
g
a a
We get f(FR1) and f(FR2|FR1) :
1 FR
f ( FR1 ) = g1 1
a a
Since
b
1 FR2 a FR1
f ( FR2 | FR1 ) = g 2
c
c
p1, 2 = f ( FR2 FR1 )dFR2 f ( FR1 )dFR1

dr1dr 2
p1, 2
FR2 b FR1
1
a
1 g1 ( FR1 )dFR1
= g2
dFR
2
c
a
c
a
dr1 dr 2
Taesik Lee 2004
Information
Contents
Multiple FR system range
Example
FR1 1 0 DP1
FR 2 1 1 DP 2
DP2
Design range
FR1: [-0.5 , 0.5]
FR2: [-2.0 , 2.0]
FR2
DP1
FR1
Taesik Lee 2004
Allowable tolerance
Information
Contents
Defined for DP
Tolerances that DPs can take while FRs still remaining completely
inside design ranges
Unconditional tolerance
Conservative tolerancing
0 DP1
FR1 A11 0

DP2
FR
2
A
21
A
22
0
=
FR3 A31 A32 A33DP3
FR1
A11
FR2 A21 DP1
DP2 =
A22
FR3 A31 DP1 A32 DP2
DP3 =
A33
DP1 =
Taesik Lee 2004
Information
Contents
Allowable tolerance
FR2
DP2
Design range
X
X
O
X
Allowable tolerance area

DP1
FR1
Design range is mapped

onto DP space
DP1 =
DP2 =
FR1
A11
FR2 A21 DP1
A22
Taesik Lee 2004
Detecting change in system range
Monitoring marginal probability of each FR is not

only inaccurate but potentially misleading
Example
Design range
FR1 1 0 DP1
FR 2 1 1 DP 2
FR1: [-0.5,0.5]
FR2: [-2,2]
Design parameter variation

Initial
DP1: [-1,1]
DP2: [0,1.5]
After change
DP1: [-1,1]
DP2: [-1,1.6]
Taesik Lee 2004
Information
Contents
Information
Contents
FR2
FR2
2.5
2
Joint p.d.f.
(FR1,FR2)
1.5
FR2
2.6
2
2.5
1
0.5
-1
1
-1
Joint p.d.f.
(FR1,FR2)
.6
p.d.f
0.5
FR1
FR2
-1
Design range
0.3846 p.d.f
FR1
-1
Design range
-2
(b)
-2
(a)
(b)
p.d.f
p.d.f
DP1: [-1,1]
DP2: [0,1.5]
0.5
-1
0.6
-1
-1
-2
(a)
2.6
1
(c)
0.5
-1
FR1
1
(c)
Before DP2 change
FR1
DP1: [-1,1]
DP2: [-1,1.6]
After DP2 change
pFR1
pFR2
pFR1 pFR2
pFR1,FR2
Before
0.5
0.9583
0.4792
0.5
After
0.5
0.9654
0.4827
0.499
Taesik Lee 2004
Information
Contents
Summary
Joint probability unless it is uncoupled design

Assuming DPs are statistically independent, working in DP domain
is typically easier.
FR
DP
FR
DP
FR
FR
DP
DP
FR
FR
FR
FR
FR
FR
FR
DP
DP
DP
DP
DP
DP
DP
Taesik Lee 2004
Information
Contents
Information
In AD information content, by imposing design
range, FR is transformed into a binary random
variable.
ui =
1 (success) with
0 (failure)
|dr|
Design Range
Common
Range,
AC
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
P(FRi = success)
1-P(FRi = success)
drl
dru
FR
|sr
|
I(ui= 1) - log2P(FRi =success)
H(X) = - pi log2 pi = E[I]

H (P )
I(p)
7
6
5
H (p ) [bit]
I (p ) [bit]
4
3
2
1
0
0
0.2
0.4
0.6
0.8
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
P (FR =success)
0.2
0.4
0.6
0.8
P (FR =success)
(a)
(b)
Taesik Lee 2004
Robustness
Robustness
In axiomatic design, robust design is defined as a design that always

satisfies the functional requirements,
FRi > FRi
even when there is a large random variation in the design parameter DPi.
Two different concepts in robustness

Insensitive to noise
Information Axiom
Traditional robust design
Adaptive to change
Independence Axiom
Hod Lipson, Jordan Pollack, and Nam P. Suh, "On the Origin of Modular Variation", Evolution,
Evolution, 56(8) pp. 1549-1556, 2002
Taesik Lee 2004
Robustness
Example: Measuring the Height of a House with a Ladder
H
Angle =
Taesik Lee 2004
Robustness
Solution:
H + H = sin ( + )L = (sin cos + cos sin ) L
For small ,
H + H = sin L + Lcos
H = Lcos
where is the mean value of the estimated angle, L the length of the ladder, and H the
height.
Carefully selecting parameter values can make a system much more robust at almost no
additional cost.
Taesik Lee 2004
Robustness
Example
FR = DP2 (1 DP)
FR
0.0315
0.0787
DP
DPb = 0.259924
Taesik Lee 2004
DPc = 0.943877
Robustness
How to make a system robust?
Where does the variation come from?
FR
r
FR
FR FR =
n +
n nr =0
DP
*
( DP DP ) +
DP = DP
Taesik Lee 2004
FR
C
(C C )
C =C
Robustness
FR
r
FR
FR FR =
n +
n nr =0
DP
*
( DP DP ) +
DP = DP
FR
C
0. Assign the largest possible tolerance

0. Eliminate the bias ( E[FR] = FR* )
1.
2.
3.
Eliminate the variation: SPC, Poka-Yoke, etc.

De-sensitize: Taguchi robust design
Compensate
FR
r
FR
(C C ) =
n +
n r
C C =C *
DP
n =0
FR
Taesik Lee 2004
( DP DP )
*
DP = DP
(C C )
C =C
Robustness
Bring the system range back into design range: Reinitialization

Example: Design of Low Friction Surface
Dominant friction mechanism: Plowing by wear debris
System range (particle size) moves out of the desired design range
Need to re-initialize
Figure removed for

copyright reasons.
Figure removed for

copyright reasons.
N. P. Suh and H.-C. Sin, Genesis of Friction, Wear, 1981
S. T. Oktay and N. P. Suh, Wear debris formation and

Agglomeration, Journal of Tribology, 1992
Taesik Lee 2004
Robustness
Design of Low Friction Surface
Periodic undulation re-initializes the system range
Figures (6-part diagram and two graphs)

removed for copyright reasons.
S. T. Oktay and N. P. Suh, Wear debris formation and agglomeration, Journal of Tribology, 1992
Taesik Lee 2004

February 16
Taesik Lee 2004
What well do today

Project discussion
Information content, Robustness
Taesik Lee 2005
Term Project Overview

Key dates
Today: Project topic discussion, kick-off
[ ~ 6 wks]
April 4: Interim progress report

[ ~ 5 wks]
May 11: Project presentation

May 16: Written project report due
Deliverables
Conceptual design solution
AD/Complexity analysis
Presentation, report
Taesik Lee 2005
Project Topic Presentation
Taesik Lee 2005
Project Examples from the previous year

Engine project
CEV architecture project
Taesik Lee 2005
Information content
Design range
System range
Probability of success
(Allowable) Tolerance
Taesik Lee 2005
Information
Contents
Review
Design Range
Examples of range in FR statements

-
Maintain the speed of a vehicle at a x mph +/- 5mph

Ensure no leakage under pressure up to 100 bar
Specification for FR
Acceptable range of values of a chosen FR metric; Goal-post
Different from tolerance
Different from operating range
Target value (nominal), Upper bound, Lower bound
Taesik Lee 2005
Review
System Range
Response/performance in FR domain, resulting from the chosen

design
Here, design includes both a chosen set of DPs and the way they
deliver/affect FRs
Due to various factors such as the input (DP) variation,

internal/external noise, etc., FR takes a range of values, forming
a range
p.d.f.
f(FR1,FR2)
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
FR
|sr|
FR2
Taesik Lee 2005
FR1
Review
Information content
dr u
P ( FR ) =
f ( FR )dFR
dr l
dr u
I = log2 P = log2 P(FR) = log2
f (FR)dFR
dr l
|dr|
Design Range
Common Range,
AC
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
drl
dru
|sr|
Taesik Lee 2005
FR
Review
Example
FR1 1 0 DP1
FR 2 1 1 DP 2
DP2
Design range
FR1: [-0.5 , 0.5]
FR2: [-2.0 , 2.0]
FR2
DP1
FR1
Taesik Lee 2005

Example
Design range
FR1 1 0 DP1
FR 2 1 1 DP 2
FR1: [-0.5,0.5]
FR2: [-2,2]

Initial
DP1: U[-1,1]
DP2: U[0,1.5]
After change
DP1: U[-1,1]
DP2: U[-1,1.6]
Taesik Lee 2005
Information
Contents
Information
Contents
FR2
FR2
2.5
2
Joint p.d.f.
(FR1,FR2)
1.5
FR2
2.6
2
2.5
1
0.5
-1
1
-1
Joint p.d.f.
(FR1,FR2)
.6
0.5
FR1
FR2
p.d.f
-1
Design range
0.3846 p.d.f
FR1
-1
Design range
-2
(b)
-2
(a)
(b)
p.d.f
p.d.f
DP1: U[-1,1]
DP2: U[0,1.5]
0.5
-1
0.6
-1
-1
-2
(a)
2.6
0.5
-1
FR1
(c)
FR1
DP1: U[-1,1]
DP2: U[-1,1.6]
(c)
Before DP2 change
After DP2 change
pFR1
pFR2
pFR1 pFR2
pFR1,FR2
Before
0.5
0.9583
0.4792
0.5
After
0.5
0.9654
0.4827
0.499
Taesik Lee 2005
Review
Allowable tolerance
Defined for DP
FR2
DP2
X
X
Design range
O
X
FR1
DP1 =
DP2 =
DP1
FR1
A11
FR2 A21 DP1
Taesik Lee 2005
A22

onto DP space
Linear tolerancing vs. Statistical tolerancing

FR1 1 0 DP1
FR 2 0.4 1 DP2
Linear tolerancing
(-0.6,2.04)
Design range
FR1: [-0.6,0.6]
FR2: [-1.8,1.8]
Statistical tolerancing
DP2
(0.6,1.56)
-0.6
0.6
DP1
3FR1 = 0.6 FR1 = 0.2

Therefore, DP1 = 0.2
Allowable tolerance
Var(FR2) = 0.42Var(DP1) + Var(DP2)

Thus, DP2 = 0.5946
DP1: [-0.6,0.6]
DP2: [-1.56,1.56]
3 DP1 = 0.6
3 DP2 = 1.784
Taesik Lee 2005
Review
Robustness
In axiomatic design, robust design is defined as a design that always
satisfies the functional requirements,
FRi > FRi
even when there is a large random variation in the design parameter DPi.
Two different concepts in robustness

Insensitive to noise
Information Axiom
Traditional robust design
Adaptive to change
Independence Axiom
Hod Lipson, Jordan Pollack, and Nam P. Suh, "On the Origin of Modular Variation", Evolution,
Evolution, 56(8) pp. 1549-1556, 2002
Taesik Lee 2005
Review
L
H
Angle =
H + H = sin L + Lcos
H = Lcos
What if L also has uncertainty?
Taesik Lee 2005
Review
r
FR
FR FR =
n +
n nr =0
DP
*
FR
( DP DP ) +
DP = DP
FR
C
0. Assign the largest possible tolerance

0. Eliminate the bias ( E[FR] = FR* )
1.
2.
3.
Eliminate the variation: SPC, Poka-Yoke, etc.

De-sensitize: Taguchi robust design
Compensate
FR
r
FR
(C C ) =
n +
n r
C C =C *
DP
n =0
FR
Taesik Lee 2005
( DP DP )
*
DP = DP
(C C )
C =C
Robustness
Robustness built into a system by design

Graph and diagram removed for copyright reasons.

Taesik Lee 2005
Two diagrams removed for copyright reasons.

Robustness

Two figures (6-part diagram and pair of graphs) removed for copyright reasons.
Taesik Lee 2005
February 28
Taesik Lee 2004
What well do today
Information content for multi-FR
Basic statistics/probability
Allowable tolerance (linear tolerancing) vs. statistical

tolerancing
Taesik Lee 2005
Review
Information content
dr u
P ( FR ) =
f ( FR )dFR
dr l
dr u
I = - log2 P = - log2 P(FR) = - log2
f (FR)dFR
dr l
|dr|
Design Range
Common Range,
AC
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
dru
drl
|sr|
Taesik Lee 2005
FR
Normal Distribution
We want to know this area

(probability)
p.d.f.
f(FR)
F(z), value from the standard

normal distribution function
-1s
X ~ N(m, s2)
1s
FR
Then, Z = (X-m)/s ~ N(0, 1)

z
1
2
3
4
5
6
phi(z)
0.84134474
0.977249938
0.998650033
0.999968314
0.999999713
0.999999999
1-phi(z)
prob between +z, -z
0.15865526
0.68268948
0.022750062
0.954499876
0.001349967
0.997300066
3.1686E-05
0.999936628
2.87105E-07
0.999999426
9.90122E-10
0.999999998
Taesik Lee 2005
ppm
ppb
317310.5
45500.12
2699.934
63.37207
0.57421
574.21
0.00198 1.980244
HW1, #5
System range, FR1 ~ N(m,s2)

Design range drl =< FR1 =< dru
l = (dru + drl)/2
Q: Information Content ?
l
p.d.f.
f(FR)
N(m,s2)
drl
(drl+dru)/2
l
u)/2 + l
drl - (dr+dr
In terms of s multiple: {drl - (drl+dru)/2 + l}/s
Taesik Lee 2005
dru
FR
dru (drl+dru)/2 + l
{dru (drl+dru)/2 + l}/s
Review
Example
FR1
1
0
DP1
FR 2
1
1
DP2
Design range
FR1: [-0.5 , 0.5]
FR2: [-2.0 , 2.0]
Design range
DP2
FR2
DP1
FR1
DP variation (joint pdf) is

mapped onto FR space;
only a projection is shown
here
Assuming statistical
independence between
DP1 and DP2, the joint pdf
of (DP1,DP2) is a product
of pdf(DP1) and pdf(DP2)
Taesik Lee 2005

Example
Design range
FR1
1
0
DP1
FR 2
1
1
DP2
FR1: [-0.5,0.5]
FR2: [-2,2]

Initial
DP1: U[-1,1]
DP2: U[0,1.5]
After change
DP1: U[-1,1]
DP2: U[-1,1.6]
Taesik Lee 2005
Information
Contents
Information
Contents
FR2
FR2
2.5
2
Joint p.d.f.
(FR1,FR2)
1.5
-1
FR2
2.6
2
2.5
1 pFR2
0.5
1
0.5
FR1
-1
FR2
Marginal
pdf (FR2)
Joint p.d.f.
(FR1,FR2)
.6
p.d.f
-1
Design range
Marginal
pdf (FR1)
0.5
pFR1
-1
-1
-2
(b)
-2
(a)
p.d.f
DP1: U[-1,1]
DP2: U[0,1.5]
Marginal
pdf (FR1)
0.5
pFR1
-1
FR1
(c)
FR1
DP1: U[-1,1]
DP2: U[-1,1.6]
(c)
Before DP2 change
After DP2 change

Wrong
Correct
pFR1
pFR2
pFR1 pFR2
pFR1,FR2
Before
0.5
0.9583
0.4792
0.5
After
0.5
0.9654
0.4827
0.499
Taesik Lee 2005
Marginal
pdf (FR2)
0.3846 p.d.f
FR1
Design range
(b)
p.d.f
0.6 p
FR2
-1
-1
-2
(a)
2.6
Review
Allowable tolerance
Defined for DP
FR2
DP2
X
X
Design range
FR1
DFR1
DDP1 =
A11
DFR2 - A21 DDP1

DDP2 =
A22
Taesik Lee 2005
DP1
onto DP space
Linear tolerancing vs. Statistical tolerancing
FR1
1 0
DP1
FR2
0.4 1
DP2
Linear tolerancing
(-0.6,2.04)
Design range
FR1: [-0.6,0.6]
FR2: [-1.8,1.8]
Statistical tolerancing
DP2
(0.6,1.56)
-0.6
0.6
DP1
(-0.6,-1.56)
3sFR1 = 0.6 sFR1 = 0.2

Therefore, sDP1 = 0.2
Allowable tolerance
Var(FR2) = 0.42Var(DP1) + 12 Var(DP2)

Thus, sDP2 = 0.5946
DP1: [-0.6,0.6]
DP2: [-1.56,1.56]
3 sDP1 = 0.6
3 sDP2 = 1.784
Taesik Lee 2005
MIT, 2.882 Spring 2005, Taesik Lee
Supplementary Lecture Note on

Information Content
2.882
Spring 2005
by Taesik Lee
1. Information content for one-FR problem
We all know the definition of the information contents in Axiomatic Design. For a oneFR design,
dr u
I = log 2 P = log 2 P( FR) = log 2
f ( FR)dFR
(1)
dr l
P(FR) is a probability of success for the functional requirement, i.e. the probability that a
value of FR output by the designed system falls within the acceptable range. f(FR) is a
probability density function (or probability mass function in the case that the FR is a
discrete variable) of the FR. Given f(FR), integrating the function over the acceptable
range, from drl to dru, gives the probability of success. A visual illustration of this is the
following graph that we have seen many times in the class and the textbook.
|dr|
Design Range
Common Range,
AC
p.d.f.
f(FR)
System Range,
p.d.f. f(FR)
dru
drl
FR
|sr|
Figure 1. Probability of success is obtained by integrating the probability density function of the FR,
f(FR), over the Design Range. The area of the hatched section is equal to the value of the integration.
Example: Computing probability of success for an FR ~ N(,2)

Normal distribution (a.k.a. Gaussian distribution) is probably the most popular
distribution among many engineers. It is a bell-shaped curve centered on a mean, ,
with variance, 2. It is common to see outputs of a manufacturing process/system
being assumed to follow a normal distribution. Use of normal distribution in
statistical modeling is not an arbitrary choice of a modeler. It is justified by the fact
that when many independent random factors act in an additive manner to create
variability, data will follow the normal distribution. In statistics, this is formalized as
the Central Limit Theorem.
If a random variable, X, follows the normal distribution, N(,2), the probability of X
being between a and b is the area, A, of the hatched section in Figure 2. A can be
computed by subtracting A2 from A1 as shown in the figure.
p.d.f.
f(FR)
=
a
FR
A1 b
FR
FR
A2 a
Figure 2. Computing probability of X being between a and b by considering two

integrations. The area, A equals to (A1 A2).
A1 and A2 can be easily obtained by using a standard normal distribution table found
in most probability and statistics textbooks. Standard normal distribution is a normal
distribution where a random variable, Z, follows normal distribution with its mean at
0 and a variance of 1, i.e. Z ~ N(0,1). The standard normal distribution table shows
the values (numerical approximations) for integrals of the normal distribution
function.
( z ) = P( Z z ) =
f ( z )dz
(2)
If we imagine that Figure 2 is drawn for Z, the probability that a Z b is
P (a Z b) = A1 A2 = (b) ( a )
(3)
and the value (a) and (b) can be found from the standard normal distribution
table.
In order to use the standard normal distribution table, the random variable X~N(,2)
has to be transformed to Z~N(0,1). That can be done by the following transformation:
Z=
(4)
Thus, for any FR~N(,2), we can use the standard normal distribution table to
compute the probability of success.
For example, suppose we have designed a system that yields FR ~ N(6,1). The design
range is given as 4FR7. See Figure 3.
Design Range
p.d.f.
f(FR)
FR
Figure 3. FR ~ N(6,1). The design range is indicated by two vertical lines.
Substituting the given values into equation (4), we get Z+ = 1 and Z- = -2. Thus, the
probability of success (the hatched area in Figure 3) can be computed as
76
46
P (4 FR 7) = P
Z
1
1
= P(2 Z 1) = (1) (2)
(5)
Note that (-2) equals to 1-(2) from the symmetry of the normal distribution.
Equation (5) becomes,
P (4 FR 7) = P (2 Z 1) = (1) (1 ( 2))
(6)
Reading from the standard normal distribution table, we get (1) = 0.8413 and (2)
= 0.9772. Therefore, P(4FR7) = 0.8185.
As you may have noticed, transforming X = x to Z = z is equivalent to identifying how far

x is from the mean of X in the unit of its standard deviation. In the above example, what
we saw was that FR=7 is (+1) apart from its mean and FR=4 is (-2) apart, thus Z+=1
and Z-=-2.
It is common to hear someone saying our systems quality performance is at 3 sigma
level. This means (with little possibility of different interpretation) that the probability
of getting a good output from the system is 99.73%. In other words, the system range
follows a distribution such that the design range corresponds to +/- 3 of the distribution.
You can see from Table 1 that +/- 3 level performance would yield about 2700 bad
outputs per 1 million outputs (or ~3 bad outputs per 1,000 outputs). As the quality level
goes up, say to +/- 6 level, it becomes ~2 bad outputs per billion outputs. That is why
the term six sigma has become a catchphrase for the popular quality improvement
initiative.
Table 1. Sigma value and its quality level. For 3 level, it is on the order of one bad output per
thousand; for 6, it is on the order of one ppb(part per billion).
z
1
2
3
4
5
6
phi(z)
0.84134474
0.977249938
0.998650033
0.999968314
0.999999713
0.999999999
1-phi(z)
prob between +z, -z
0.15865526
0.68268948
0.022750062
0.954499876
0.001349967
0.997300066
3.1686E-05
0.999936628
2.87105E-07
0.999999426
9.90122E-10
0.999999998
ppm
ppb
317310.5
45500.12
2699.934
63.37207
0.57421
574.21
0.00198 1.980244
It should be noted that equation (1) does not include any information (information as in
common use of the term) on the design parameter at all. The information content is only a
function of the distribution of the functional requirement and the design range. As the
design range is given also in a functional domain, you can see that the information
content is defined solely in a functional domain. How we achieve the FR is behind the
scene, and the information content addresses the resulting performance, i.e. how well the
functional requirement is satisfied, of the complete system. Therefore, in the strictest
sense, the true information content can only be measured by observing the output of a
complete system. This, in some sense, makes the information content less useful in
practice: if we can measure the information content only after the system design is
complete, how does it help us in the design phase? Although computing the true value of
information content is impossible before the design is finished, some qualitative
conclusions derived from the concept turn out to be useful in a design process. For
example, means to reduce the information content or the relationship between the
information content and coupling hold true regardless of our ability to compute the value
of the information content. Also, estimating information content from the knowledge
about design parameters and design matrix can be useful in comparing potential solutions.
If we assume that the variation of FRs is entirely attributed to the variation of DPs, then
we can estimate the information content of the FR given the knowledge about the DP. A
design equation for one-FR design problem is,
FR = [ A]DP = f (DP)
(7)
Given the distribution of DP and the mapping relationship from DP to FR (A or f), we can
derive the distribution of FR, a system range (Figure 4). This can be done by either the
distribution function technique or the change-of-variable technique [refer to any statistics
textbook such as the one by Hogg & Tanis]. Although it is particularly easy when the
mapping is linear or strictly in(de)creasing, it still takes some effort to do the derivation.
Instead, it is much easier to map the design range to physical domain (Figure 5) because
the design range is simply an interval rather than a function. For example, in linear
mapping, it is simply either stretching or shrinking the design range. This turns out to be
particularly convenient for multi-FR problems, which is discussed in the next section.
Distribution of
Design Parameter
p.d.f. g(DP)
p.d.f. f(FR)
Derived distribution of
Functional Requirement
DP
FR
Physical Domain
Functional Domain
Figure 4. System range can be derived from the distribution of DP. The task may become non-trivial
depending on the mapping relationship and DP distribution.
Design Range
drl
Allowable tolerance
dru
FR
cl
cu
DP
Physical Domain
Functional Domain
Figure 5. An alternative is mapping the design range onto physical domain, which can be easily done
in most cases. The mapped range in DP domain is called allowable tolerance.
The range in DP domain, mapped from the design range, is called the allowable tolerance
to avoid confusion with the design range while denoting the fact that once a DP value is
within this range, the FR values will be within the design range. The probability of
success for FR is then equal to the probability of DPs falling within the allowable
tolerance. This probability is obtained by integrating the p.d.f. of DP, g(DP), over the
allowable tolerance. (Figure 6)
p.d.f. g(DP)
Design Range
drl
dru
FR
Allowable tolerance
Functional Domain
DP
Physical Domain
Figure 6. Probability of success for FR is equal to the probability of DP's falling within the allowable
tolerance.
2. Information content for multi-FR problem

We can easily extend the definition of information content, equation (1), to multi-FR case:
I ( FR1 , FR 2 , L , FR n ) = log 2 p1, 2,..., n
(8)
where p1,2,,n denotes the probability of success for all of the FRs. p1,2,,n is the
probability that the value of FR1 output by the designed system falls within FR1 design
range, the value of FR2 within FR2 design range, etc. p1,2,,n is defined as
p1, 2,..., n =
f ( FR , FR ,L, FR
1
(9)
)dFR1 dFR2 L dFRn
design hyperspace
As explained in the textbook, it should be emphasized that, unless the design is

uncoupled, we have to make sure the probability of success is properly computed/
estimated using the joint probability.
In computing the information content, the same arguments presented in section 1 remains
valid: 1) under the assumption that the only source of variation is DP, we can calculate
the probability of success given DP variation and [DM], and 2) in doing so, it is much
easier to map the design range onto DP domain than deriving the joint system range from
the DP distributions.
Example: Computing probability of success for two-FR problem
Consider the following two-FR design problem.
FR1 1 0 DP1
FR 2 1 1 DP 2
(10)
The design ranges for FRs are given as -0.5FR10.5 and -2FR22. The
variation of DPs is given by two uniform distribution: DP1 ~ U[-1,1] and DP2
~ U[0,1.5] where U[a,b] denotes a uniform distribution between a and b. Let
g1 denote the probability density function of DP1 and g2 be the p.d.f. of DP2.
Then, g1 and g2 are constant within the two bounds: g1 = 0.5 for -1DP11,
and g2 = 2/3 for 0DP21.5. (see Figure 7)
First, we need to assume that DPs are statistically independent to each other to
simplify the problem. Under the assumption, the joint probability density
function of (DP1,DP2), g1,2, is simply a product of two uniform probabilities.
This, again, yields a uniform p.d.f. in DP1-DP2 plane:
g1,2 (DP1,DP2) = 1/3
0
if, -1DP11 and 0DP21.5

otherwise
Figure 7 is a graphical illustration of obtaining g1,2.
(11)
U[-1,1]
0.5
DP2
h = 1/3
1.5
1 DP1
-1
g12(DP1,DP2)
=
g1(DP1)*g2(DP2)
U[0,1.5]
2/3
h
0
-1
1.5 DP2
DP1
p.d.f. g(DP1,DP2) = h = 1/3
Figure 7. Assuming DP1 and DP2 are statistically independent to each other, the joint
p.d.f. g1,2(DP1,DP2) is a product of g1 and g2. g1,2 is another uniform density function,
i.e. constant 1/3.
Now, we can map the design range onto DP domain to obtain the allowable
tolerance (in this case, an area). The design range is represented by a rectangle
on the FR-plane (Figure 8a). From the design matrix,
FR1 = DP1
FR2 = DP1 + DP2
If we rearrange these equations such that DPs are expressed in terms of FRs,
DP1 = FR1
DP2 = -DP1 + FR2
FR1 design range is given as -0.5FR10.5. Thus, the first equation of the
above represents a vertical line in the area between -0.5DP10.5 as shown in
Figure 8b. The second equation represents a set of straight line with the slope
of -1, and its intersection with DP2 axis ranges from -2 to 2 (given by FR2
design range). See Figure 8c. Finally the allowable tolerance for DP1 and DP2
is obtained by taking intersection of 8b and 8c.
DP2
2
FR2
DP2
DP1
FR1
-0.5
0.5
DP2
DP1
DP1
-0.5
0.5
-2
-2
(a)
(b)
(c)
(d)
Figure 8. Mapping the design range onto DP space to obtain the allowable tolerance.
Alternatively, since the design matrix is constant (linear mapping), we know

that four vertices of the design range will become the vertices of the allowable
tolerance. These vertices can be easily calculated by multiplying the inverse of
the design matrix by the vertices of the design range.
Now that we have the allowable tolerance and the joint probability density
function of (DP1,DP2), the probability of DPs being within the allowable
tolerance is readily computed by integrating g1,2 over the allowable tolerance.
Since g1,2 is constant, the integral is simply the product of g1,2 and the
overlapping area between the allowable tolerance and the projection of g1,2:
Probability = 1/3 * (1*1.5) =
DP2
2
1.5
-1 -0.5
DP joint p.d.f.
0.5
DP1
Allowable
Tolerance
-2
Figure 9. Probability of success is the integral of g1,2 over the allowable tolerance
Assuming DPs are the only source of FR variation, we can say the probability
of success of FR1 and FR2 equals to the probability we computed above.
We know that, if the design matrix is coupled or decoupled, the probability of success for
FRs is not equal to the product of the probability of success for individual FRs: p1,2,,N
p1p2pN. This individual probability, pi, is called the marginal probability (of success) of
FRi. The marginal probability is defined based on the joint probability. For example, in a
2-FR case:
p1 =
dr u , FR1
dr l , FR1
f1 ( FR1)dFR1
(12)
f1 ( FR1) =
1, 2
( FR1, FR 2)dFR 2
f1(FR1) is called the marginal p.d.f. for FR1. Suppose we have a system that has FR1 and
FR2. Les us say we only care about FR1 and measure the FR1 samples to obtain its
distribution. Then we can compute the probability of success for FR1 by examining the
distribution with the FR1 design range. This probability is the marginal probability of
success for FR1.
It should be noted that the probability of success computed by multiplying the marginal
probabilities is not only inaccurate, but potentially misleading in drawing a conclusion
from the observation. The following example illustrates the point.
Example: Marginal probability vs. Joint probability
The two-FR design problem we saw in the earlier example is revisited.
FR1 1 0 DP1
FR 2 1 1 DP 2
(10)
The probability of success was obtained by the joint probability calculation,

and it was 0.5.
In order to compute the marginal probability of FR1 and FR2, we must first
derive the joint p.d.f. for (FR1,FR2). Since the DPs follow uniform
distribution and the mapping is linear, the derivation is easy: the derived joint
p.d.f. will also be that of a uniform distribution. We can map the projection of
the DPs joint p.d.f. onto the FR domain, and the joint p.d.f. of FR will simply
be the uniform p.d.f. over the mapped area. As the integral of any p.d.f. must
be 1, the height of the uniform p.d.f. of (FR1,FR2) is 1/(projection area) = 1/3.
FR2
DP2
h = 1/3
h = 1/2
1.5
2.5
S
L
1
h
0
-1
-1
DP1
p.d.f. g(DP1,DP2) = h = 1/3
FR1
1
-1
p.d.f. f(DP1,DP2) = h = 1/3
Figure 10. Deriving f1,2 from g1,2. DP's having the uniform distribution and its linear
mapping make it simple.
Now, the second equation of equation (12) becomes,
f1 ( FR1) =
f1, 2 ( FR1, FR 2)dFR 2 =
3 dFR 2
for -1FR11
(13)
You may notice that the integral
dFR 2 is the length of the projected area
along FR2 axis (L in Figure 10). The figure suggests that the length is 1.5
throughout the -1FR11 range. Thus,
f1(FR1) = 1/3 * 1.5 = 0.5
0
when -1FR11
otherwise
For FR2,
f 2 ( FR 2) =
1, 2
( FR1, FR 2)dFR1 =
3 dFR1
for -1FR22.5
(14)
dFR1 is the length of the projected area along FR1 axis (S in Figure 10),
and, unlike L, S is a function of FR2. By inspecting Figure 10, you can see
that S linearly increases from 0 to 1.5 as FR2 increases from -1 to 0.5. Then,
for 0.5<FR2<1.0, S is constant at 1.5. From FR2=1.0 to FR2=2.5, S decreases
from 1.5 back to zero. S, a function of FR2, multiplied by 1/3 yields the
marginal p.d.f. f2 that has a trapezoidal shape. Figure 11 summarizes the above
discussion with the design range also shown.
FR2
FR2
2.5
2
Joint p.d.f.
(FR1,FR2)
1.5
-1
2.5
1
0.5
1
0.5
FR1
-1
p.d.f
-1
Design range
-2
(a)
(b)
p.d.f
0.5
-1
1
(c)
FR1
Figure 11. The marginal p.d.f. for FR1 and FR2 can be computed from the joint p.d.f. of
(FR1,FR2). Each of the two marginal p.d.f. is integrated within the design range to give
the marginal probability of success for FR1 and FR2, respectively.
Integrating each of the two marginal p.d.f. over its relevant design range
yields the marginal probability of success. From Figure 11, you can figure out
that p1 = 0.5 and p2 = 0.9583. Product of these two marginal probabilities is
0.4792. As expected, this value is a bit off from the correct value, p1,2 = 0.5.
10
Now suppose that DP2 variation has changed somehow (e.g. change of a
supplier) from U[0,1.5] to U[-1,1.6]. Following the same steps described
above, we can obtain p1,2, p1, and p2. It turns out that p1,2 = 0.499; p1 = 0.5;
p2 = 0.9654. Again, p1*p2 (= 0.4827) is slightly off from the correct value.
FR2
FR2
2.6
2
Joint p.d.f.
(FR1,FR2)
0.6
.6
-1
2.6
0.3846 p.d.f
FR1
-1
-1
Design range
-2
(a)
-2
(b)
p.d.f
0.5
-1
1
(c)
FR1
Figure 12. Joint p.d.f. of (FR1,FR2) and the marginal p.d.f.'s for FRs.
What is important here is that even though the probability of success by using
the joint p.d.f. shows a hint of performance degradation (p1,2 < p1,2) , the other
probability of success by using two marginal probabilities indicates the
opposite trend (p1*p2 > p1*p2). This can be a significant problem when
detecting mal-functioning is time-critical for recovering the system. Therefore,
as emphasized earlier, it is very important to consider the joint p.d.f. when
there are multiple functional requirements that are not completely uncoupled.
This section explained how to compute information contents for multiple-FR problem
using 2-FR uniform distribution examples. Applying the same technique to three or
more-FR problems with general distribution will not be effective. Indeed, when it is not
the simplest case, computing information contents is perhaps best handled by a
probabilistic simulation technique such as Monte Carlo simulation.
11
Chapter 4.
Design of Systems
Introduction
What is a system?
CMP machine
Engineering Systems & Bio System
Space Shuttle, Mars Rover,

etc
System
Sense of sight, smell and

taste
Nervous system
[Figures of spacecrafts
removed for copyright
reasons.]
Element
[Figures of human
biology removed for
copyright reasons.]
Eyeball
Subsystem
Individual parts
Cellular structure in
the retina
Component
:
Leaf
Chemistry in these cells
Definition of a System:
An assemblage of
sub-systems,
hardware and software components, and
people
designed to perform a set of tasks so as to
satisfy the specified FRs
and constraints.
Examples of Systems
Software
Machines
Manufacturing systems
Materials
Products
Government
Issues Related to System Design

1. How should a complex system be designed?
2. How should the complex relationships

between various components of a system be
coordinated and managed?
Issues Related to System Design

3. How can the stability and controllability of a
system be guaranteed?
4. What is the role of human operators in a

system?
Classification of Systems
Are all systems alike?
How should we classify systems?
Should the system be classified based on the
physical size of the system?
or
Should it be based on the number and nature of
the functions
that the system must perform?
Why does the classification of systems
based on functions
rather than physical size
make more sense?
Large systems from Small systems
Static systems from Dynamic systems
Fixed systems from Flexible systems
Passive systems from Active systems
Open systems from Closed systems
Axiomatic Design Theory

for Fixed Systems
Is the design of systems different from the

design of other things?
Axiomatic Design Theory

for Fixed Systems
i. The First Step in Designing a System: Define FRs of the
System
ii. Mapping between the Domains: a Step in Creating
System Architecture
iii. The Independence of System Functions
iv. Information Content for Systems: the Best Design
v. Decomposition
vi. System architecture
Measurement of Information content
How do we measure the information content of

a system that has many decomposed layers in its
hierarchy?
Information Content of Systems

How do we measure the information content of a
system that has many decomposed layers in its
hierarchy?
Isystem = Ihighest level FRi = - log (Ac)highest level FRi
where (Ac)highest level Fri is the area of the common

range associated with each one of the highest level
FRi.
Information Content of a Large System
Isystem= Ihighest level FRi= - log (Ac)highest level FRi
where (Ac)highest level FRi is the area of the common range

associated with each one of the highest level FRi.
How do we determine Isystem when

Ihighest level FRi is not known?
Isystem = log(p leaf) = - log (Ac)leaf
where (Ac)leaf is the area of the common range associated with each leaf.
log (Ac)leaf = log (Ac)highest level FRi
In the case of a coupled design, it is expected that in

most cases,
log (Ac)leaf < log (Ac)highest level FRi

since any change in any other FR in the same set of FRs
at a given level will affect the Ac.
Information associated with physical

integration (i.e., assembly)
Isystem = log(p leaf) + Ia = - log (Ac)leaf + Ia

where Ia is the information associated with assembly of
modules.
log (Ac)leaf + Ia < log (Ac)highest level FRi
Definition of Module
FR1 a 0 DP1
FR2 b cDP2
Definition of Module --- Mi
FR1 = a DP1 = M1 * DP1
FR2 = b DP1 + c DP2 = M2 * DP2
where M2 = b (DP1 / DP2) + c.
Definition of Modules
FR1 = a DP1 = M1 * DP1
FR2 = b DP1 + c DP2 = M2 * DP2
Decomposition of {FRs}, {DPs}, and {PVs}
How does the decomposition process affect the

outcome of the design process?
Definition S1
(Equivalent Designs)
Two designs are defined to be

"equivalent"
if they satisfy the same set of the highest level
FRs within the bounds established by the same
set of constraints,
even though the mapping and decomposition
process might have yielded designs that have
substantially different lower level FRs and all
DPs for each of these designs.
Definition S2: (Identical Designs)

Designs that fulfill the same set of the highest
level FRs and satisfy the Independence Axiom
with zero information content are defined to be
"identical" if their lower level FRs and all DPs
are also the same.
Theorem S1 (Decomposition and System

Performance)
The decomposition process does not affect

the overall performance of the design if
the highest level FRs and Cs are satisfied
and if the information content is zero,
irrespective of the specific decomposition
process.
Theorem S2: (Cost of Equivalent Systems)
Two "equivalent" designs can have a

substantially different cost structure,
although they perform the same set of
functions and they may even have the
same information content.
Design and Operation of Large Systems

What is a Large System?
The telephone system for Boston,
The government bureaucracy,
An assembly plant for automobiles,
A software system that controls nuclear power
plants, and
Boeing 747 airplanes
Design and Operation of Large Systems

Is it the physical size, the number of
components, or the number of functions that
make it large?
What is a Large Flexible System?

Definition of a Large Flexible System
A system is a large flexible system if the
total number of FRs that the system must
satisfy during its lifetime is large and if at
different times, the system is required to
satisfy different subsets of FRs.
Axiomatic Design of a Large Flexible System

How do we design a large flexible system?
Define FRs and Constraints
Knowledge base -- DPs for FRs
Develop design concepts -- A set of DPs for the
design task
Physical integration
Develop alternative designs
Choose the best based on information measure
The knowledge base can be structured as follows:

FR1 $ (DP1a, DP1b,...................., DP1m)
FR2 $ (DP2a, DP2b,...................., DP2q)
FR3 $ (DP3a, DP3b,...................., DP3w)
............................................................
............................................................
FRn $ (DPna, DPnb,...................., DPns)
Synthesis of a Large Flexible System
SAAB,
Developer of Defense Airplanes of Sweden
SAABs Research: Letter from Exec. VP,

Professor Billy Fredriksson
We have an interesting ongoing research on product
development utilizing your Axiomatic Design Theory.
Building on your theory Gunnar Holmberg (PhD-student
from Saab) is using AD on high systems architecture level to
design systems for life cycle flexibility. This is in order to
efficiently add new unknown functionality through the life of
the systems.
It would be interesting and valuable to us to discuss this with
you. I am planning to go to US in April and plan to be at MIT
23 April. Would you be available at MIT?
Synthesis of a Large Flexible System

Suppose the subsets of FRs change as a function of time as
follows:
@ t = 0,
the subsets are

{FRs}0 = {FR1, FR5, FR7, FRn}
@ t = T1, {FRs}1 = {FR3, FR5, FR8, FRm}

@ t = T2, {FRs}2= {FR3, FR9, FR10, FRn}
How shall we choose DPs?
System Synthesis through Physical Integration

of DPs
How do we combine the lower-level DPs to
synthesize the higher-level DPs?
V-Model
System
Needs
Biological
System
Establish
Interfaces
Identify
Molecular
Entities
Decompose
hy
Define
Modules
In
te
gr
at
e
(B Ph
ot ys
to ica
m
-U l En
p)
tit
c
ar
er
Hi n)
P
-D ow
FR p-D
ild (To
Bu
Map to
DPs
ies
Determine
System
Morphology
Define
FRs
Map DPs to
Biological Entities
Figure1. V-model overview of system analysis using the Design Matrix. The V-Model
describes how the Design Matrix of AD is used to study hierarchical nature of biological
systems
System Design & Development

TECHNICAL CREDIBILITY
CREDIBLE COST ASSESSMENT

Technical Scope Defined
is Scope Estimated
NASA Requirements
(Level 1 & 2)
Customer
Needs
Identifies Lowest
Level Requirements
& Interactions
Define
FRs
Estimate the
Systems Physical
Solutions
Satisfy system
morphology
System changes
are assessed
Construct local
assemblies
Map to
DPs
Build FR -DP
hierarchy
(Top -down)
Establish
interfaces
Decompose
Define
Modules
Identify
physical
components
Mapping DPs into

physical entities
Detailed
system
Integrate
physical entities
(Bottom -up)
The V-Model
Customer
Attributes
Define FRs
Establish Interfaces
Mapping
ch
ar
ier h)
eh c
ar roa
ftw pp
so n A
the ow
ild - D
Bu (Top
y
CA Domain
Bu
ild
th
(B e o
ott
b
om ject
o
-U
p A rien
pp ted
ro
ac mod
h)
el
Coding with System

Architecture
Software
Product
Identify classes
Decomposition
Module
Definition
Identify Leaves
(Full Design Matrix)
FR / DP Domain
DP / PV Domain
Theorem S3 (Importance of High Level

Decisions)
The quality of design depends on the selection
of FRs and the mapping from domain to
domain. Wrong selection of FRs made at the
highest levels of design domains cannot be
rectified through the lower level design
decisions.
Theorem S4 (The Best Design)

The best design for a large flexible system
that satisfies n FRs
can be chosen among the proposed designs
that satisfy the Independence Axiom
if the complete set of the subsets of {FRs}
that the large flexible system must satisfy
over its life
is known a priori.
Theorem S5 (The Need for a Better Design)

When the complete set of the subsets of {FRs}
that a given large flexible system must satisfy
over its life
is not known a priori,
there is no guarantee that
a specific design will always have the minimum information
content for all possible subsets and thus, there is no
guarantee that the same design
is the best at all times.
Theorem S6 (Improving the probability of

success)
The probability of choosing the best design for a large
flexible system increases as the known subsets of {FRs}
that the system must satisfy approach the complete set
that the system is likely to encounter during its life.
Theorem S7 (Infinite
Completeness)
Adaptability
versus
The large flexible system with an infinite

adaptability (or flexibility) may not represent the
best design when the large system is used in a
situation where the complete set of the subsets of
{FRs} that the system must satisfy is known a priori.
Theorem S8 (Complexity of a Large Flexible

System)
A large system is not necessarily complex if it has a
high probability of satisfying the {FRs} specified for
the system.
Theorem S9 (Quality of Design)

The quality of design of a large flexible system is
determined by the quality of the database, the
proper selection of FRs, and the mapping process.
Representation of the System

Architecture of Fixed Systems
Is there a need to represent the system

architecture?
How do we present it in a concise manner?
Three different but

representing a system:
equivalent
ways
of
(1) FR/DP/PV hierarchies with corresponding design

matrices,
(2) Module function diagram, and
(3) Flow diagram.
Hierarchies in Design Domains through

Decomposition of {FRs}, {DPs}, and {PVs}:
Representation of the System Architecture
What is a design hierarchy and how does that

represent the system architecture?
Suppose that we have completed a system design such

that the FR and the DP hierarchies are:
FR1 A11
=
FR2 0
0 DP1
A22 DP2
FR11 X O DP11
=
FR12 X X DP12
FR21 X

FR22 = X

FR23 0
0
X
0
0 DP21
0 DP22
X DP23
FR121 X

FR122 = X
FR123 X
0
X
0
0 DP121
0 DP122
X DP123
FR1231 a 0DP1231
FR1232 b cDP1232

See Figure 4.1 in Suh, Axiomatic Design (2001).
Modules
FR1231 = a DP1231 = M1231 * DP1231
FR1232 = b DP1231 + c DP1232 = M1232 * DP1232
where M1232 = b (DP1231 / DP1232) + c.
Design Matrix and Module-Junction Diagrams Another

Means of System Representation
Since there can be many modules distributed throughout a system, how can
we represent the inter-relationship among modules in a system design?

See Figures 4.2-4.6 in Suh, Axiomatic Design (2001).
Summation Junction
Control Junction
M2
M21
M214
C
M213
M212
M22
S
M211
Hardware Module
Software Module
M23
S
M1
M12
M121
M122
M123
M11
M1232
M112
M12323
S
M111
M1231
M12322
M12321
System Control Command (SCC)
How do we operate a system?
Example 4.1 Design of Wafer Processing Equipment

Vapor
Prime
VP chill
Adhesion promoter
Coating
Photoresist film
Soft Bake
SB chill
Solvent evaporation
Substrate
Developed image
(negative resist)
HB chill
Hard Bake
Developing
Chemical reaction
in exposed area
PEB chill
PEB
(Post Exposure
Bake)
Exposure
Example 4.1 Design of Wafer Processing Equipment
Stack o f mo dules
Trac k
Robo t
Loading
Station
Stack o f mo dules
Unloading
statio n
Example 4.1 Design of Wafer Processing

Equipment
Constraints are:
Cost
Footprint
Reliability
Safety
Serviceability
Manufacturability
Contamination
Minimization of wafer temperature variation
The Highest Level FRs, DPs, and the Design

Matrix
FR1 = coat wafers with

desired resist film at desired
throughput rate
FR2 = develop exposed film
at desired rate
FR3 = transport wafer from
input pt. to modules to
output pt.
FR4 = control the system
functions
X O O X DP1 = coating process modules
O X O X DP2 = developing process modules

X X X X DP3 = transport system
O O O X DP4 = system architecture
The design equation has a triangular matrix:

FR4 = control the system functions
FR1 = coat wafers with desired resist
film at desired throughput rate
FR2 = develop exposed film at desired
rate
FR3 = transport wafer from input pt.
to modules to output pt.
X O O O DP4 = system architecture

X X O O DP1 = coating process
modules
X O X O DP2 = developing process
modules
X X X X DP3 = transport system
Description of the Decisions Made

FR1 : Coating thickness = 0.5 0.8m
Coating uniformity within wafer = 15 A (3)
wafer-to-wafer = 10
cassette-to-cassette(24Hr.) = 13A
FR2 : Critical Dimension (C.D.)
within wafer = 0.010m
wafer-to-wafer = 0.005m
cassette-to-cassette(24Hr) = 0.010m
FR3 : various flow capability
Transfer time 10 sec - Overhead time should
not be a throughput limiter
FR4 : process recipe/plan generating high-level
command (e.g. On/Off, Accel/Decel)
The Second Level FRs, DPs, and the Design Matrix
Parent FR : FR1 = coat wafers with desired resist film at desired throughput rate
Parent DP: DP1 = coating process modules
Constraints : Decomposition of DP1 must not affect FR2, FR4
FR11 = prepare wafer for coating X O O DP11 = thermal process module1
FR12 = coat the wafer with resist O X O DP12 = (N1) spin coater
FR13 = complete coating process O O X DP13 = thermal process module2
Description of the Decision Made at the Second

Level
FR11 = wafer surface adhesion, surface temperature
uniformity before coating, etc. (This will be taken into
account at the decomposition of FR11/DP11)
FR12 = various photoresist capability to produce
uniform and repeatable film (same spec. as with FR1)
FR13 : final resist film thickness control. (This will be
taken into account when FR13/DP13 are decomposed.)
Master Matrix Chart
FR11
FR12
FR13
FR21
FR22
FR23
FR31
FR32
FR41
FR42
FR43
DP11
DP12
DP13
DP21
DP22
DP23
DP31
DP32
DP41
X
O
O
O1
O
O1
X
X3
O
O5
O*
O
X
O
O
O2
O
X
O4
O
O5
O*
O
O
X
O1
O
O1
X
X3
O
O5
O*
O1
O
O1
X
O
O
X
X3
O
O5
O*
O
O2
O
O
X
O
X
O4
O
O5
O*
O1
O
O1
O
O
X
X
X3
O
O5
O*
O
O
O
O6
O
O
X
X
O
O
O
O
O
O
O6
O
O
O
X
O
O
O
X
X
X
X
X
X
X
O
X
O
X
DP42 DP43
O
O
O
O
O
O
O
O
O
X
O
X
X
X
X
X
X
O
X
O
O
X
Notes :
1. Thermal effect must be considered among
thermal process modules - we can use some
kinds of thermal shields or we can do that
with appropriate layout.
2. Spin module must not affect each other in the
sense of vibration, particle generation, etc.
3. The evidence of this X is the utilization of
IBTA robot.
4. Standard method for wafer hand-off is required.
5. Flexible or standard way of supervising
(sensing)
6. There must be no delay from stepper to
thermal process module 3.
M1
M11
M12
M4
M41
M13
M43
M3
S
M42
M21
M22
M2
M23
Flow chart at the second level
M31
M32
M1
M11
M111
S
M112
M12
M121
M122
M127
M124
M123
M126
M125
M131
M13
S
M132
Figure b Flow Chart [FR1x/FR1xx]
M2
M21
M211
S
M212
M22
M221
M222
M226
C
S
M223
M224
M126
M225
M23
M231
S
M232
Flow Chart [FR2x/FR2xx]
Mathematical Modeling, Simulation, and

Optimization of Systems
Application of the Flow Diagram of the System

Architecture
What is the system architecture good for?
Application of the Flow Diagram of the System

Architecture
1. Diagnosis of system failure
2. Engineering change orders
3. Job assignment and management of a system
development team
4. Distributed systems
5. System design through assembly of modules
6. System consisting of hardware and software
Exam Revisit (I)
Diagonal term vs. Off-diagonal term
Matrix rearrangement
FR1 X
FR 2 X

FR3 O
FR 4 O
O
X
X
X
O
X
X
X
O DP1
O DP 2
O DP3
X DP 4
Taesik Lee 2005
Exam Revisit (II)
Allowable tolerance / Probability of Success
DPa
DPb
2 0 2 2 0
FR1 2 1
DPc
FR 2 = 1 = 0.5 1 0 0 1 0
DPd
FR3 3 0.1 0.2 0 0 3 0.5

DPe
DPf
FR 2 2 0.5 0 0 DPa
FR1 = 1 = 1 2 0 DPd
FR3 3 0.1 0 0.5 DPf
DP 2 + = 2FR 2 + = 0.2
DP1+ = 0.5DP 2 + + 0.5FR1+ = 0.05
DP3 = 0.2DP 2 + 2FR3 = 0.16
Taesik Lee 2005
Exam Revisit (III)

DP1
DP1
0.2
0.19
0.16
0.05
0.1
-0.2
0.2
-0.1
DP2
0.1
-0.2
-0.1 -0.05
0.05
0.2
DP2
-0.05
-0.19
-0.2
DP1
DP3
0.2
Joint p.d.f. for (DP1,DP2)
0.16
0.1
0.1
0.1
-0.2
0.2
-0.1
DP2
0.1
-0.2
0.2
-0.1
DP2
-0.1
-0.1
Joint p.d.f. for (DP2,DP3)

A
Taesik Lee 2005
-0.2
Design of Manufacturing Systems
Taesik Lee 2005
What is a manufacturing system?
Photo removed for copyright reasons.
1910 Ford Motor Company
Taesik Lee 2005
2010 Semiconductor Fab
Design of fixed manufacturing systems

for discrete identical parts
Small Scale Problems
Taesik Lee 2005
. Simple deterministic scheduling problem

C
F
A
Robot
IN OUT
Part
Process module for X
X
Design a manufacturing
system
to eliminate the root cause
of a problem (symptom)
In/Out buffer
Taesik Lee 2005
Photoresist processing
Vapor
Prime
VP chill
Adhesion promoter
Coating
Photoresist film
Soft Bake
SB chill
Solvent evaporation
Substrate
Developed image
(negative resist)
HB chill
Hard Bake
Developing
Taesik Lee 2005
Chemical reaction
in exposed area
PEB chill
PEB
(Post Exposure
Bake)
Exposure
Deterministic scheduling problem

Machine diagram removed for
copyright reasons.
F
Robot
IN OUT
Part
In/Out buffer
Taesik Lee 2005
Process
Time (sec)
# of
modules
40
20
17
60
15
40
35
Level 1
FRs
#.1
#.2
Perform process steps

with desirable quality
Satisfy process flow
and throughput
DPs
Process modules
System configuration
FR1 X
FR2 =
X DP1
X DP2
Level 2
FRs
#.1
Manage the recipe
#.2
Support the system

physically
Move wafer when
process is over
#.3
DPs
Recipe handling
module
System layout
Transport system
Taesik Lee 2005
FR2.1 X
FR2.2 = O

FR2.3 X
O
X
O DP2.1
X DP2.2
X DP2.3
Level 3 - Sub FRs/DPs of FR2.1

FRs
#.1
#.2
#.3
Keep TAKTprocess below

TAKTsystem
Maintain # of moves by main
robot not to degrade target
throughput
Locate process modules into
200-APS frame
DPs
Number of each process
module
Number of IBTA
Layout (module
arrangement)
FR2.2.1 X
FR2.2.2 = O

FR2.2.3 X
O
X
O DP2.2.1
X DP2.2.2
X DP2.2.3
Level 3 - Sub FRs/DPs of FR2.2

FRs
#.1
#.2
#.3
#.4
#.
#.
Coordinate transport
function
Move wafer from CES
to VP
From VP to VPC
From VPC to CT
M
From HB to HBC
From HBC to CES
DPs)
Command and control
algorithm
CES handler
IBTA
Central handler
M
Central handler
SI handler
Taesik Lee 2005
* Design matrix depends on

a process plan and selection
of DPs.
FR1: move wafer from process 1 to 2

:
DP1: robot 1
DP2: robot 2
t=0
t = t1
t = t2
FR = {FR1}
FR = {FR4}
FR = {FR2, FR3, FR5}
DP = {DP1}
DP = {DP2}
DP = {DP1, DP2}
Coupling due to an insufficient number of DPs
Taesik Lee 2005
Problem definition
Conflict : more than one modules

competing for a robot
The conflicts make the waiting time of
wafers inconsistent, which degrades onwafer result variation.
Robot
SP
T1
P2
P1
SP
T2
T1
P1
SP
SP
P1
SP
T2
T1
In/Out buffer P3
P2
P1
T2
T1
P3
P2
T2
P3
Delay time
T1
P1
P3
P2
Conflict
P1
T3
T2
T1
P3
P2
IN OUT
Part
P1
P2
T2
T1
P2
P1
SP
P3
T2
P3
T1
P2
P1
SP
T3
T2
T1
P1
T3
P3
P2
T1
T2
P2
P3
T2
Example : Process timing diagram with a sending period(6 unit)

Taesik Lee 2005
T3
P3
Deterministic scheduling problem

i
i 1
j =1
j =0
j =1
t i = Pj + MvPk j + MvPl j + n SP , n = 0,1,2,...

Dividing both sides by its SP yields
i 1
t i = Pj + MvPk j + MvPl j + n , n = 0,1,2,...

j =1
j =0
j =1
Original pick-up time
Taking only the decimal,

G
i = ti int(ti )
moment of ith transport task

within a period
E
Process
i indicates the (normalized)
D
C
B
A
IN
Conflict
0
0.5
1.5
Time (normalized by sending period)

Taesik Lee 2005
Solution
Basic concept
Break the conflicts among number of transport requests from process
modules
Use predetermined queue as a decoupler between process and
transport
Insert optimum queue at possible process steps
i 1
j =1
j =0
j =1
j =1
t = Pj + MvPk j + MvPl j + n SP + q j , n = 0,1,2,...

*
i
Taesik Lee 2005
Solution
Condition for no-conflict:
~ * * ~
tmax
1 tmax
i
j
Where
*
i j = i j +
k =1
for i = 1,2,... , N ; j = 1,2, , (i 1)
qk
~
tmax : longest transport time
qk = i j +
k =1
(aik a jk ) qk'
k =1
Optimize values of qk along with sending period, subject to no-conflict

condition and process constraint (qcritical = 0 sec)
N
min
q 'j
j =1
Taesik Lee 2005
Solution
Process
Time (sec)
Delay (sec)
40
20
17
60
15
40
35
Adjusted pick-up time
A
D
B
IN
0
0.5
1.5
Transforming a potentially combinatorial

complexity problem to a periodic problem
Solution is obtained for one (and repeating) period
Taesik Lee 2005
Manufacturing Systems Design

Large Scale Problems
Taesik Lee 2005
Customers view on Toyota products

Models rated
at or below average
GM
Chrysler
28
7
3
19
10
Ford
10
1
Nissan
Honda
Models rated
above average
*The Wall Street Journal

May 4, 2000
Toyota
10
Worlds No.2 Automaker

$12B profit (2003)
No1. JD Power Initial Quality Prize
Market capitalization of Toyota ($104B) >
GM ($24B) + Ford ($23B) + DC ($37B) (2003.11.1)
Taesik Lee 2005
TPS / Lean manufacturing system
Set of 19 slides removed for copyright reasons.

Source: Production System Design presentation by Dr. David Cochran
Taesik Lee 2005
Conclusion
Cartoon removed for copyright reasons.
Taesik Lee 2005
Axiomatic Design of Manufacturing Systems
Introduction to Manufacturing System

"What is a manufacturing system?"
"What is an ideal manufacturing system?"
"How should we design a manufacturing
system?"
"Can a manufacturing system be designed in a
rational way based on scientific principles?"

The design of manufacturing systems has an
important effect on
Manufacturing productivity
Return on investment
Market share
Workers satisfaction
Environment

Manufacturing systems consist of
machines, materials, people, and information.
The goal of a manufacturing system is to improve
customer satisfaction
through improvements in
the quality of products, short delivery time, and
high labor productivity.
Manufacturing systems must be designed to
satisfy a specific set of FRs and constraints.
Evolution of cost, quality and workers satisfaction
1
KEY
Cost
1 = Handicraft
2 = Taylorism
3 = High Volume Automation
4 = Flexible Automation
5 = Concurrent Engineering
Worker's
satisfaction
Quality customer's satisfaction
Figure by MIT OCW. After Sohlenius, 2005.

Is there "the" IDEAL MANUFACTURING SYSTEM?
NO.
An ideal manufacturing system is not time-invariant!
The design of an ideal manufacturing system depends

Functional requirements (FRs)
Constraints (C).

2. To reduce the capital investment
Mass production
Capacity planning for machines
Group technology
Transfer lines
Flexible manufacturing systems
3. In recent years, to reduce the materials cost

Manufacturing system design
Lean manufacturing system
Driving Forces for Manufacturing Systems

(From Sohlenius 2005)
The 50's
The 60's
The 70's
Efficiency of
manual labor
Efficiency of
machine-tool
utilization
Minimize capital
cost for products
in process
Order control
customer adapt
Figure by MIT OCW.
The 80's
Quality &
productivity
holistically
The 90's
Customer,
concurrent
design, lean
production,
environment
Numerical Control Machine Tool in 1951

J. T. Parsons, (Traverse City, Michigan)
Small machine shop with milling machine and dial
gages
U.S. Air Force funding of the Servo-mechanisms Lab

(Gordon Brown, Jay Forrester, etc.)
Controversy
development
on
credit
for
NC
machine
tool
Parsons received a National Technology Medal from

President Ronald Reagan
Manufacturing Systems Concept proposed in

1960s by M. Eugene Merchant (1970)
NEEDS
(product requirements)
CREATIVITY
(product concepts)
COST AND CAPABILITIES
PRODUCT
DESIGN
PRODUCTION
PLANNING
PRODUCTION
CONTROL
PRODUCTION
EQUIPMENT
PRODUCTION
PROCESSES
FINISHED
PRODUCTS
for production
programming
feedback,
supervisory,
adaptive
optimizing
including
machine tools
removal,
forming,
consolidative
fully assembled,
inspected &
ready for use
PERFORMANCE
Figure by MIT OCW.
CIRP Definition of Manufacturing System

The manufacturing system is defined as an
organization in the manufacturing industry for the
creation of production. In the mechanical and electrical
engineering industries a manufacturing system in
general has an integrated groups of functions: They are
the sales, design, manufacturing production, and
shipping functions. A research function may provide a
service to one or more of the other functions.

FRs and Cs have been selected implicitly and intuitively
and are affected by
1.
Cost structure of the manufacturing

enterprise
2.
Available technologies
3.
Performance measurements
4.
Competitive factors

Historically the FRs of a manufacturing system
have changed to deal with the most costly item
in manufacturing.
1.
At the beginning of the 20th century, labor cost
was thought to be the dominant cost item.
To minimize the labor content.
Division of labor (F. W. Taylor)
More efficient and automated machines

The Effect of Competitive Factors on Manufacturing
Systems
1.
Demand Rate >> Manufacturing Capacity

Mass production Henry Ford's era for
automobiles
2.
Demand Rate << Manufacturing Capacity

Flexible manufacturing system that responds
to customer preference

How do societal changes affect the design of manufacturing
systems?
Globalization leading to specialization among nations
Environmental concerns leading to
Clean mfg operations
Life cycle issues
Re-manufacturing
Over-capacity of many consumer oriented products (cars, DRAMs, etc.)
Supply exceeds the demand
Customer satisfaction
Customer demands
Mass customization
Flexible manufacturing system

Much of the research in the field of manufacturing systems
has been
1.
To increase the efficiencies of existing

operations and manufacturing processes
2.
To optimize inventory levels
3.
To maximize the utilization of

manufacturing capacity
rather than through

a rational design of manufacturing systems.

What are typical manufacturing systems?
Typically, manufacturing systems are classified in terms of
the physical machine arrangement.
Production job shop
Transfer lines
Flow lines
group technology
Lean, linked cell manufacturing systems

The cost of most manufacturing systems varies
depending on
Production volume
Degree of automation
Labor cost
Equipment costs
Location.

What is wrong with the way manufacturing systems have
been designed in recent years?

Two manufacturing systems that have two identical
sets of machines may have significantly different
production rates and cost, depending on the
design of manufacturing system.
Basic Requirements of a Manufacturing System

A manufacturing system must be designed consistent with the
design axioms to be most efficient and reliable.
In a "push" system, we must be sure that a decoupled system
is created to maintain the independence of FRs.
In these machines, "decouplers" are used to create functional
independence and satisfy various functional requirements
while the part is subject to various processes in the
manufacturing system.
Basic Requirements of a Manufacturing System

For manufacture of random sets of different types of parts
that require a number of different processes, a pull
system appears to be the better than a push system.
In a "pull system", the production rate is gated by the rate
at which the finished part leaves the manufacturing system.
The "pull" system is neither inherently better nor worse
than a "push" system it is simply a matter of having to
satisfy different sets of FRs.
Theorem 25 ("Push" System vs "Pull" System)

When identical parts are processed through a
manufacturing system, a "push" system can be
designed with the use of decouplers (that control
queues) to maximize the productivity, whereas
when irregular parts requiring different
operations are processed, a "pull" system is the
most effective system.
Elements of Manufacturing Systems

What are the key elements of a manufacturing system?
A manufacturing system a subset of the production
system consists of people, information and "things".
Elements of Manufacturing Systems

To make all these machines and people work together for the
common goal of the system,
we must deal with
information -- generation, collection, processing, transmission,
and interpretation of information.
Information is typically embedded in hardware and software.
It forms the interface between diverse elements of a
manufacturing system, and integrates the functions of the
system. Without the proper management of information, a
modern manufacturing system cannot function.
Axiomatic Design of Fixed Manufacturing

Systems for Identical Parts
Highest-Level Design of a Fixed Manufacturing System
FR1 = Maximize the return on investment

(ROI)
Axiomatic Design of Fixed Manufacturing

Systems for Identical Parts
DP1 = Dedicated automated machine that

can produce the desired part at
the
specified production rate.
Axiomatic Design of Fixed Manufacturing Systems

for Identical Parts
DP1 = Dedicated automated machine that can produce the desired part at
the specified production rate.
The FRs of the dedicated and automated machine

may be written as
FR11 = Process the wafers
FR12 = Transport the wafers
Axiomatic Design of Fixed Manufacturing Systems

for Identical Parts
FR11 = Process the wafers
FR12 = Transport the wafers
Corresponding DPs :
DP11 = Process modules

DP12 = Robots
Cs:
C1= Throughput rate
C2 = Manufacturing cost
C3 = Quality of the product
C4 = Yield (production of acceptable products divided by the
total output)
Push system
A conflict can arise
in scheduling the robot motion
if two processes are completed
within the time it takes
for robot to pick up and deliver a wafer
(i.e., single robot motion).
Decoupling of processes from transport

Design equation
FR11 X X DP11
FR12 X X DP12
The original system was a coupled design.
Decoupling of processes from transport

Design equation
FR11 X
=
FR12 X
0 DP11
X DP12
(6.1)
Since this machine processes exactly identical parts, a

"push" system may be designed, where the part will be
supplied to the machine on a regular time interval T.
Push system
Number of modules
corresponding to T.
ni
needed
for
each
i
i
( 6 .2 )
ni = Int = Int
3,600
T
where
T= Cycle time
M = No. of wafers per hour
process
Push system
Total number of modules, n, required to process the
wafers:
N
n = ni
i =1
where
N is the number of tasks, i.e., processes.
(6.3)
Push system
Decompose FR12 (Transport the wafers) as
FR121 = Decouple the process times

FR122 = Minimize the total process time
TC (or minimize the number of
modules)
The corresponding DPs are:
DP121 = Decouplers with q's

DP122 = The minimum value of TC
Push system
Design equation
FR121 X 0
=
FR122 X X
DP121
DP122
(6.4)
Push system
Total process time Tc
T C = t P + tT +
where
tP is the total process time
tT is the total transport time.
i =1
qi
Push system
To minimize TC, we must satisfy the following two
conditions:
TC
qi = 0
i =1
N
2 TC
> 0
2
i =1 qi
N
( 6.5)
where N = number of processes.

Since TC is not a continuous function, we have to solve it approximately.
Push system
Analytical Solution for Queues in Decouplers
Having designed the manufacturing system, we

must replace those X's with mathematical
expressions if they can be modeled.
Push system
Ti = Time the wafer has to be picked up upon the
completion of process j in Module i
Ti =tP +tT
(6.6)
where
tP = total process time up to and including the i-th module
tT = accumulated transport time up to i-1 (normalized
times)
Push system
Ti (Time the wafer has to be picked up upon the
completion of process j in Module i), may be written as
i
i1
Ti =tP +tT = tP,j +tT,j

j=1
(6.7)
j=1
Note: Total process tP = sum of the individual process times, tP,j

Transport times, tT, = sum of the all robot transport time, tT,j
Push System
nR = number of pick-up moves the robot can make
in a given period T
T
nR =
tT
(6.8)
Push System
Time i
-- measured from the beginning of each period -at which the robot has to pick up a wafer
i1
i1
i1
i
i
i
i =Ti InttP,j +tT,j =tP,j +tT,jInttP, j +tT,j

j=1
j=1
j=1 j=1
j=1
j=1
(6.9)
where Int(x) is a function that rounds x down to the next nearest integer.
Push System
T T1 = q1
*
1
T2 T2 = q1 + q 2
*
T T3 = q1 + q 2 + q 3
*
3
etc.
Push System
i = qj = aijqj
*
i
j=1
wherethe matrixaij is definedas

0
aij =
1
when i > j
when i < j
(6.11)
Push System
We can approximately determine i* by solving Eq. (6.9), by determining
where the decouplers may be needed, and by approximating the values of
queues.
[ ]{}
{i i}= aij qj
(6.12)
Push System
1
{q}= [a] {} = [A]{} = [A]{}
aij
1
where
= i i *
[a] matrix with elements aij

N
a ji determinant of matrix [a ] = akk = 1
[ A] = Adj [aij ]= [A ji ]
A ji = ( 1)
i+j
M ji minor
M ji
of a ji
i =1
(6.13)
Equation (6.13) can be solved iteratively. To solve Eq. (6.13), we need to know {},
which can be approximated by estimating reasonable values for i* and by solving Eq. (6.9)
for i. The value for j* can be estimated by adding transport time to i*, since |i*- j*|> tT,
for all j's except j=i. The solution can be improved by successive substitution of the
improved values of i*. The determinant |aij| of the triangular matrix {a} is equal to the
product of the diagonal elements.
Since the best solution is the one that makes the total cycle time TC a minimum, we must
seek for a set of values of qi that yield a minimum value for the total queue, qi. When the
precise control of processing time is critical, the queue qi associated with the process should
be set to equal to zero.
To solve Eq. (6.12) for the best set of queues q's, Oh (1998) and Oh and Lee (1999)
developed an optimization software program based on genetic algorithm. Multiplying these
qs by T, we can obtain actual values of queues.
Example 6.1 Determination of the queues of a fixed

manufacturing system that processes identical parts
Steps
1
2
3
4
5
Modules Temp. (C)Duration +- tolerance (Seconds)

A
B
C
D
E
35
80
10
50
68
50 + 25
45 +/- 0
60 + 20
70 +10
35 +/- 0
The process times in Modules B, and E must be precise because of the

critical nature of the process. The cycle time is assumed to be the process
time plus the transport time both for placement and pick-up of the wafer.
The robot must pick up the wafers from a supply bin (load-lock) and
deliver it Module A and when the process is finished, it must pick up the
container from Module G and place it on a cassette. These operations
take 6 seconds each.
Solution
The number of modules is dependent on the process time TC and the desired
throughput rate. The required number of modules is as follows:
Modules
A
B
C
D
E
Number of Modules
2
1
2
2
1
Without any decouplers, there are simultaneous demands for the service of the
robot as shown in Fig. ex.6.1.a, which shows the time the process is finished in
each of the modules within a given period T.
Original Pick-up Time

E
Process
D
C
B
conflicts
A
IN
0
0.2
0.4
0.6
0.8
1.2
1.4
1.6
1.8
The original pick-up time without any decouplers. There are conflicts
among the transport demands from Modules A, B, C, E, and IN, since it
takes the robot 0.1 normalized time to travel between the modules.
Push Systetm
Solution for qs
The best solution was obtained by finding a set of values that give the shortest cycle time
TC solving Eq. (6.12) repeatedly and using a genetic algorithm. The solution yields the
following values for q's.
qA = 19.7 sec
qB = 0 sec
qC = 9.3 sec
qD= 9.3 sec
qE = 0 sec
The queues for B and E zeros since the tolerance on these two modules is specified to be
zero. Therefore, the queues of other modules has been adjusted to make these two queues
to be zero.
Decoupled Design for Pick-up Time

E
D
C
B
A
IN
0
0.2
0.4
0.6
0.8
1.2
1.4
1.6
1.8
The actual pick-up time for Modules A, B, C, D, and E. A set of queues, qA=19.7,
qB=0, qC=9.3, qD=9.3, and qE=0, are added to the original process times to resolve
the transport conflicts. It takes the robot 0.1 normalized time to travel between the
modules. Oh (1998) and Oh and Lee (2000).
Push System
One of the interesting results of this solution is that the

number of combinations for part flow reduces down from
several thousands to a few, because the parts flow
through the manufacturing system along deterministic
paths.
What the concept of decouplers has done is to change a
combinatorial problem into a periodic function that
repeats itself with a given cycle that is deterministic.
What is the main message of this section?

1.
When identical parts processed by identical processes, we can
use a Push System.
2.
By decoupling a coupled design, we have reduced a problem that
used to be treated as a combinatorial problem to a deterministic problem
that can be solved "exactly"!
3.
The number of possible combinations was reduced from several
thousands to two! The throughput rate is also controlled precisely
because the system has been designed with a given throughput in mind.
4.
The decoupled design of a manufacturing system that processes
identical parts can be modified without much effort.
What is the main message of this section?

5.
The shortcomings of the traditional approach to these problems
such as operations research (OR) can be overcome by developing designs
that satisfy the Independence Axiom.
6.
The ideas presented in this section can be applied to a variety of
system problems that involve design, analysis, and optimization.
7.
The important message is that if the system design violates the
Independence Axiom, it is extremely difficult to achieve the functional
requirements of a manufacturing system.
8.
This solution does not allow random variations in process times.
How would you solve the problem with random variations?
Chapter 6
Design of Manufacturing Systems II
Pull System
Evolution of cost, quality and workers satisfaction:

1. Handicraft, 2.Taylorism, 3. High volume automation, 4.
Flexible automation, 5. Concurrent engineering
(From Sohlenius, 2005)
1
KEY
Cost
1 = Handicraft
2 = Taylorism
3 = High Volume Automation
4 = Flexible Automation
5 = Concurrent Engineering
Worker's
satisfaction
Quality customer's satisfaction
Figure by MIT OCW. After Sohlenius, 2005.
Driving Forces for Manufacturing Systems

(From Sohlenius 2005)
The 50's
The 60's
The 70's
Efficiency of
manual labor
Efficiency of
machine-tool
utilization
Minimize capital
cost for products
in process
Order control
customer adapt
Figure by MIT OCW.
The 80's
Quality &
productivity
holistically
The 90's
Customer,
concurrent
design, lean
production,
environment
Numerical Control Machine Tool in 1951

J. T. Parsons, (Traverse City, Michigan)
Small machine shop with milling machine and dial
gages
U.S. Air Force funding of the Servo-mechanisms Lab
(Gordon Brown, Jay Forrester, etc.)
Controversy on credit for NC machine tool
development
Parsons received a National Technology Medal from
President Ronald Reagan
Manufacturing Systems Concept proposed in 1960s

by M. Eugene Merchant (1970)
NEEDS
(product requirements)
CREATIVITY
(product concepts)
COST AND CAPABILITIES
PRODUCT
DESIGN
PRODUCTION
PLANNING
PRODUCTION
CONTROL
PRODUCTION
EQUIPMENT
PRODUCTION
PROCESSES
FINISHED
PRODUCTS
for production
programming
feedback,
supervisory,
adaptive
optimizing
including
machine tools
removal,
forming,
consolidative
fully assembled,
inspected &
ready for use
PERFORMANCE
Figure by MIT OCW.
CIRP Definition of Manufacturing System

The manufacturing system is defined as
An organization in the manufacturing industry
for the creation of production. In the
mechanical and electrical engineering
industries a manufacturing system in general
has an integrated groups of functions: They
are the sales, design, manufacturing
production, and shipping functions. A
research function may provide a service to one
or more of the other functions.

What are typical manufacturing systems?
Typically, manufacturing systems are classified in terms of the
physical machine arrangement.
Production job shop
Transfer lines
Flow lines
group technology
Lean, linked cell manufacturing systems

The cost of most manufacturing systems varies
depending on
Production volume
Degree of automation
Labor cost
Equipment costs
Location.
Axiomatic Design Steps

for
a Pull System
Step 1. Choose FRs in the Functional
Domain
FR1 = Maximize the return on
investment (ROI)
SalesCost
ROI=
Investment

Step. 2. Mapping of FRs in the Physical
Domain to Determine DPs
DP1a = Manufacturing system to provide
products at a minimum cost
or
DP1b = Manufacturing system design to provide

products customers demand

Step 3. Decompose FR1 in the Functional
Domain Zigzagging between the domains
FR11 = Increase the sales revenue
FR12 = Minimize the manufacturing cost
FR13 = Minimize manufacturing investment
Sales Cost
ROI =
Investment

Step 4. Find the Corresponding DP1x's by Mapping
FR1x's in the Physical Domain
Decomposition of DP1a
DP1a1 = Maximization of production output
DP1a2 = Unit cost minimization
DP1a3 = Machine utilization
Decomposition of DP1b
DP1b1 = Product design and manufacture to
maximize customer satisfaction
DP1b2 = Target production cost
DP1b3 = Investment in production with a systems
thinking approach

Step 5. Determine the Design Matrix
The second design represented by DP1ax's satisfies FR1x.
a
FR11 000 DP 1 1
a
FR13 = XX 0 DP 1 3
FR12 XXX DP 1a 2
(2)
Similarly, the second design represented by DP1bx's satisfies

FR1x.
b
FR11 X00 DP1 1
b
FR12 = XX 0 DP1 2
FR13 XXX
b
DP1 3
(4)

Step 6. Zigzag
Decompose FR11, FR12 and FR13 by going from the Physical
to the Functional Domain and determine the corresponding
DPs (Level 3)
Step 6-a. Decompose FR11 (Increase the sales revenue) and
DP11 (Product design and manufacture to
maximize customer satisfaction) and determine DPs
FR111 = Sell products at the highest acceptable price
FR112 = Increase market share (volume)
DP111 = Customer perceived value of product improved
DP112 = Broad product applications
FR111 X0 DP111
=
FR112 XXDP112
(6)

Step 6. Zigzag
SR = (Pr icei Volume )

i
i =1
Step 6-a. Decompose FR11 (Increase the sales revenue) and

DP11 (Product design and manufacture to
maximize customer satisfaction) and determine DPs
FR111 = Sell products at the highest acceptable price
FR112 = Increase market share (volume)
DP111 = Customer perceived value of product improved
DP112 = Broad product applications
(6)
FR111 X0 DP111
=
FR112 XXDP112

Step 6-b. Decompose FR12 (Minimize the manufacturing
cost) and Determine DPs
FR12 (Minimize the production cost) may be decomposed with
DP12 (Target production cost) in mind as
FR121 = Reduce material costs
FR122 = Reduce operational activity costs
FR123 = Reduce overhead
The corresponding DPs may be stated as:
DP121 = Target price given to suppliers
DP122 = Targeted performance of operational activities
DP123 = Right size business processes
The elements of the production cost are the cost of raw materials and
components, the direct cost, indirect cost, and administrative costs or
overhead.

Step 6-b. Decompose FR12 (Minimize the manufacturing
cost) and Determine DPs
FR12 (Minimize the production cost) may be decomposed with
DP12 (Target production cost) in mind as
FR121 = Reduce material costs
FR122 = Reduce operational activity costs
FR123 = Reduce overhead
DP121 = Target price given to suppliers
DP122 = Targeted performance of operational activities
DP123 = Right size business processes
Design Matrix
FR121 X00 DP121
FR122 = 0X 0 DP122
FR123 00X DP123

Step 6-c. Decompose FR13 (Minimize manufacturing
investment) and select DPs
FR13 (Minimize production investment) may be decomposed with
DP13 (Investment in production with a system thinking approach) in
mind as
FR131 = Acquire machines with cycle time
the minimum takt time

FR132 = Ensure flexibility to accommodate capacity
increments at lowest cost
FR133 = Develop flexible tooling
FR134 = Ensure flexibility to accommodate future
products

FR131 = Acquire machines with cycle time (less than or equal to) the
minimum takt time
FR132 = Ensure flexibility to accommodate capacity
increments at lowest cost
FR133 = Develop flexible tooling
FR134 = Ensure flexibility to accommodate future
products

DP131 = Machine design focused on customer demand
pace and value added work
DP132 = Linked cell manufacturing systems
DP133 = Flexible tooling design
DP134 = Movable machines and reconfigurable
stations to enable new cell design

Step 6-c. Decompose FR13 (Minimize manufacturing
investment) and select DPs
FR13 (Minimize production investment) may be decomposed with
DP13 (Investment in production with a system thinking approach) in
mind as
Design Matrix
FR 131
FR 132
FR 133
FR 134
X 000 DP 131
XX 00 DP 132
=
00 X 0 DP 133

DP 134
000
X

Step 7. Fourth Level Decomposition
Step 7-a. FR11 Sales Revenue Branch
Functional requirement FR111 (Sell products at the highest
acceptable price) must be decomposed with DP111 (Customer
perceived value of product improved).
FR111 may be decomposed as follows:
FR1111 = Increase the appeal of products by
providing desired functions and features
FR1112 = Increase the reliability of products
FR1113 = On time delivery (for a variety of products)
FR1114 = Decrease variation of the delivery time
FR1115 = Provide effective after sales service

DP1111 = Design of high quality products that meet
customer needs as specified by FRs and Cs
DP1112 = Robust design of products
DP1113 = Production based on actual demand
DP1114 = Predictable production output
DP1115 = Service network
The design equation and matrices are as follows:
FR 1111
FR 1112
FR 1113 =
FR 1114
FR 1115
X 0000
XX 000
0 XX 00
0 XXX 0
0 X 00 X
DP 1111
DP 1112
DP 1113
DP 1114
DP 1115

FR112 (Increase market share (volume)) must be decomposed with
DP112 (Broad product applications).
FR1121 = Development of niche (new or custom)
products
FR1122 = Development of multiple solutions within
the product line
DP1121 = Short product development process
DP1122 = Product variety
FR1121 X0 DP1121
DP1122
FR1122
XX
(10)

Step 7-b. FR12 Production Cost Branch
FR 122 (Reduce operational activity costs) and DP122 (Target production
cost) may be decomposed as
FR1221 = Reduce transport costs
FR1222 = Reduce setup costs
FR1223 = Reduce costs of manual operations (mach.
load/unload, assembly, inspect.)
FR1224 = Reduce fabrication costs
FR1225 = Reduce maintenance costs
DP1221 = Product-flow oriented layout
DP1222 = Setup performed with reduced resources
DP1223 = Effective use of the workforce
DP1224 = Fabrication parameters based on takt time to increase
tool life
DP1225 = Total productive maintenance program
The design is an uncoupled design.

Step 8. Fifth Level Decomposition
FR 1112 (Increase the reliability of products) and DP1112 (Robust design of products)
may be decomposed as
FR11121 = Determine the lowest tolerable stiffness of the product
FR11122 = Determine the design range for manufacturing tolerance
FR11123 = Select manufacturing operations with a system range that is
within the design range
DP11121 = Mathematical model for stiffness determination
DP11122 = Mathematical model for derivation of design range for PVs
DP11123 = Selected machines with appropriate system range for PVs
FR11121 X00 DP11121

FR11122 = XX0 DP11122
FR11123 XXX DP11123
(12)
The design matrix is triangular and thus, it is a decoupled design.

Functional requirement FR1113 (Decrease mean delivery time) must be decomposed
with DP1113 (Production based on actual demand).
FR11131 = Produce at the customer demand cycle time (or takt time)
FR11132 = Produce the mix of each part type demanded per time interval
FR11133 = Be responsive to the downstream customers demand time
interval
DP11131 = Linked-cell manufacturing system balanced to customer demand
DP11132 = Level production
DP11133 = Reduced response time across the production system
The design matrix
FR11131 X00 DP11131

FR11132 = 0X 0 DP11132
FR11133 XXX DP11133
A Linked-cell Manufacturing System to Illustrate the Concept.

The machines are arranged in a cellular structure. In this figure, two workers are
moving in two loops opposite to the flow of the work piece (Cochran, 1998).
Straighten
Crack detect
Buff OD rack
Super finish OD
Assemble pistons
Lap journal
10
Broach teeth
2
1
IN
Wash
Centerless grinder
3
Final wash
(rust prevent)
Resistance harden teeth
OUT
Induction temper
11
Straighten
Scan harden journal
Deburr
Turn piston groves
Drill cross holes
Broach end slots
Machine both
ends
Figure by MIT OCW.

Functional requirement FR1114 (Decrease variation of the delivery time) and DP1114
(Predictable production output) must be decomposed.
FR11141 = Respond quickly to production problems
FR11142 = Produce with a predictable quality output
FR11143 = Produce with a predictable time output
DP11141 = Visual control system to provide rapid response
DP11142 = Production with no defects and the ability to identify root cause
DP11143 = Predictable production resources
FR11141 X00 DP11141

FR11142 = XX0 DP11142
FR11143 XXX DP11143
(14)

Step 8-b. FR12 Manufacturing Cost Branch

FR 1223 (Reduce costs of manual operations (mach. load/unload, assembly,
inspect.)) may be decomposed with DP122 (Effective use of the workforce) in mind as
FR12231 = Reduce tasks that tie the operator to the machine
FR12232 = Enable worker to operate more than one machine or station
FR12233 = Plan the resources to produce with different production volumes
DP12231 = Machines & stations designed to run autonomously
DP12232 = Work-loops implemented in a cell layout
DP12233 = Standardized work-loops designed for different volumes
FR12231 X00 DP12231

FR12232 = XX0 DP12232
FR12233 XXX DP12233
(15)

Step 9. Sixth Level Decomposition: FR11 Sales Revenue Branch
Functional requirement FR11131 (Produce at the customer demand cycle time (or
takt time) and DP11131 (Linked-cell manufacturing system balanced to customer
demand) may be decomposed as follows:
FR111311 = Define customers, parts, and volumes for each sub-system or
cell within production
FR111312 = Design sub-system for a range of volume fluctuation
DP111311 = Configuration of sub-systems to enable flow at the ideal range
of cycle times
DP111312 = Cell or sub-system designed to meet the minimum takt time
FR111311
X 0 DP 11131
FR111312
XX
DP
11132
(16)

FR11132 (produce the mix of each part type demanded per time interval) and DP11132
(Level production) may be decomposed as follows:
FR111321 = Produce in small run sizes
FR111322 = Convey in small and consistent quantities
FR111323 = Produce and supply only the parts needed
DP111321 = Short setup time
DP111322 = Standard containers that hold small amounts of parts
DP111323 = Information system to produce only the parts needed (Pull
system)
FR111321 X00 DP111321

FR111322 = XX 0 DP111322
FR111323 XXX DP111323
(17)
This design is a decoupled design and thus, satisfies the Independence Axiom.

FR11133 (be responsive to the downstream customers demand time interval) must be
decomposed with DP11133 (reduced response time across the production system) in
mind. FR11133 may be decomposed as
FR111331 = Reduce sub-system replenishment time to less than the
customer demand interval.
FR111332 = Ensure that sufficient parts are available to satisfy the
customer demand interval.
The corresponding DPs are
DP111331 = Elimination of wastes that cause excess lead-time
DP111332 = Standard work-in-process (swip) quantity of parts
FR111331 X 0 DP111331
= X X
FR111332
DP111332
(18)
This design is a decoupled design and thus, satisfies the Independence Axiom.

FR11142 (produce with a predictable quality of output) must be decomposed with
DP11142 (production with no defects and with the ability to identify root cause) in
mind.
FR111421 = Ensure capable processes.
FR111422 = Decrease sources of variation due to multiple flow paths.
FR111423 = Prevent making defects throughout.
FR111424 = Do not advance defects to the next operation.
DP111421 = Capable machines, equipment, tools, and fixtures
DP111422 = Single path through manufacturing system and external
supplier (no parallel processing)
DP111423 = Use of standards and devices to prevent defects
DP111424 = Use of successive checks to detect defects if they do occur
FR111421
FR111422
FR111423 =
FR111424
X
X
X
0
X
0
0
X
0 DP111421
0 DP111422
0 DP111423
X
DP111424
(19)

Step 10. Seventh-Level Decomposition: FR11 Sales Revenue Branch
FR111312 (design sub-system for a range of volume fluctuations) must be
decomposed with DP111312 (sub-system designed to meet the minimum TAKT time) in
mind.
FR1113121 = Select appropriate manufacturing process.
FR1113122 = Design manufacturing process cycle time at each station to
meet minimum TAKT time.
FR1113123 = Design station fixtures to enable minimum TAKT time.
DP1113121 = Physics of the manufacturing process
DP1113122 = Manufacturing process work content defined to be less than
the minimum TAKT time
DP1113123 = Fixture design to provide quick load/unload (within
required tolerance)
The design equation and matrices are
FR 1113121 X
FR1113122 = X
FR 1113123 X
0
X
X
0 DP 1113121
0 DP 1113122
X DP 1113123
(20)
Flow Chart of the Manufacturing System Design

M1
M 11
M 111
M 112
M 1111
M 1121
M 1112
M 11121
M 11123
M 1 1122
M 1113
M 11131
M 111311
M 11133
M 111312
M 111331
M 1113122
M 1113121
M 111332
M 1113123
M 11132
M 111321
M 111323
M 111322
M 1114
M 11141
M 11142
M 11143
M 111421
M 111422
<111423
M 111424
M 1115
M 12
M 13
M 121
M 131
M 122
M 133
M 1221
M 134
M 1222
M 1223
M 12231
M 1224
M 1225
S um m ation J uncti on
C ontrol Junction
M 123
M 12232
M 12233
M 132
M 1122
Use of the Flow Chart of

the Manufacturing System Design
1.
Diagnosis
2.
Engineering changes
3.
Job assignment and management
4.
Distributed systems
5.
Software development
Complexity
Reference:
Nam P. Suh, Complexity: Theory and Applications,
Oxford University Press, 2005
Why complexity?
In many fields (engineering, economics, politics, academic
administration, etc.), one of the major goals is to reduce
complexity.
In many fields, we want to understand the causes for
complexity.
There is a great deal of confusion what we mean by
complexity.
Computer scientists and mathematicians have considered
complexity, which may not be applicable to design and
engineering fields.
(David P. Feldman and Jim Crutchfield, A Survey of Complexity Measures, 1998 Complex Systems Summer School,
Santa Fe Institute, 11 June 1998.
Issues related to complexity

Why is something complex?
Why is something that appeared to be complex is
not complex once we understand it?
What is complexity?
Is it better to reduce complexity of an engineered
system?
How do we reduce complexity?
Is the complexity of engineering systems different
from that of socio-economic-political systems?
Is the complexity of natural systems different from
engineered systems?
Consider the task of cutting a rod?

Is this a complex task?
Suppose we must cut it to 1 m +/- 1 cm. Is
this a complex task?
Suppose we must cut the rod to 1 m +/- 1
micron?
Consider the task of controlling the flow rate

and the temperature of water using the faucet
shown below?

See Figure E3.3a in Suh, N. P.
Axiomatic Design: Advances and Applications. New York, NY:
Oxford University Press, 2001. ISBN: 0195134664.
Is this a complex task?

Suppose the task is to control the flow rate 1 +/- 0.3
gallons per minute and the temperature within 90
+/- 25 F.
Is it a complex task? Why? Why not?
If the flow rate has to be controlled to 1 gallon +/0.01 gallons/minute and the temperature within 60
+/- 0.2 F, it appears to be complex.
The task is more complex when these two FRs are
coupled by the design!
International Space Station Beta Gimbal

Assembly Failure
Other basic questions:
How do we guarantee the long-term

stability of engineered systems?
Why is there so much wasted effort in
developing new products?
What is complexity?
Departure from Conventional View of

Complexity
Most people have been trying to
understand complexity in terms of
physical things.
Many Different Definitions of

Complexity
Computational complexity
Algorithmic complexity
Probabilistic complexity
Departure from Conventional View of

Complexity
Most people have been trying to
understand complexity in terms of
physical things.
Complexity must be viewed in the
functional domain.
Functional Domain vs Physical Domain
"What we
wa nt to
ac hieve or
kn ow"
"H ow w e
plan to
ac hieve
the goa l"
{FR s}
{DP s}
Consider the problem associated with

satisfying an FR.
Prob.
Density
Design
range
FR
If the system range is inside the design range, FR

can easily be achieved. Therefore, it is not complex.
Prob. Density
Design
Range
System
Range
Functional
Requirement
In this case, FR cannot always be satisfied. There is

a finite uncertainty. Therefore, the task of
achieving the FR appears to be complex.
Prob. Density
Design
Range
System
Range
Functional
Requirement
Definition of Complexity
Complexity is defined as a measure of

uncertainty in satisfying the functional
requirements.
Definition of Complexity as applied

to natural science
uncertainty in our ability to predict a certain
natural phenomenon to the desired accuracy.
Complexity,
which is defined as a measure of uncertainty
in satisfying the FRs,
is a relative quantity.
Four Different Kinds of Complexity
Time-Independent Real Complexity

Time-Independent Imaginary Complexity
Time-Dependent Combinatorial Complexity
Time-Dependent Periodic Complexity
Complexity can be reduced by taking the

following actions:
Reduce Time-Independent Real Complexity
Eliminate Time-Independent Imaginary Complexity
Transform Time-Dependent Combinatorial
Complexity into Time-Dependent Periodic Complexity
Important concept: Functional Periodicity
Functional Periodicity
Temporal periodicity
Geometric periodicity
Biological periodicity
Manufacturing process periodicity
Chemical periodicity
Thermal periodicity
Information process periodicity
Electrical periodicity
Circadian periodicity
Material periodicity

Time-Independent Real Complexity --> Infinity.
Prob. Density
Design
Range
System
Range
Functional
Requirement
What happens when there are many FRs?

Most engineered systems must satisfy many
FRs at each level of the system hierarchy.
The relationship between the FRs determine
how difficult it will be to satisfy the FRs within
the desired certainty and thus complexity.
If FRs are not independent from each other,

the following situation may exist.
Pro b. De n s i ty
Pro b. De n s ity
De si g n
Ra n ge
De si g n
Ra n g e
Syste m
Ra n ge
Sy ste m
Ra n g e
FR1
FR2
Axiomatic design:
Mapping, hierarchies, and zigzagging
Customer
Domain
Functional
Domain
What?
Physical
Domain
How ?
What?
Customer Needs
Functional
Requirements
Process
Domain
Design
Parameters
How ?
Process Variables
Design Axioms
Axiom 1 The Independence Axiom
Maintain the independence of functional
requirements.
Axiom 2 The Information Axiom
Minimize the information content.
Design Equation and Matrix

The relationship between {FRs} and {DPs} can be written as
{FRs} = [A] {DPs}

When the above equation is written in a differential form as
{dFRs} = [A] {dDPs}

[A] is defined as the Design Matrix given by elements :
Aij = FRi/DPi
Uncoupled, Decoupled, and Coupled Design

Uncoupled Design
A11 0
[ A] = 0 A22
0
0
0
0
A33
Decoupled Design
0
A11 0
[A] = A21 A22 0
A31 A32 A33
Coupled Design
All other design matrices
Design Range, System Range, and

Common Range
Probab.
Density
Target
Bias
System
Rang e
Design Rang e
Area o f
Com mon
Rang e (Ac)
Variation
from the
peak valu e
FR
Coupling Increases Time-Independent Real

Complexity!
Real complexity of a decoupled system is, in general, larger than
that of a uncoupled design.
Uncoupled
FR1 A11 0

FR
A22
2
= 0
FR3 0
0

FR1
DP1 =
A11
FR2
DP2 =
A22
FR3
DP3 =
A33
Decoupled
0 DP1
0 DP 2
A33 DP3
FR1 A11 0

FR
2
= A12 A22
FR3 A13 A23
DP1 =
DP 2 =
DP3 =
0 DP1
0 DP 2
A33 DP3
FR1
A11
FR 2 A21 DP1
A22
FR3 A31 DP1 A32 DP 2
A33
Complexity of a Knob Design
M ill ed Fl a t
En d o f th e
s h af t
Me ta l
S h af t
S lot
I n jec tio n
m o ld ed
n ylo n K n o b
Knob designs
Which is more complex?

Which knob has a higher com plexity?
M ill ed Flat
En d o f th e
sh af t
M ill ed Flat En d
o f th e sha ft
Slot
Me tal
Sh af t
A
Injec tio n
m old ed
nylo n K n ob
(b)
(a)
Se cti on view A A
Knob designs
Conventional Engine: Real Complexity
Schematic removed for copyright reasons.
Conventional Engine
A New Engine
Schematic removed for copyright reasons.
New Engine
Simulation Results of the New Engine (?)
New Engine
What is Time-Independent Imaginary

Complexity?
Imaginary complexity is defined as:
Uncertainty that is not real uncertainty,
but arises because of the designers
lack of knowledge and understanding of a
specific design itself.
What is Time-Independent Imaginary

Complexity?
Example:
Combination Lock
Imaginary complexity changes with a
priori knowledge:
(a) Without knowing any of the numbers
(b) Know the numbers but the sequence is not
given
(c) Know both the numbers and the sequence

What is Imaginary Complexity ?
Prob. Density
Design
Range
System
Range
Functional
Requirement
Example of Imaginary Complexity

Light
Original
image
Image is
created
Seleniumcoated
Aluminum
cylinder
PaperFeed
Roll
Paper
Wiper
Roll
Toner
container
Toner is coated
on surfaces of
Selenium with
electric charges
Schematic drawing of the xerography based printing machine

Consider a triangular design matrix for a decoupled
design:
FR1 X00 DP1
FR2 = XX0DP2
FR3 XXXDP3
CI = - log2 P = log2 n!
n
1
2
3
4
5
6
7
8
n!
1
2
3
24
120
720
5,040
40,320
p = 1/ n!
1
0.5
0.1667
0.04167
0.8333 x 10-2
0.1389 x 10-2
0.1984 x 10-3
0.2480 x 10--4

CR = 0
(CI)max = log2 m!
Prob. Density
Design
Range
System
Range
Functional
Requirement
CMP Machine Designed and Built by Four

New Graduate Students
CAD model
Fabricated machine

See Figure 8.20 and 8.21 in Suh, N. P.
Axiomatic Design: Advances and Applications. New York, NY:
Oxford University Press, 2001. ISBN: 0195134664.
CMP machine
2 platen, single head (200 mm)/Multi-step wafer polishing.
Developed at MIT to meet the research needs.
9 servo motors/4 pumps/8 pressure regulators/60 on-off controllers.
Control hardware
Servo Amplifiers
Master Switches
Encoder Filter
ADwin Controller
DC Distribution Panel
Servo Amplifiers & DC Supplies in the Back Panel
AC Switch Panel
Time-Dependent Combinatorial
Complexity
Time-dependent combinatorial complexity
arises because the future events occur in
unpredictable ways and thus cannot be
predicted.
For example, it occurs when the system
range moves away from the design range
as a function of time.
Pro b. D en sity
T h e Sy st e m Ra n g e c han ge s
con tin u ou sly a s a f u nc tion of ti me.
D es ign
Ran ge
T im e
FR
Example
Airline Schedule
3. Time-Dependent Combinatorial Complexity

The combinatorial complexity is defined as the complexity
that increases as a function of time due to a continued
expansion in the number of possible combinations with
time, which may eventually lead to a chaotic state or a
system failure.
Example: Job shop scheduling

Future scheduling is affected by the decisions made
earlier and its complexity is a function of the decisions made over its
past history
4. Time-Dependent Periodic Complexity

The periodic complexity is defined
as the complexity that only exists in a finite period,
resulting in a finite and limited number of probable
combinations.
Example: Airline flight scheduling

All of the uncertainties introduced during the course of a
day terminate at the end of a 24-hour cycle, and hence the complexity
does not extend to the following day
Reduction of Time-Dependent Complexity
Transform
a design with time-dependent combinatorial
complexity
to
a design with time-dependent periodic
complexity

following actions:
Reduction of Combinatorial Complexity

How?
Through
Re-initialization of the system
by defining
a Functional Period.
Note: Functional period is defined by a repeating set of
functions, not by time period, unless time is a set of
functions.
Transformation of Time-Dependent Combinatorial

Complexity
Basic Idea
1.
Make sure that the design satisfies the
Independence Axiom.
2.
Identify a set of FRs that undergoes a cyclic
change and has a functional period.
3
Identify the functional requirement that

may undergo a combinatorial process
4.
T C com ( FR i ) C per (FR i )
6.
5.
"Reinitialization"
Set the beginning of the cycle as t=0
The functional periodicity are the following types:

(1) Temporal periodicity
(2) Geometric periodicity
(3) Biological periodicity
(4)
(5) Chemical periodicity

(6) Thermal periodicity
(7) Information process periodicity
(8) Electrical periodicity
(9) Circadian periodicity
(10) Materials periodicity
Example: Productivity of A Manufacturing

System that Consists of Many Sub-Systems
How do we maximize the productivity of

a manufacturing system?
Example: Semiconductor Manufacturing
Complexity - II
Consider the creation of a manufacturing

system by combining two sub-systems
serially.
A Manufacturing System
Composed of Two Sub-Systems
Stack of modules
Stack of modules
Track
Robot
Loading
Station
Stack of modules
Subsystem A
Stack of modules
Subsystem B
A Cluster of two machines that are physically coupled to manufacture a part.
Each subsystem consists of

many processing steps and
a robot that transports the part.
After the part is finished in
Subsystem A, the part is further
processed in Subsystem B.
We want to maximize the production rate.
If the throughput rates are the following:

Subsystem A -- 100 parts/hr and
Subsystem B -- 60 parts/hr,
what should be the throughput rate of the
manufacturing system that consists of
Subsystem A and Subsystem B?
Example: Subsystem X feeds parts to Subsystem Y
Subsystem X
IN
Process
a
Process
b
Subsystem
Y
Process
d
Process
c
Example: Physical Configuration
c
b
8
7
a
9
10
IN
M a c h in e
M a c h in e
M a c h in e
M a c h in e
M a c h in e
t' = in itia liz e d t 0

CTY
sta te
a
25
b
25
c
25
d1
ocp
d2
45
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
10
15
20
25
30
35
40
45
50
55
60
65
70
75
a
b
c
d2
ro b o t
Figure 13. Information at the instant of initialization
M a c h in e
M a c h in e
M a c h in e
M a c h in e
M a c h in e
ro b o t
t' = in itia liz e d t 0

5
10
CTY
0
sta te
a
25
a
b
25
b
c
25
c
d1
ocp
1
2
d2
45
1
15
20
25
30
35
40
45
50
55
60
65
70
75
80
10
15
20
25
30
35
40
45
50
55
60
65
70
7
3
10
a
b
c
d1
ID L E
d2
7
10
Figure 14. Result of scheduling for a period
The same approach as in case 2, with the no-transporttime adjusted, applies to case 3.

(2)

(4) Manufacturing process periodicity
(10) Material periodicity
Example: Design of Low Friction Sliding

Surfaces
Consider the following task:

Reduce friction between two sliding
surfaces under load
Friction Space
Design of Low Friction Sliding Surfaces without

Lubricants

Lubricants
What are the FRs?


Lubricants

interface
Constraint:
No lubricant

Lubricants

See Figure 7.11 and 7.13 in [Complexity]:
Suh, N. P. Complexity: Theory and Applications. New York, NY: Oxford
University Press, 2005. ISBN: 0195178769.
DP1 = Total contact area of the pad, A

DP2 = Roughness of the planar surface of pads, R
DP3 = Length of the pad in the sliding direction,
DP4 = Volume and depth of the pocket for wear particles, V

Lubricants
FR1 X 0 0 0 DP1 X 0 0 0 A

FR2 0 X x 0 DP2 0 X x 0 R
=
=

FR3 0 0 X 0 DP3 0 0 X 0

FR4 0 0 0 X DP4 0 0 0 X V

Lubricants
Test Results:
Drive sprockets, idlers, rollers, Grouser shoes
Pin-Joint Design

See Figure 7.25 and 7.26 in [Complexity].
Electrical Connectors
(Courtesy of Teradyne, Inc.)

See Figure 7.29 in [Complexity].
Conventional Electrical Connectors
Male connector
Compliant pin
(for permanent connection)
Plastic
overmolding
Plastic
overmolding
Female connector
Multiple layers will be stacked together

to obtain an entire connector.
Figure by MIT OCW.
FRs of a Data Electrical Connector

FR1 = Mechanically connect and disconnect
electrical terminals
FR2 = Control contact resistance (should be less
than 20m)
FR3 = Prevent the cross-talk (i.e., interference)
between the connections
Subject to the following constraints (Cs):
C1 = Low cost
C2 = Ease of use
C3 = Long life (> million cycles)
C4 = Maximum temperature rise of 30 oC
C5 = Low insertion force
FRs of a Power Electrical Connector

than 20m)
FR3 = Maximize power density
C1 = Low cost
C2 = Ease of use
DPs of a Data Electrical Connector
DP1 = Cylindrical assembly of the woven

tube and the pin
DP2 = Locally compliant electric contact
DP3 = Number of conducting wires
Decomposition of FR1 (Mechanically connect and

disconnect electrical terminals) and DP1 (Cylindrical
assembly of the woven tube and the pin)
FR11 = Align the rod axially inside the tube
FR12 = Locate the axial position of the rod in the tube
FR13 = Guide the pin
DP11 = Long aspect ratio of the rod and the tube
DP12 = Snap fit
DP13 = Tapered tip of the pin
Decomposition of FR2 (Control contact resistance to be

less than 20m) and DP2 (Locally compliant electric
contact)
FR21 = Prevent oxidation of the conductor
FR22 = Remove wear particles
FR23 = Control line tension/deflection of the nonconducting fiber
DP21 = Gold plated metal surface
DP22 = Space created in the crevices between fibers
DP23 = Spring
Design Matrix
FR1 X 00 DP1
FR2 = XX 0 DP2
FR3 0XX DP3
Tribotek Electrical Connectors

(Courtesy of Tribotek, Inc. Used with permission.)

Performance of Woven Power

Connectors
Power density => 200% of conventional
connectors
Insertion force => less than 5% of
conventional connectors
Electric contact resistance = 5 m ohms
Manufacturing cost
Capital Investment

(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Thermal periodicity
(10) Material
periodicity
Material Functional Periodicity

Unstable Crack growth


Stress intensity factor at the crack tip
k1 = 22 c

Crack Growth under cyclic loading


Crack Growth rate per cycle under cyclic loading
dc
k1
= A
dN
Y

1-D functional periodicity
Rope
Fibers and yarns
Fabric

Fiber reinforced composites

Functional Periodicity to Control Material

Properties


Properties
Onset of plastic
instability
Force
Workhardening
region
Yield
point
Fracture
Elongation
Functional Periodicity to Control Material Properties

D islo c at io n d e nsi ty
S tr e n gt h
S tr a in
S tr a in
A nn ea lin g
S tr a in

Properties
Functional periodicity for controlling material
properties
Thin films
Superlattices
Annealing in wire drawing

Properties -- Annealing during wire drawing


Properties -- Microcellular plastics

Complexity Theory
and
the Stability of a System
For a system to be stable and survive for a long
period of time, it must either
Be at equilibrium
or
Have a functional periodicity
Examples of Functional Periodicity

Examples of Functional
Periodicity
Examples in Nature
Examples in Engineered Systems
Planetary system, solar calendar
Airline/train schedules, computers
Crystalline solids,
Undulated surface for low friction,

woven electric connectors
Cell cycle, life-death cycle, plants,

grains,
Fermentation processes such as

wine making
Manufacturing processing
periodicity
Biological systems
Scheduling a clustered
manufacturing system,
Periodic table of chemical

elements/atoms
Polymers
Thermal periodicity
Temperature of Earth,
Heat cycles (e.g., Carnot cycle),
Information processing periodicity
Language
Re-initialization of software
systems, music
Thunder storm
LCD, alternating current,
Living beings,
Light sensitive sensors
Wavy nature of matter,

crystallinity,
Fabric, wire drawing, microcellular

plastics,
Implications of the Complexity Theory
S ta b ili ty
E q uilib rium
F un ct ion a l
Pe ri o di city

Equilibrium based physics
Newtonian Mechanics
Thermodynamics

Functional Periodicity based physics
Quantum Mechanics
String theory

Periodicity based physics
Quantum Mechanics
Can we explain the particle/wave duality
based on this stability argument?
Complexity of Socio-Economic Political Systems
Economic functional periodicity

Political functional periodicity
Organizational functional periodicity
Academic functional periodicity
Complexity of Socio-Economic Political Systems
Case Studies
Economic development of ROK (1980-85)
National Science Foundation
Mechanical Engineering Department of MIT

(2) Geometric periodicity
(3)

(10) Materials periodicity
Systems Biology
The goal is to relate the higher-level FRs to the
behavior of biological molecules.
Use the knowledge in curing patients.
Use the knowledge in drug discovery.
Systems Biology
Based on the complexity theory, one may speculate that
biological systems must have a functional periodicity.
Based on axiomatic design theory and complexity
theory, one may speculate that most biological systems
are a decoupled or uncoupled systems.
How do we relate the molecular behavior to the
behavior of biological systems?
It seems that biological system follows the same

principles as engineered systems.
Robustness through functional independence

Functional periodicity
FRs of a Cell
FR1 = Isolate the cell and its components from its
environment.
FR2 = Obtain fuel.
FR3 = Convert the fuel to energy.
FR4 = Communicate with it surrounding.
FR5 = Reproduce itself.
FR6 = Control cell functions.
DPs of a Cell
DP1 = Plasma membrane (phospholipid bilayer)
DP2 = Diffusion of ions and transport of proteins
DP3 = Synthesis of ATP in mitochondria
DP4 = Receptors and signal
transduction protein
DP5 = Reproduction mechanism
DP6 = Functional periodicity
Second Level Decomposition of

FR/DPs of a Cell
FR6 (Control cell functions) / DP6 (Functional periodicity) may
be decomposed as
FR61 = Regulate growth.
FR62 = Prevent mutation.
FR63 = Control cell functions.
DP61 = Growth factor
DP62 = DNA repair and apoptosis
DP63 = cdk protein
Design Matrix of a Cell
DP1 DP2 DP3 DP4 DP5

FR1
FR2
FR3
FR4
FR5
FR6
X
X
0
X
x
FR61 x
FR62 0
FR63 X
0
X
X
x
X
0
0
X
0
0
X
0
X
0
0
X
0
0
0
X
x
0
0
X
0
0
0
0
X
0
0
0
DP6
DP61 DP62
0
0
0
0
0
0
0
0
0
0
X
0
0
X
0
X
DP63
0
0
0
0
X
0
0
X
Decomposition of FR5 (Reproduce itself) and DP5

(Reproduction mechanism)
FR51 = Initiate the replication process.

FR52 = Start chromo some replication.
FR53 = Replicate proteins.
FR54 = Replicate DNA structur es.
FR55 = Create cytoplasm skeleton.
FR56 = Form m embranes of organelles.
FR57 = Induce mitosis.

DP51 = Start kinase/ "restriction point" for G1 phase

DP52 = Chromosome replication kinase
DP53 = Cyclin-dependent protein kinase (cdk protein)
DP54 = DNA polymerase, ligase / mRNA
DP55 = Polymerization mechanisms
DP56 = "Membrane" kinase
DP57 = Tensile force

FR51
FR52
FR53
FR54
FR55
FR56
FR57
DP51
X
X
X
X
X
X
0
DP52
0
X
0
0
0
0
0
DP53
0
0
X
0
0
0
0
DP54
0
0
0
X
0
0
0
DP55
0
0
0
0
X
0
X
DP56
0
0
0
0
0
X
X
DP57
0
0
0
0
0
0
X
Decomposition of FR57 (Induce mitosis) and

DP57 (Tensile force)
FR571 = Adhere to a reference surface

FR572 = Activate WASp family protein.
FR573 = Nucleate cellular actin.
FR574 = Polymerize G-actin to make F-actin filament.
FR575 = Terminate the polymerization reaction.
FR576 = De-polymerize the F-actin filament.

DP571 = Active sites on the surface (protein
fibronectin)
DP572 = Integrin receptor on cell membrane that
binds to fibronectin and initiates signal
transduction.
DP573 = Arp2/3 complex
DP574 = Free actin/profilin
DP575 = Capping protein
DP576 = ATP hydrolysis within actin filaments
and dissociation of Phosphate/
ADF(actin depolymerization factor)

DP571
DP572
DP573
DP574
DP575
DP576
FR571
FR572
FR573
FR574
FR575
FR576
Cell and Cell Cycle

(From Alberts, et al., 3rd Edition)
Possible Lessons from Biological Systems

Every subsystems that make up the cell undergo
periodic cycles as the entire cell.
One of the important FRs of a biological seems
to be the control of functional period.
The consequence of the breakdown of the
functional periods of cell may lead to failure of
biological systems.
Lesson from Biological Systems

(from Alberts, et al., 3 rd edition)
Cell cycle control:

Controlled centrally by cyclin and cdk (cyclindependent protein kinases)
Not by a mechanism of each of the major
essential processes triggering the next event
This behavior of biological systems is consistent
with the complexity theory.
Ultimate Goals of Systems Biology

Better drug discovery process
Better cure of disease
Application of biological knowledge in
engineering systems
Linkage between biology and engineered
systems
Conclusions
The field of complexity may be a new emerging field in
science and technology.
The complexity presented in this talk provides a 40,000
feet view of the engineering systems and natural
systems.
It provides guidelines what to do and what not to do.
Conclusions
Complexity must be defined in the functional
domain, not in the physical domain.
There are the following four types of complexity:
Time-independent real complexity

Time-independent imaginary complexity
Time-dependent periodic complexity
Conclusions
Time-independent real complexity is a measure of

uncertainty in achieving a given set of FRs and thus is
related to information content.
two orthogonal components: real complexity and imaginary
complexity
Time-dependent complexity is the complexity that

changes as a function of time.
two types of complexity: combinatorial complexity and periodic
complexity.
Conclusions
Transforming a time-dependent combinatorial

complexity into a periodic complexity increases the
stability of the system and the system performance.
The transformation is achieved by introducing (or

maintaining) functional periodicity.
The periodicity is introduced by re-initializing subsystems
on a periodic interval of a set of FRs.
2.882
Complexity
April 20, 2005
Taesik Lee 2005
Complexity in AD
Complexity
Measure of uncertainty in achieving the desired functional

requirements of a system
Difficulty
Relativity
Information
Ignorance
Taesik Lee 2005
Four types of complexity in AD
Complexity
Measure of uncertainty in achieving FR
Does uncertainty
change with time?
Time-independent
Complexity
Real
Imaginary
complexity complexity
Time-dependent
Complexity
Combinatorial
complexity
Taesik Lee 2005
Periodic
complexity
Complexity: A measure of uncertainty in achieving the desired set
of FRs of a system

Measure of uncertainty when the probability of achieving the functional
requirements is less than 1.0 (because the common range is not identical to the
system range)
Time-independent imaginary complexity

Uncertainty that arises because of the designers lack of knowledge and
understanding of a specific design itself

Time -dependent combinatorial complexity arises because in many situations,
future events cannot be predicted a priori. This type of time-dependent
complexity will be defined as time-dependent combinatorial complexity.
Time-dependent periodic complexity

Consider the problem of scheduling airline flights. it is periodic and thus
uncertainties created during the prior period are irrelevant. This type of timedependent complexity will be defined as time-dependent periodic complexity.
Taesik Lee 2005
Time-independent Real Complexity

caused by system ranges being outside of the design range.
Real complexity ~ Information content
Take ui as a random variable
ui =
1 (success) with P(FRi = success)
0 (failure)
1-P(FRi = success)
Information content:
I(ui= 1) - log2P(FRi =success)
|dr|
Design Range
Common
Range,
AC
System Range,
p.d.f. f(FR)
I(p) [bit]
p.d.f.
f(FR)
5
4
3
2
1
0
drl
dru
FR
0.2
0.4
0.6
0.8
P(FRi =success)
|sr|
Taesik Lee 2005
Time-independent Imaginary complexity
Imaginary complexity ~ Ignorance

Ignorance causes complexity.
Types of ignorance
Functional requirement
Knowledge required to synthesize(or identify) design parameters
Ignorance about the interactions between FRs and DPs
p (probability of selecting a right sequence)

For uncoupled design, p= 1
For decoupled design, p= z/n!
For coupled design, p = 0
Taesik Lee 2005
Time-dependent complexity
Time-dependency
Complexity Uncertainty in achieving a set of FR
Complexity is time-dependent if
1) uncertainty (probabilistic) is time-dependent
Time-varying system range
2) behavior of FR is time-dependent
FR = FR(t)
Combinatorial / Periodic complexity

Uncertainty increases indefinitely : combinatorial complexity
Uncertainty in one period is irrelevant to the next period :
periodic complexity
Taesik Lee 2005
Origins of complexity and reduction
Taesik Lee 2005
Time-independent
Minimize Real complexity by

Eliminating source of variation
Desensitizing w.r.t. variation
Compensating error
Eliminate Imaginary complexity by
Achieving uncoupled design

Identifying design matrix
Taesik Lee 2005
Time-dependent complexity
Combinatorial complexity
Taesik Lee 2005
Periodic complexity
Time-varying system range
Detect changes in system range

Prevent system range deterioration by design
Bring the system range back into design range
by re-initialization
Taesik Lee 2005
Prevent system range deterioration by design
By eliminating coupling between turn and grasp,

one can effectively delay system range deterioration.
Milled Flat end

of the shaft
Metal
shaft
Slot
Milled Flat end

of the shaft
A
A
Injection molded
nylon Knob
(a)
(a)
(b)
(b)
Section view AA
N. P. Suh, Axiomatic Design: Advances and Applications, 2001

Taesik Lee 2005
Bring the system range back into design range: Re-
initialization

Taesik Lee 2005

Taesik Lee 2005
Example: Scheduling of mfg system
Periodicity should be introduced & maintained to prevent

the system from developing chaotic behavior
Subsystem X
Taesik Lee 2005
Subsystem Y
Problem description
Subsys
X
Station
PTi or CTY
(sec)
Number of
machines
MvPki
(sec)
MvPli
(sec)
IN
30
40
50
80
605
1
1
1
2
1
5
5
5
5
-
5
5
5
5
5
a
b
c
d
IN
Fastest speed *
70 seconds when X determines the
system speed
65-75 seconds when Y determines
the system speed
Inter
face
Subsystem Y
Objective
Maximum utilization rate for the machine Y
Constraint
Transport from C to D must be immediate
* Speed is measured by throughput time: shorter time means faster speed

Taesik Lee 2005
6th part
ready at X
5th part
ready at X
4th part
ready at X
3rd part
ready at X
FPX
1st part
ready at X
2 FPX
2nd part
ready at X
3 FPX
Taesik Lee 2005
4 FPX
Time, t
Y process
starts
5th demand
from Y
D5
z4
z3
1st demand
from Y
Y process
strats
z2
z6
FPY,5
FPY,4
6th part
ready at X
FPY,3
5th part
ready at X
FPY,2
4th part
ready at X
FPY,1
3rd part
ready at X
FPX
1st part
ready at X
2 FPX
2nd part
ready at X
3 FPX
Taesik Lee 2005
4 FPX
Time, t
CTY : 60sec - 60sec - 60sec - 60sec - 55sec -
Y FINISH
Y START
Y FINISH
CT Y = 60 sec
1 2
Transport
(next period)
Y START
Max[ CT Y ] = 6 5 sec
Machine d1
1 2
Min[ CT Y ] = 55 sec
Machine c
Machine d2
3 4
Machine b
Machine a
5 6
6
Machine c
7 8
SP=70sec
3 4
4
Machine b
9 10
Machine d1
5 6
Machine a
Machine c
7 8
SP=70sec
Machine b
9 10 Machine a
New panel
New panel
Figure 10. Steady state operation with 70 seconds sending period

From Lee, Taesik. "Complexity Theory in Axiomatic Design." MIT PhD Thesis, 2003.
Taesik Lee 2005
CTY : 60sec - 60sec - 55sec - 65sec - 65sec
Y FINISH
Max[CT ] = 6 5 sec
Y START
Panel #
2
Machine d1
Y FINISH
Y START
Max[CT ] = 65 sec
Machine d2
Machine c
7
Machine b
SP = 95 sec
Machine d1
5
3
9 10
X X X X
No Transport
9 10
Machine c
Machine a
New panel
Machine b
X X X X
9 10
9 10
SP = 75 sec
Figure 11 (b)
Y FINISH
Max[ CT ] = 6 5 sec
Y START
Panel #
Machine d1
5
Y FINISH
Y START
Max[ CT ] = 6 5 sec
Machine d2
5
Machine a
Machine c
7
Machine d1
Machine b
9 10
X X
SP = 90 sec
5
9 10
6
Machine a
Machine c
7
8
X X X X
No Transport
New panel
Figure 11 (c)
Taesik Lee 2005
Average throughput time = (75+75+75+100)/4 = 81.25
Taesik Lee 2005
Single perturbation from subsystem Y causes incomplete

period in downstream
The system regains periodicity after the perturbation is
removed but with undesirable performance
Throughput time is 81.25 seconds in average
Slower than the system capability
Taesik Lee 2005
Re-initialization scheme in scheduling
Define a renewal event that imposes period

u(0)
= { u0(t), u1(t), , uk-1(t), uk(t), uk+1(t), , uN(t)}

= {0, 0, , 0, 0, 0, , 0}
:
u(T-D) = {1, 1, , 1, 0, 1, , 1}
u(T)
= {1, 1, , 1, 1, 1, , 1}
u(T+e) = {0, 0, , 0, 0, 0, , 0} = u(0)
Scheduling activity is confined within such a period with a goal of

maintaining periodicity
Conditional renewal event
tini = trequest
if trequest 70 sec ( FPX )
if trequest < 70 sec
tini = 70sec
Taesik Lee 2005
3rd demand
from Y
2nd demand
from Y
1st demand
from Y
Y process
strats
Y process
strats
FPY,1
Y process
strats
FPY,2
5th demand
from Y
Y
4th
process
demand
starts
from Y
Y process
starts
D5
D4
FPY,3
t4
t3
t5
5th part
ready at X
4th part
ready at X
t1
Time, t
z1
z2
2nd part
ready at X
3rd part
ready at X
Each period is independent (memoryless)

Taesik Lee 2005
z6
Conclusion
Breakdown of functional periodicity results in sub-optimal

throughput rate
Periodicity should be introduced & maintained to prevent
the system from developing chaotic behavior
Taesik Lee 2005
Example: Cell division
A cell has a mechanism to coordinate cycles of two

subsystems such that the overall periodicity is
maintained
Break-down of functional periodicity leads to anomaly of
cell division and further chaotic behavior of the system
Maintaining functional periodicity in the cell cycle is an

important functional requirement for cell division
Taesik Lee 2005
Overview of the Cell Cycle
* Figure taken from Molecular Biology of the Cell, Alberts, Garland Science
Taesik Lee 2005
Chromosome cycle & Centrosome cycle

* Figure taken from Molecular Biology of the Cell, Alberts, Garland Science
Taesik Lee 2005
Importance of the correct number of chromosomes and
centrosomes
Centrosomal abnormalities
Chromosome missegregation
Aneuploidy

Figure 2 in Nigg, E. A. "Centrosome abberation: cause or
consequence of cancer progression?" Nature Reviews Cancer 2 (2002): 815-825.
* Figure taken from http://www.sivf.com.au/chromosomes.htm
Taesik Lee 2005
See Figure 1 in Nigg, E. A. "Centrosome abberation: cause or consequence of
cancer progression?" Nature Reviews Cancer 2 (2002): 815-825.
Machine
X
IN
Inter
face
Machine Y
Taesik Lee 2005
Mechanism
Cdk2 initiates both cycle

ensuring one level of
synchronization
G1Cdk
Cdk2 (S-Cdk)
promote
accumulation
Mitogen-dependent
mechanism
(Extracellular signal)
inactivate
Rb
inactivate
inactivate
G1/Scyclin
G1/SCdk
E2F
Firing origin of
replication
Centriole
split
Replicating DNA
Duplicate
centrosome
Check for completion

of DNA replication
DNA replication
checkpoint ensures
completion of
chromosome
duplication
S-cyclin
S-Cdk
inactivate
Check for
completion
Centrosome cycle lacks a

mechanism to check its completion
Prophase
Prometaphase
Metaphase
Check for
spindle
attachment
Taesik Lee 2005
Anaphase
Telophase
Hct1
APC
mutually
inhibits
Centrosome
duplication
Chromosome
duplication
gene
transcription
mutually
inhibits
Cytokinesis
mutually
inhibits
CKI:p27
Conclusion
A cell has a mechanism to coordinate cycles of two

subsystems such that the overall periodicity is
maintained
Maintaining functional periodicity in the cell cycle is an
important functional requirement for cell division
Can pose questions with new perspective
Taesik Lee 2005
Time-dependent FR
u(t) = {u1(t), u2(t), , uN(t)}
T1
(a) Periodic
T2
u5
u4
u3
u2
u1
u0
FR State
u5
u4
u3
u2
u1
u0
FR State
FR State
Periodic There exist Ti s.t. u(Ti ) = u(Tj) with regular transition pattern
Semi-periodic There exist Ti s.t. u(Ti) = u(Tj) without regular transition
pattern
Aperiodic None of the above
T1
(b) Semi-periodic
Taesik Lee 2005
T2
u5
5
u4
4
u3
3
u2
2
u1
1
u0
t
(c) Aperiodic
Uncertainty and functional periodicity
Ps(t) = P(u(t) = u*(t))

For periodic & semi-periodic FR(t), Ps returns to one at the
beginning of a new period
Predictability of FR
(Periodicity) (Predictability)
(Unpredictability) (Aperiodicity)
~ (Aperiodicity) ~ (Unpredictability)
Uncertainty in current period is independent of a prior
period only if the initial state is properly established
Taesik Lee 2005
Reduction of Complexity through the

Use of
Geometric Functional Periodicity
Reference:
Chapter 7 of
Nam P. Suh, Complexity: Theory and Applications,
Oxford University Press, 2005

uncertainty in satisfying the FRs.
According to this definition,
complexity is a relative quantity.
Four Different Kinds of Complexity

Time-Dependent Periodic Complexity

following actions:
Important concept: Functional Periodicity
Thermal periodicity
Time-Dependent Combinatorial
Complexity
arises because the future events occur in
unpredictable ways and thus cannot be
predicted.
For example, it occurs when the system
range moves away from the design range
as a function of time.
Prob. Density
T h e System Range changes

continuously as a function of time.
D es ign
Ran ge
T im e
FR

following actions:
Reduction of Combinatorial Complexity

How?
Through
Re-initialization of the system
by defining
a Functional Period.
Note: Functional period is defined by a repeating set of
functions, not by time period, unless time is a set of
functions.
Transformation of Time-Dependent Combinatorial

Complexity
Basic Idea
1.
Make sure that the design satisfies the
Independence Axiom.
2.
Identify a set of FRs that undergoes a cyclic
change and has a functional period.
3
Identify the functional requirement that

may undergo a combinatorial process
4.
T C com ( FR i ) C per (FR i )
6.
5.
"Reinitialization"
Set the beginning of the cycle as t=0

(2)

(10) Material periodicity
Example: Design of Low Friction Sliding

Surfaces
Consider the following task:

Reduce friction between two sliding
surfaces under load
Coefficient of friction versus sliding distance
i
Distance slid
(a)
Distance slid
(b)
Effect of removing wear particles for an Armco

iron slider sliding against an Armco iron specimen
s
I
Wear particles removed
i
Distance slid
Friction Space

Plowing Mechanism

Particle Agglomeration
Why do particles agglomerate?

Height of Agglomerated Particles

Friction Coefficient and the Number of
Agglomerated Particles

Reduction of Friction by Elimination of Particles

Lubricants
What are the FRs?


Lubricants

interface
Constraint:
No lubricant

Lubricants

See Figure 7.11 and 7.13 in [Complexity]:
DP1 = Total contact area of the pad, A

DP2 = Roughness of the planar surface of pads, R
DP3 = Length of the pad in the sliding direction,
DP4 = Volume and depth of the pocket for wear particles, V

Lubricants
FR1 X 0 0 0 DP1 X 0 0 0 A

FR2 0 X x 0 DP2 0 X x 0 R
=
=

FR3 0 0 X 0 DP3 0 0 X 0

FR4 0 0 0 X DP4 0 0 0 X V

Lubricants
Test Results:

Undulated Surface for Elimination of Particles
Friction at Lubricated Interface
Effect of Boundary Lubrication

~ 0.1
Cause?
Plowing
What is the role of a lubricant?
Lower shear stress

Transport particles
Prevent particle agglomeration
Prevent adhesion
Geometric Functional Periodicity to Decrease

Face Seal Wear

See Figure 7.16 through 7.21 in [Complexity].
FRs of Face Seal

FRs
FR2 = Prevent particle migration across the seal
FR3 = Prevent agglomeration of particles in the seal/metal
interface
FR4 = Provide lubricant to the interface
FR5 = Prevent the flow of the lubricant out of the sealed area
FR6 = Delay the initiation of the wear process
DPs of Face Seal

DPs
DP1 = Total Area of the pad, A
DP2 = Seal lip
DP3 = Diameter of the pad,
DP4 = Lubricant
DP5 = Length of the sealed tip, L
DP6 = Seal material
Design Matrix for Face Seal Wear
Reduction of Wear of Pin-Joints By the

Introduction of Geometric Functional
Periodicity
Drive sprockets, idlers, rollers, Grouser shoes
FRs and DPs of Conventional Pin-Joints

FRs
FR1 = Carry the load
FR2 = Allow rotational motion about the axis of the
bushing
FR3 = Minimize friction force
FR4 = Minimize wear
DPs
DP1 = Shaft/bushing dimensions
DP2 = Clearance between the shaft and the bushing
DP3 = Clearance between the shaft and the bushing
DP4 = Material properties
Friction in Geometrically Confined Space

See Mosleh, M., and N. P. Suh. High Vacuum Undulated Sliding Bearings.
SLTE Transactions 38 (1995): 277-284.
FRs and DPs of Pin-Joints with Geometric

FRs
FR1 = Carry the load
FR2 = Allow rotational motion about the axis of the
bushing
FR3 = Minimize friction force
FR4 = Minimize wear
DPs
DP1 = Dimensions of the pin and the bushing structure
DP2 = Cylindrical pin/bushing joint
DP3 = Low-friction system
DP4 = Wear-minimization system

FR 1 X 000 DP 1

FR 2 0 Xxx DP 2
=
FR 3 00 Xx DP 3
FR 4 000 X DP 4
Decomposition of FR3 and DP3 of Pin-Joints

with Geometric Functional Periodicity
FR3s
FR31 = Remove Wear Particles from the interface
FR32 = Apply lubricant
DP3s
DP31 = Particle removal mechanism
DP32 = Externally supplied grease

FR 31 X 0 DP 31
FR 32 xX DP 32

FR31s (Remove wear particles from the interface)
FR312 = Do not allow the increase in the normal force due
to the presence of particles at the interface
FR313 = Trap wear particles
DP31s (Particle removal mechanism)
DP311 = The length of the pad on the liner that actually
comes in contact with the pin
DP312 = Flexible ring
DP313 = The indented part of the undulated surface



FR 311 X 00 DP 311
FR 312 = xX 0 DP 312
00
X
DP
313
FR
313

FR4s (Minimize wear)
FR41 = Prevent particle penetration into both surfaces
FR42 = Minimize the shear deformation of the pin and the
flexible liner
DP4s (Wear-minimization system)
DP41 = Hardness ration of the two materials
DP42 = Coating of the pin with a thin layer of low shear
strength material

FR 1 X 0000000 DP 1
FR 2 0 XxxxX 00 DP 2
FR 311 00 X 00000 DP 311
FR 312 00 xX 0000 DP 312

=
FR 313 0000 X 000 DP 313

FR 32 0000 xX 00 DP 32
FR 41 000000 X 0 DP 41
FR 42 00000 X 0 X DP 42
Pin-Joint Design

Geometric Functional Periodicity for


Conventional Electrical Connectors
Male connector
Compliant pin
(for permanent connection)
Plastic
overmolding
Plastic
overmolding
Female connector
Multiple layers will be stacked together

to obtain an entire connector.
Figure by MIT OCW.
FRs of a Data Electrical Connector

than 20m)
FR3 = Prevent the cross-talk (i.e., interference)
between the connections
C1 = Low cost
C2 = Ease of use
FRs of a Power Electrical Connector

than 20m)
FR3 = Maximize power density
C1 = Low cost
C2 = Ease of use
DPs of a Data Electrical Connector
DP1 = Cylindrical assembly of the woven

tube and the pin
DP2 = Locally compliant electric contact
DP3 = Number of conducting wires
Decomposition of FR1 (Mechanically connect and

disconnect electrical terminals) and DP1 (Cylindrical
assembly of the woven tube and the pin)
FR11 = Align the rod axially inside the tube
FR12 = Locate the axial position of the rod in the tube
FR13 = Guide the pin
DP11 = Long aspect ratio of the rod and the tube
DP12 = Snap fit
DP13 = Tapered tip of the pin
Decomposition of FR2 (Control contact resistance to be

less than 20m) and DP2 (Locally compliant electric
contact)
FR21 = Prevent oxidation of the conductor
FR22 = Remove wear particles
FR23 = Control line tension/deflection of the nonconducting fiber
DP21 = Gold plated metal surface
DP22 = Space created in the crevices between fibers
DP23 = Spring
Design Matrix
FR1 X 00 DP1
FR2 = XX 0 DP2
FR3 0XX DP3


Performance of Woven Power

Connectors
Power density => 200% of conventional
connectors
Insertion force => less than 5% of conventional
connectors
Electric contact resistance = 5 m ohms
Manufacturing cost -- Flexible manufacturing
Capital Investment
Tribotek, Inc.
Busbar Application Embedded
Configuration
Features
LowR 125 socket
Press-Fit into busbar
130 A
Benefits
Highest current
density
Low insertion force
Low contact
resistance, min voltage
drop
Available in various
contact sizes
Figures removed for

copyright reasons.
Tribotek, Inc.
Busbar Application - Mounted Configuration
Figures removed for

copyright reasons.
Features
LowR 125 socket
Press-Fit onto busbar
130 A
Benefits
Highest current density
Low insertion force
Low contact resistance,
min voltage drop
Available in various
contact sizes
2.882
Chapter 8 of Complexity
Reduction of Complexity in Materials through
Functional Periodicity related to materials

found in Nature
Periodic Table of chemical elements

Crystalline solids
Atomic structure (electrons in valence bands)
Biological systems
Engineered Materials
Functional Periodicity to to prevent unstable
crack growth
Wire rope
Fabric
Composites
Crack Growth
Unstable crack growth
k1 = 22

See Figure 8.1 in [Complexity]:
Crack Growth
Fatigue crack growth under cyclic loading
n
k
dc
= A 1
dN
Y

Composites

Functional Periodicity for Control of

Material Properties
Edge and Screw Dislocations


Material Properties
Stress-Strain Relationship
showing work-hardening


Material Properties
Annealing to eliminate dislocations


Material Properties
Wire Drawing

Microcellular Plastics
(Distortion of injection molded parts)
Request of Our Research Sponsor

Eastman Kodak (Gordon Brown)
Provide a means of reducing the

consumption of plastics
Maintaining the same toughness of plastics and
the same geometric shape
The customer needs may be stated

in terms of the following FRs and Cs:
Functional Requirements:
FR1 = Reduce consumption of plastics
FR2 = Maintain the toughness of parts
Constraint:
The shape of products must remain the same.
Design of Microcellular Plastics

The highest level FRs
FR1 = Reduce the amount of plastic used
FR2 = Increase the toughness of the plastic
product
FR3 = Make three-dimensional geometrical
shape

Conceptual Solution
A large number of microscale bubbles

FR1 = Reduce the amount of plastic used
FR2 = Increase the toughness of the plastic
product
FR3 = Make three-dimensional geometrical
shape
The corresponding highest level design

parameters (DPs) are:
DP1 = Number of cells
DP2 = Cell size
DP3 = Die or mold design

The design equation for the product:
FR1
FR2 =
FR3
XX0 DP1
0 X0 DP2
00X DP3
What is a Microcellular Plastic?

Microcellular Plastics (MCP) is defined as
the plastic that has a large number of
small bubbles, typically less than 30
microns.
What is a Microcellular Plastic?

Cell Size (m)
0.1
1
10
Cell Density (Cells/ cm3)

1015
1012
109
Morphology of MuCell
Polystyrene Products
PP Products
PVC profiles
Injection Molded Printer Chassis
Injection Molding of
Microcellular Plastics vs Solid Plastics
TRW -- Air bag Canister (Material: 33% glass filled Nylon)
Solid
365 gms
45 sec
150 tons
Part weight
Cycle time
Tonnage
MuCell
252 gms
35 sec
15 tons
(Courtesy of Mar Lee Companies)
% red.
31%
22%
90%
Advantages of Microcellular Plastics

Reduction of material consumption (from 5 to
95 %)
Faster cycle time
Higher productivity
Greater toughness in some plastics
Dimensional accuracy
Dimensional stability
No warping
Appearance (no visible cells)

Thin sections
No sink marks
Low temperature process
Low pressure process
Large number of cavities or smaller machines
Most polymers
Use of non-hydrocarbon solvents -- CO2 and N2

No additives for nucleation
No reactive components such as viscosity
modifiers
No special equipment other than gas supply
system -- similar conventional machines
Question: What is the physical basis

for MuCell technology?
Simultaneous nucleation
of an extremely large number of cells!!
How do we achieve the

simultaneous nucleation of an
extremely large number of
cells?
Design of the Process Technique for

(The first student to work on the batch process -Jane Martini, SM Thesis, MIT)
The processing technique consists of

dissolving a large amount of gas to form
polymer/gas solution
and then inducing sudden
thermodynamic instability
by either lowering the pressure or raising the
temperature to change the solubility of the gas.
Sudden Change in Solubility

The solubility is a function of two
thermodynamic properties, temperature and
pressure:
S = S(p, T)
The change in the solubility can be expressed as:
S
S
S =
p +
T
p
T
by Batch Process
(From Cha, Ph.D.Thesis, MIT)
Cell sizes of the microcellular foamed polymers (Ambient temperature foaming).

Note: Saturation pressure and temperature of COs were not the same for all
polymers.
by Batch Process
(From Cha, Ph.D.Thesis, MIT)
Polymers
Cell density of the microcellular foamed polymers (Ambient temperature foaming).

Note: Saturation pressure and temperature of COs were not the same for all
polymers.
What is the basic physics behind

continuous processes?
- Design of a Continuous Process
To create a continuous process, we must be able
to design a process and associated equipment
to perform following four functions:
(1) Rapid dissolution of gas into molten, flowing
polymer to form a solution,
(2) Nucleation of a large number of cells,
(3) Control of the cell size, and
(4) Control of the geometry of the final product.
- Design of a Continuous Process
Extruder
(1) Rapid dissolution of gas into molten, flowing
polymer to form a solution,
Die/Mold
(2) Nucleation of a large number of cells,
(3) Control of the cell size, and
(4) Control of the geometry of the final product.
Physics of the Continuous Process

1. Polymer/Gas Solution in Extruders
Important Parameters for Formation of
Polymer/Gas Solution:
1. Temperature
2. Pressure
3. Degree of mixing of gas and polymer
Introduction to Physics of the Continuous

Process
i. Gas diffusion and formation of polymer/gas
solution
Brief Introduction to Physics of the

Continuous Process
Polymer
PS
PP
PE
HDPE
LDPE
PTFE
PVC
D of CO2 (cm2/s)
At 188 C
At 200 C
-1.3x10-5
4.2x10-5
--2.6x10-6
5.7x10-5
2.4x10-5
-1.1x10-4
-7.0x10-6
-3.8x10-5
D of N2 (cm2/s)
At 188 C
At 200 C
-1.5x10-5
3.5x10-5
--8.8x10-7
6.0x10-5
2.5x10-5
-1.5x10-4
-8.3x10-6
-4.3x10-5
Table 7.1 Estimated diffusion coefficients of gases in polymers at elevated

temperatures (From Durril, P. L., Griskey, R. G., AIChE Journal, Vol. 12, p
1147 (1960 and Vol. 15, p 106 (1969)
Brief Introduction to Physics of the Continuous

Process
Table 7.2 Estimated diffusion time at various striation thickness and

diffusion coefficients. (From Park 1996)
Brief Introduction to Physics of the

Continuous Process
Deformation of a spherical bubble in a shear field to form an ellipsoid. The

distance between the ellipsoids (measured perpendicular to the major axis
of the ellipsoids) is the striation thickness. The dissolution rate of gas
increases as the striation thickness becomes smaller and as the interfacial
area of gas/polymer increases.

Gas solubility
Polymer
CO2 weight gain (%)
N2 Weight gain (%)
PE
14
PP
11
PS
11
PMMA
13
Table 7.3
Estimated gas solubility in polymers at 200C and 27.6
MPa (4,000 psi) (Park, 1993).
How high should the pressure be in

the extruder or the plasticating
section of the injection molding
machine?
How high should the pressure be in the extruder

or the plasticating section of the injection molding
machine?
Pressure should be high enough to

prevent the formation of two phase
throughout the system until ready for
cell nucleation.

2.
Nucleation of Cells

Classical Nucleation Theory
Free energy change due to the formation of bubbles

G = Gv + Gs
Critical cluster is formed when
d
(G) = 0
dr

Classical Nucleation Theory
(From Sanyal, Ph.D.Thesis, MIT)
Nucleation rate
dN
16
RT
= J = N0 f exp
3
3kT
Pg
dt

*
Pg A zU(1 2) + kT ln(
) + RT ln
1
P
Nucleation Theory
Nucleation rate:
dN
[G G *(gas concentration)]
= N 0 exp
dt
kT
G = Activation energy barrier
= f ( impurities, mol.orientation, etc.)

G* = Energy change due to supersaturation of gas
Possible Nucleation Sites and G

(Approximately Lowest to Highest)
Solid/polymer interface (heterogeneous nucl.)

Non-polar polymer/polar polymer interface
High strain region
Free volume
Crystalline/amorphous interface in a polymer
Interface between crystallites
Morphological defects in a polymer
Polar groups of polymers
Possible Nucleation Sites and G

(Approximately Lowest to Highest)
Potential sites
Solid/polymer interface
Non-polar polymer/polar
polymer interface
High strain region
Free volume
Crystalline/amorphous
interface in a polymer
Interface between crystallites
Morphological defects in a
polymer
Polar groups of polymers
Rough estimation of potential

number of nucleation sites (??)
105 to 106 /cc
--
--109 /cc
1012/cc
1018 /cc
---
1022 /cc
Probability Distribution of Activation

Energy
Prob.
D en sity
pdG
f of G
A ctiv ati on
Ene rgy Leve l
Effect of Gas on the Probability Density of

Activation Energy
Prob.
De nsity
pd f of
(G-G*)
GG*
Effe ct of gas
on pdf of G
pd f of
G
G
Energy
Level
N0 as a function of % gas dissolved

N0
(N0)max
(N0)min
%gas
dissolved
How do we achieve
the simultaneous nucleation
of an extremely large number of cells?
Answer #1:
By creating a large driving force that can
easily overcome all activation energy
barriers for nucleation
How do we achieve
Answer #2:
By making sure that the nucleation
rate is faster than the diffusion rate.
Nucleation Theory
Condition for Simultaneous Nucleation:
(From Baldwin, Ph.D.Thesis, MIT)
Characteristic nucleation time

<< 1
Characteristic diffusion time
dN
dc
dt
<< 1
Nucleation Theory
Condition for Simultaneous Nucleation:
(From Baldwin, Ph.D.Thesis, MIT)
Characteristic gas diffusion distance

<< 1
Characteristic spacing between stable nuclei
2 c1/ 3 ( tD )1 / 2 << 1
How do we achieve
Answers:
1. By creating a large driving force
2. By making sure that the nucleation
rate is faster than the diffusion rate.
How can we make the nucleation

time as short as possible and the
driving force as large as possible?

time as short as possible and make
the driving force as large as
possible?
1. Make (-dp/dt) large
Why do we need to make (-dp/dt) large?

pressu re
Time
How large should dp/dt be?

Answer
It depends on the following:
(1) the temperature of the plastic
(2) materials
(3) die or mold design

Polymer
PS
PE
HDPE
LDPE
PTFE
PVC
PP
D of CO2 (cm2/s)
@ 188
@ 200 C
-1.3x10-5
-2.6x10-6
5.7x10-5 2.4x10-5
-1.1x10-4
-7.0x10-6
-3.8x10-5
4.2x10-5 --
D of N2 (cm2/s)
@ 188 C @ 200 C
-1.5x10-5
-8.8x10-7
6.0x10-5 2.5x10-5
-1.5x10-4
-8.3x10-6
-4.3x10-5
3.5x10-5 --
Table 7.1 Estimated diffusion coefficients of gases in polymers at

elevated temperatures (From Durril, P. L., Griskey, R. G., AIChE
Journal, Vol. 12, p 1147 (1960 and Vol. 15, p 106 (1969)
How large should (-dp/dt) be?

Typical value:
|dp/dt| > 1 G Pascal / second

cite (N0) as large as possible?
1. Increase the level of supersaturation!!
2. Dont put nucleating agent!!
3. Introduce internal strains by stretching,
etc.
Cell Density vs % Gas Dissolved

C ell
De n si ty
C o n ce n trati on of
D i ss o l v e d G as
CELL NUCLEATION DENSITY

Polystyrene as a function of N2 Sat Pressure
(from Kumar, Ph.D. Thesis, MIT)
-3
CELL DENSITY, cm
1010
10 9
10 8
10 7
10 6
0
500
1000
1500
2000
SATURATION PRESSURE, psi
2500
CELL NUCLEATION DENSITY

Polycarbonate as a function of N2 Sat. Pressure
CELL DENSITY(*10 ), Cm
-3
(From V. Kumar, 2000)
35
30
25
20
15
10
5
0
0
SATURATION PRESSURE, MPa

3. Cell Growth
Physics of the Process

Cell Growth in a Batch or Continuous
Process (From Baldwin, Ph.D. Thesis, MIT)
Governing Relationships
2
3 dR 2
d
R
dR
2
cell
cell
cell + Rcell
=
-4
+ Pg P
2
dt
Rcell dt
2 dt
Rcell
P
R
g cell
2 c
| Rcell
= 3 p Dr
t RT
r
dRcell
c Rcell dt c 1 2 c
+
= 2 r D
2
t
r
r r r
r
2

Process
Initial condition
c(r, 0) = c i
Boundary conditions
c(Rcell ,t) = Ks P
c
=0

t t, r

Process
Cell Size Control in Free Expansion

Forces due to the pressure in the bubble
Viscous forces resisting the expansion
f ( p, Tgas , r )
=1
g( Texterior layer , T r
, )
exterior layer
=1

Cell Growth in a Batch or Continuous Process
Cell Size Control by Imposition of Geometric Constraint

Free expansion of bubbles
=1
Geometric constraints
g( Geomtry )
=1

For Uniform Cell Growth in an
Intermittent Processes
Characteristic flow rate in the mold
Rate of expansion of cells
>1
f (Vinjection )
g( , T polymer , c total gas , gas in
>1
polymer
Design of a Continuous Process

FRs --> DPs
DPs --> PVs
Design of a Continuous Process for Sheet

Extrusion
FR1 = Control cell size
FR2 = Control the number of cells
FR3 = Control the geometry of the
extrudate
Design of a Tube Die

H
Di e
P0
Pi
L
(Len gt h o f
th e Die Lip )
Cen ter
Lin e
Design of Microcellular Plastics Process

The corresponding design parameters (DPs) are:
DP1 = Pi*
DP2 = dp/dt
DP3 = Die shape & Accessories
* Assuming that Pi* is the saturation pressure for
the dissolved gas.

The design equation may be written as
Cell size X x 0 Pi
Cell density = X X 0 dp dt
x
0
X
Geometry
Die
&
Acc.
Design of Microcellular Plastics Process
FR3 and DP3 must be decomposed to develop

detailed means of controlling the geometry.

The corresponding Process Variables (PVs) are:
PV1 = Extruder rpm,
PV2 = Die length, L
PV3 = Means of controlling the geometry of
the extrudate

The corresponding design equation for the
process :
Pi
X x 0
dp dt
= XX 0 L
Die & Acc. 0 0 X Means ...
Typical Nucleation Rate in a Parallel Die

Schematic Diagram Adapted from Sanyal, PhD. Thesis, MIT
Nu cl.
Ra te
0.5
X/L
Typical Cell Growth in a Parallel Die

Schematic Diagram Adapted from Sanyal, PhD. Thesis, MIT
Ce ll
Ra diu s
0.5
X/L
Design of a continuous Process
Graph removed for copyright reasons.
Representative pressure profile along the polymer flow field in

the extruder and die (From Baldwin, Park and Suh, 1997)
Design of a continuous Process
Diagram removed for copyright reasons.
Figure 7.16 Schematic of the Microcellular extrusion system

used for the shaping and cell growth control experiments
(From Baldwin, Park and Suh, 1997)
Unanticipated Processing Advantages

of Microcellular Plastics
Decrease in viscosity
Lowering of the melting point and glass
transition temperature
Elimination of shrinkage
Effect of CO2 on Physical Properties:
Viscosity
% Gas
D i ss o l v e d
Rigid PVC @ 340F

Rigid PVC @ 340F
Viscosity(Ns/m )
1000
800
600
No gas
400
0.5% CO2
200
0
0
1000
2000
3000
Shear Rate(1/s)
4000
5000
Santoprene @ 340F
Santoprene @ 340F
Viscosity(Ns/m )
200
150
No gas
100
1.5% CO2
50
0
0
5000
10000 15000
Shear Rate(1/s)
20000 25000
ABS @ 370F
ABS @ 370F
Viscosity(Ns/m )
600
500
No gas
400
3% CO2
300
8% CO2
200
12% CO2
100
0
0
10000
20000
30000
Shear Rate(1/s)
40000
Glass Transition Temperature

DMA result (PETG) [From Cha & Yoon, 2001]
Graph removed for copyright reasons.
Behavior of PETG material as a function of temperature.

Elastic modulus sharply decreases around the glass transition temperature
Models for estimating Tg [From Cha & Yoon, 2001]

Chows Model (1980)
Tg
ln
T
go
= {(1 ) ln(1 ) + ln }
Cha-Yoon Model (1998)
Tg = Tgo exp (M p )
1/ 3
()
1/ 4
: specific density of polymer

: material constant
: % solubility for carbon dioxide in polymer (weight gain)
Glass Transition Temperature

[From Cha & Yoon, 2001]
Comparison of the Two Models
Two graphs removed for copyright reasons.
Experimental results of Tg of
PETG
Comparison of Cha-Yoon model vs.

Chows model
Microcellular Foaming/Forming
Thermoforming : the process of manufacturing products from a

thermoplastic sheet heated to its softening point and formed by
pressure difference into a molded shape
Experimental
- Cha-Yoon model indicates when the weight fraction of
CO2 dissolved in PETG is 7.7% the Tgis lowered to room
temperature
Photographs of two cups
made through microcellular
foaming/forming experiment
at room temperature
Effect of Dissolved Gas on

Processing of Microcellular Plastics
Higher throughput rate by as much as
50%
Faster cycle time by as much as x2
Precision parts
Cycle Time Reduction

Start of
Injection
Hold
Cooling
Solid
Hold
Cooling
20% - 50% overall

cycle savings
MuCell
Hold & Pack
is eliminated
25% less
cooling time
MuCell Molding Technology

Office Equipment
Dimensional Stability Required:
Glass filled engineering resins such as
PPO, PC or PC/ABS
Out of spec product made to spec. using
50% reduction in warpage - .060 to .027
25% reduction in cycle time, 8% weight
reduction
Photo removed for

copyright reasons.
Printer Chassis

Impact Performance
Printer Chassis:
35% Glass/Mineral filled
PPO/HIPS
Photo removed for

copyright reasons.
8% weight reduction
Drop Weight Impact
Notched Izod Impact
Solid
6.7 ft-lb
7.3 kJ/m2
8% Wt. Red.
9.0 ft-lb
9.7 kJ/m2
Complexity of Socio-Economic Political

Systems
Application of Functional Periodicity
Complexity of Socio-Economic Political

Systems
Axiomatic design and complexity theory
appear to be equally applicable to nontechnical fields such as economic
development, government, and educational
institutions.
Socio-Economic Political Systems

Korean Economic Development Plan 1980-85
National Science Foundation 1984-88
MIT Department of Mechanical Engineering
1991-2001
Transformed a system with time-dependent

combinatorial complexity to a system with
time-dependent periodic complexity
In some cases, introduced new FRs
Development of the Five-Year (1980-85)

Economic Development Plan of the Republic
of Korea
History of Industrialization of South Korea

1945 -- Liberation of Korea with the ending of the
Second World War (no industrial output)
1948 -- Republic of South Korea under Syngman Rhee
1961 -- Military coup detat (General Park, Junghee)
1961 to 1970 -- Labor intensive industries (textile,
apparel, shoes, etc.)
1970 -- Heavy industries (automobiles, shipbuilding,
steel making, machine tools, earthmoving equipment,
etc.)
1979 -- Assassination of President Park
July 1980 --
Status of Industrialization in 1980

Ambitious industrialization plan based on the
success of the labor intensive business and overseas
construction business
Capital formation through the concentration of
capital and privileges in a few business groups
Government guaranteed loans
Rapid pace of investment
Corruption
Lack of domestic market for automobiles, etc.
Export oriented economy
Lack of technologies
Poor planning (duplication of efforts, market, etc.)
Status of Industrialization in 1980

Time-Dependent Combinatorial Complexity Spiral
was created.
Heavily borrowed capital

Heavy losses
More borrowing by going into another business
Cash flow driven business
Unsustainable debt load
Under investment in some areas
Corruption created by the government approval and
control of loans, etc.
Development of the Economic Plan for 1980-1985

Accepted the highest-level FR (FR= Develop heavy
industries for Korea)
Assessed the status of the following industries/sectors
Automobiles
Shipbuilding
Machine tools
Power plants and machinery
Earthmoving equipment
Small businesses
Research infrastructure

Many of the FRs were not changed.
Some FRs, DPs, and PVs were changed.
Imposed Constraints
Automobiles
Shipbuilding
Machine tools
Power plants and machinery
Earthmoving equipment
Small businesses
Research infrastructure

Changed DPs.
Imposed Constraints
Automobiles
The domestic market must be large enough to support the
industry. A country must have at least 30 million people.
Korea is large enough to support automobile industry.
Minimum production volume of 400,000 passenger cars/year
Volume too small to support three competing companies
One company should specialize in passenger cars, another in buses,
and the third in trucks until the volume can justify expansion.
Increase the domestic demand rate.

Changed DPs and PVs
Imposed Constraints
Shipbuilding
Export oriented industry.
Korea can be competitive in shipbuilding -- labor intensive, human
resource, etc.
Insufficient capital and too high debt load
Convert bank loans to equity.
Combine the businesses to create an internationally competitive
firm.

Changed FRs
Imposed Constraints
Power Plant equipment
Korea cannot be competitive.
Too much investment for the available market.
Should convert the plant being built to other purposes.

Changed FRs
Imposed Constraints
Machine tool industry
Small market worldwide
Korea needs machine tool industry.
Needs more technology and human resource.

Changed FRs
Imposed Constraints
Research Infrastructure
Combine KIST and KAIST
KIST should specialize in a few fields, e.g., automobile related
technology.
Support small business
Low interest loans

Regional engineering experimental stations
Require large firms to subcontract government work
Simplify the government approval procedure
Transformation of the National Science

Foundation Engineering Directorate (198488)
NSF Act of 1950, as Amended

Promote progress of science and engineering
To provide welfare, health and prosperity
To secure national defense
others
Organization of NSF Engineering
See Figure 10.1 in Suh, N. P. Complexity: Theory and Applications.

New York, NY: Oxford University Press, 2005.
Problem Definition
A wrong set of FRs.
Even for the right FRs, there was timedependent combinatorial complexity
problem.
Not enough fund for engineering research
and education.
Entrenched PVs.
Greater presence of engineering in Federal
Government
Vicious cycle for universities and funding
agencies
New Set of FRs

Strengthen the engineering science base.
Create the science base for fields in which the
science base is absent.
Support emerging technologies.
Support critical areas of technology.
Promote engineering systems research by
supporting group efforts.
Encourage innovative research.
Strengthen undergraduate engineering
education.
New (1985)
NSF Engineering Directorate Structure
See Figure 10.2 in Suh, N. P. Complexity: Theory and Applications.

New York, NY: Oxford University Press, 2005.
Transformation of the MIT Department of

Mechanical Engineering (1991-2001)
Objective of MITs ME Department

Transformation of the field of mechanical
engineering from a discipline that has been primarily
based on
physics
into one that is based on
physics, information, and biology,
while maintaining a strong foundation in design.
New FRs for the Department

Information technology
Bio-Instrumentation
Engineering systems (manufacturing systems)
MEMS and nanotechnology
Energy
Specific Goals of the Department

Research Emphasis
Near the Two Ends of the Research
Spectrum
Research Emphasis of the ME Department

R e se a r c h
A cti vi t y
L e ve l
I mp ac t
of
R e se a r ch
Bas ic ( F u nda me n ta l )
R e se a r ch
R e s ear c h Sp ec tr u m
T e ch no logy
I n nova ti on
Characteristic of Academic Research

Academic research often deals with
legitimate, well-understood problems within
their disciplines
rather than
the exceptional, unorthodox work that

creates revolutions in science and
technology
-- From Chaos by Jame Gleick --
How did we manage the change?

To deal with the challenges and the opportunities of
our era, we established new goals and examined how
we should manage the required changes.
We identified the new fields/topics in which we
should pursue excellence in education and research.
We hired new faculty members with different
expertise to complement the background of the
existing faculty members.

Out of 26 or new faculty members hired, about 50%
has degrees in physics, computer science, electrical
engineering, applied mathematics, biology,
chemistry optics, and materials.
We created new research laboratories and facilities.
We changed the undergraduate program.
New Laboratories and Facilities

We were able to receive a number of large gifts for our
programs and laboratories.
We renovated 75% of the physical facilities.
The AMP Material Laboratory
Rohsenow Heat and Mass Transfer Laboratory
Pappalardo Undergraduate Teaching Laboratories
DArbeloff Laboratory for Information Systems
Der Torossian Computational Laboratory
Hatsopoulos Micro-Fluids Laboratory
Laboratory for The 21st Century Energy
Cross CAD/CAM Laboratory
Cross Student Lounge
The Park Lecture Halls

We created funds for a number of faculty chairs,
textbooks, etc.
MIT/Pappalardo Series of Mechanical Engineering
Books by Oxford University Press
We partially changed doctoral programs.
Cost Engineering in Axiomatic Design

2.882
May 2nd, 2005
MIT
Cost: Whats the problem?

Cost nightmare
Ingalls Shipbuilding Co.

Continuous design change, required
by the Navy
$2.7B unsettled claim
MIT
USS Peleliu
$2.6B (1982) to $14.6B (2004)
Bostons Big Dig
The Big Dig
Freiman Curve
20
Final Program Cost
The Freeman Curve
1 = Underestimates lead to
disaster
2 = Realistic estimates
minimize final costs
10
3 = Overestimates become
self-fulfilling prophecies
1
10
20
0
0
Estimated Cost
MIT
Figure by MIT OCW. Adapted from Freiman, F. R. The Fast Cost Estimating Models. AACE Transactions (1983).
Fail to address:
Credible cost estimation

Cost drivers
Cost management
Schedule risk
WHY does a system

cost how much?
We need
Systematic approach
Better utilization of cost
data
MIT
TRACEABILITY
Goals & Approaches

Goals
Enhance the credibility of life cycle cost estimation
Quickly predict the cost impact of engineering changes to
the system
Identify key cost drivers to guide the cost reduction effort
Approaches
Develop a general method for integrating cost information
into the system architecture
Development, Production, Operation phase
Develop a method to predict the cost of system changes

Requirement changes, Solutions changes
Integrate the method into a usable tool for designers/

engineers
MIT
Value
Cost Credibility

Credibility
Cost information is tied to system design information via

Axiomatic Design (AD) framework
Scope of cost estimation becomes apparent
Unified framework for cost estimation at different stages of

system design by constructing a hierarchical structure
Completeness, Visibility
Capability to Assess the Cost Impact of Changes

Changes
Ramification of changes in a requirement and/or design

solution is captured by AD framework
The estimate of cost impact is quickly generated to support

decision making process
Key cost drivers can be identified

Traceability
MIT
System Design & Development

TECHNICAL CREDIBILITY
CREDIBLE COST ASSESSMENT
NASA Requirements
(Level 1 & 2)
Customer
Needs
Technical Scope Defined

is the Scope Estimated
Identifies Lowest
Level Requirements
& Interactions
Define
FRs
Estimate the
Systems Physical
Solutions
Detailed
system
Satisfy system
morphology
System changes
are assessed
Construct local
assemblies
Map to
DPs
Establish
interfaces
Build FR -DP
hierarchy
(Top -down)
Decompose
Define
Modules
Identify
physical
components
Mapping DPs into

physical entities
MIT
Integrate
physical entities
(Bottom -up)
Software Tool to Aid the Process

MIT
Software Tool Demonstration

SDCM Development Module
DP1.3.1 is Evaluated
for Change
Nature of a change
is further specified
Simulation input (change, project phase), Result

(cost impact), and Details are summarized
% completion
(hr)
Adjusted Cost
Baseline Cost
50%
Time when the change is introduced

(User selects this point)
MIT
Development Module Model Structure
Task
Model
FR111
FR112
FR121
FR122
DP111
DP112
DP121
DP122
CU2, $
FR1111
FR1112 FR1211 FR1212
DP1111
DP1112
DP1211
CU21, $
CU22, $
DP1212
Comment
MIT
Affected Design
Parameters Identified
u0, Work Vector (hours)
Installation
Test
Fabrication
Tooling
Design
Installation
Test
Fabrication
Tooling
Design
Installation
Test
Fabrication
CU4: Insulation
400
40
100
0.05 0.10
Design
Tooling
Fabrication
Test
Installation
0.05 0.10
Design
Tooling
Fabrication
Test
Installation
0.05 0.10
Design
Tooling
Fabrication
Test
Installation
0.05 0.10
0.10 0.10 0.50 0.10

0.50
0.20 0.10
0.50
0.70
150
150
75
20
10
0.10
0.40
0.80 0.80
0.10 0.10 0.50 0.10
0.50
150
150
75
20
10
0.20 0.10
0.50
0.70
0.10
0.40
0.80 0.80
0.10 0.10 0.50 0.10
0.50
150
150
75
20
10
0.20 0.10
0.50
0.70
0.10
0.40
0.80 0.80
Input
Output
CU3: Nose TPS
Conceptually Design
Assemble
Test
Design
Tooling
Fabrication
Test
Installation
CU1: Body
Winward
CU13, $
CU0:
Thermal
Protecti
on
System
DP12
CU2: Wing
Winward
DP11
CU3: Nose
TPS
FR12
CU2: Wing Winward
Tooling
CU12, $
FR11
Installation
Process
Matrix
CU1: Body Winward
Test
CU11, $
Test
CU1, $
CU4:
Insulation
Functional
Design
Change
DP1
Assemble
FR1
Conceptually Design
CU0: Thermal
Protection
System
Design
Cost
Domain
Tooling
Physical
Domain
Task/Process
model
Design
Functional
Domain
Cost
Domain
Fabrication
Axiomatic Design
Framework
Linking Functional
and Costing Domains
0.10 0.10 0.50 0.10

0.50
0.20 0.10
0.50
0.70
0.10
0.40
0.80 0.80
150
150
75
20
10
Development Module Model Structure
Identify affected
physical interfaces
between
components
CU 1
CU
2
......
......
CU k
Affected
Components
Input
Output
Comment
MIT
design
tooling
fabricate
test
design
code
test
.....
.....
design
perform analysis
tooling
fabricatio
n
test
Tasks Required to
Implement
Changes
Iterative Model
Determines Time
to Complete the
Changes
$ - -

$ - -
$ - -

$ - -
$--

$ - -

$ - -
.....
.....
$ - -

$ - -
$ - -

$ - -
$ - -

Cost Required to
Implement Changes
Iteration model
Task A Task B
Task A
Task B
0.5
0.3
Work Transformation Matrix

% completion
(hr)
un = WT un 1
U = u0 + u1 + Lu N
= u0 + WT u0 + WT (WT u0 )L
= (1 + WT + WT 2 + L + WT N ) u0
Work completion curve
for the original design
Time (iteration step)

MIT
Cost Impact of Change
Cost penalty = Y - X
% completion
(hr)
Remaining work for

original design, X
50%
User selects this point

MIT
Summary
Complete Traceability from Design to Cost Information
Task
Model
CU13, $
FR121
FR122
DP111
DP112
DP121
DP122
CU2, $
CU21, $
CU1: Body
Winward
FR112
DP1112
DP1211
DP1212
MIT
40
100
0.05 0.10
Tooling
0.10 0.10 0.50 0.10
150
0.20 0.10
150
0.50
Fabrication
0.10
75
0.80 0.80
20
10
0.50
0.70
Test
Installation
0.40
0.05 0.10
Tooling
Fabrication
0.50
Test
0.70
Design
Tooling
0.10 0.10 0.50 0.10
150
0.20 0.10
150
75
0.50
0.10
20
0.40
0.80 0.80
10
0.05 0.10
0.10 0.10 0.50 0.10

0.50
Fabrication
Tooling
Fabrication
Test
Installation
150
150
0.20 0.10
0.50
0.70
Test
Installation
Design
u0, Work Vector (hours)
Installation
Test
Fabrication
Tooling
Design
Installation
Test
Fabrication
Tooling
Design
Installation
400
Installation
CU4:
Insulation
DP1111
CU2: Wing
Winward
FR1211 FR1212
CU3: Nose
TPS
FR1111 FR1112
CU4: Insulation
Assemble
Test
Design
CU22, $
CU3: Nose TPS
Conceptually Design
Design
FR111
CU2: Wing Winward
Test
DP12
Installation
DP11
CU0:
Thermal
Protecti
on
System
FR12
Test
CU12, $
FR11
Fabrication
Process
Matrix
Tooling
CU11, $
Design
CU1, $
Test
DP1
Assemble
FR1
Conceptually Design
CU0: Thermal
Protection
CU1: Body Winward
System
Fabrication
Cost
Domain
Tooling
Physical
Domain
Design
Functional
Domain
75
0.10
0.40
20
10
0.80 0.80
0.05 0.10
0.10 0.10 0.50 0.10
150
0.20 0.10
0.10
150
75
20
10
0.50
0.50
0.70
0.40
0.80 0.80
Steps
1.
2.
3.
4.
5.
MIT
Identify affected DPs from FR change

Identify CU corresponding to DPs
Identify CU from CU
Estimate % rework input
Estimate total change-workload

2882 Spring 2005

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2882 Spring 2005

Uploaded by

Copyright:

Available Formats

2.

Major Topics to be covered

Introduction -- Read Chapter 1 of AD

Will email Homework Problems

Why Axiomatic Design

Relationship between design and analysis

Feedback loop between analysis and

Scientific paradigm -- reductionism

Synthesis -- Many FRs

Relationship between design and analysis

Figure by MIT OCW.

Typical Approach to Realization and

Poor Planning, purely experience-based

Product Development: Typical Approach

Product Development:Axiomatic Approach

TMA Projection System

Several slides describing TMA projection

The MIT CMP machine

Copper Damascene Process

Photo removed for

How do you establish science base in

Axiomatic Design applies to all

manufacturing, materials, etc.

LCD Projector Design

Several slides removed for copyright reasons.

A C lu ste r o f tw o m a ch in e s th a t a re p h y sica lly co u p le d to

Is the IC engine a good design?

How would you improve the design?

Conventional Engine is highly coupled!

There is no way we can satisfy the EPA

February 7, 2005 Lecture

Think functionally first !!

Is this knob a good design or a poor design?

Figure removed for copyright reasons.

Is this knob a good design or a poor

What are the functional requirements of

Which is a better design?

Solution: The one on the right. Why?

Typical Design Process

Figure by MIT OCW.

Who are the Designers?

Example: Refrigerator Door Design

Figure removed for copyright reasons.

Mapping from Customer Needs to Functional

Arrow's Impossibility Theorem

Consider the case of having three choices, A, B and

The results show that the group is confused as to what it wants. It

Creativity and Axiomatic Design

Axiomatic design enhances creativity

Historical Perspective on Axiomatic Design

Axiomatic Design Framework

Characteristics of the four domains of the design

Physical Domain {DPs} Process Domain {PVs}

People and other

Human and financial

Table by MIT OCW. After Table 1.1 in [Suh 2001].

Process variables that

The Design Axioms

The Information Axiom

Minimize the information content of the

Example 1.3 Beverage Can Design

How many physical parts does it have?