October 3, 2002
© Massachusetts Institute of Technology
Ph.D. Student: Eun Suk Suh
Sponsor: Dr. David Chang, General Motors R&D
Olivier de Weck
Assistant Professor
Dept. of Aeronautics & Astronautics
Engineering Systems Division
A Two-Level Optimization Approach
Platform Architecture
October 3, 2002
GM Corvette
© Massachusetts Institute of Technology
VW Golf
3 - Discussion
2 - Automotive Platforming Example:
A Two-Level Optimization Approach
1 - Introduction to Platform
Architecture in Products
Outline
October 3, 2002
© Massachusetts Institute of Technology
1 - Introduction to Platform
Architecture in Products
October 3, 2002
© Massachusetts Institute of Technology
The embodiment of concept, and the allocation of
functionality and definition of interfaces among the
elements. (Crawley)
The arrangement of the functional elements into
physical blocks. (Ulrich & Eppinger)
Definition of Architecture
October 3, 2002
Dining
Room
© Massachusetts Institute of Technology
Residing
Food
Eating
Entertaining
Parlor
Food
Preparing
Kitchen
Connects
to
1st Floor
Relaxing
Moving
around
Porch
...
2nd floor
Hall
House
Example: Private Residence
A defined set of common or shared
elements and its interface definition
© Massachusetts Institute of Technology
Product 2
E
Product 1
B
A
D
A
C
October 3, 2002
B
C
B
E
Product 3
A
C
Elements can be all kind of architectural elements ,
e.g. parts, components, systems, processes,
organizations - objects or processes
Platform =
Definition of Platform
C
P3
A
B
E
P2
October 3, 2002
B
P2
C
E
P3
D
Interconnections: A
P1
A
© Massachusetts Institute of Technology
E
D
C
B
A
DSM:
Platform = common set= ABC
P1 ∩ P 2 I P3 = { A, B, C}
D
P1
Set Theory:
Set Theory - DSM
VW
SKODA
Fabia*
Felicia
Octavia
Beetle
Bora
Golf
Polo
Lupo
new W8
model?
Passat
new W8
model?
October 3, 2002
SEAT
Arosa
Cordoba
Toledo
Leon
new W8
model?
AUDI
A2**
TT
A3
A4
A6
A8
© Massachusetts Institute of Technology
Ca. 65% common parts
Segments A&B
“Small Class”
Segment C
“Medium”
Segment D
“Medium High”
Segment E
Upper
Class”
“
Concept
D***
(A00, A01, A02,
A03)
Platform A0
(A4, A5)
Platform A
(B5, C6?)
Platform B(C?)
(D1)
(incl.Bugatti EB118)
Platform D
Platforming - VW Example
October 3, 2002
Ref: K. Otto
© Massachusetts Institute of Technology
Leverage for rapid next generation
development
Generational Platforms
Same function but different capacity
Scalable Platforms
Create functionally different variants
Modular (Functional) Platforms
Platforms
Computers
BWB
Photo Films
BWB
Examples
Classification of Platforms
October 3, 2002
200
250
350
400
© Massachusetts Institute of Technology
300
Scaling in size
• Identical Wings
• Identical Cockpit
• Identical & Similar Bays
BWB Family covering 200-450
passengers with:
Source: Boeing
450
Boeing Blended Wing Body - 1
C2ISR
Bomber
Commercial Family
Tanker
Global Reach Freighter
Bomber
October 3, 2002
© Massachusetts Institute of Technology
Source: Boeing
Share Common Wing, Cockpit and Centerbody Elements
Tanker
Representative
Cross Sections
C2ISR
Global Range Transport/Tanker
BWB Common Fleet
© Massachusetts Institute of Technology
… based on empirical evidence
Systems with common basic sets of attributes
Long lifecycle, distributed ownership
Highly interconnected systems with need for future growth
Products in rapidly changing environment
Products that include “fast clockspeed” technologies
Products with stable core functionality but variability in
secondary functions and/or external styling
Products with peripheral customization architecture
October 3, 2002
•
•
•
•
•
•
•
When we have ….
Platforming makes sense
© Massachusetts Institute of Technology
Ref: E. Fricke
Single-use or short lifecycle products without need
for product variety - commodities
Systems insensitive to change over time
Single function products
Stable, unchanging packaging or design
Slowly changing markets
Ultra-high performance markets with no
performance loss allowables
October 3, 2002
•
•
•
•
•
•
When to avoid Platforming
October 3, 2002
Low-End
Mid-Range
High-End
Luxury
Brand A
Brand B
Brand C
Brand D
© Massachusetts Institute of Technology
Vertical
Leveraging
No Leveraging
“Market
Segment”
Ref: Meyer,Simpson
Beachhead
Approach
Horizontal
Leveraging
Usually start with some market segmentation grid
Platforming Strategies
October 3, 2002
© Massachusetts Institute of Technology
• Standard manufacturing processes and tooling
• Faster response to changing market needs (always?)
• Reduce design risk and cost
• Easier coverage of market niches (Meyer and Lehnerd 1997)
• Shortening product design lead times (Ulrich 1995)
• More standard parts (Martin and Ishii 1996)
• Reduction of inventory (Baker et al. 1986)
Advantages - Reported Benefits
October 3, 2002
© Massachusetts Institute of Technology
• Effect of platforming on long-term product innovation
• Loss of performance competivity if degree of commonality is
chosen too high and market segment is price insensitive
• Cannibalization of high end products by low end products of
the same platform product family, when customer awareness
is high (e.g. Golf versus Skoda)
• Introduction of undesirable functions and unexpected
technical problems in different variants based on the same
platform (e.g. Audi TT problems with rear wheel pressure)
Disadvantages - Potential Downsides
How does the competitor’s position affect the
platform strategy?
(3)
October 3, 2002
What is the right trade-off between degree of
commonality between variants and loss of
distinctiveness (performance compromise)?
(2)
© Massachusetts Institute of Technology
… there are many more
Given a product family with N products to be placed
in M market segments - how many platforms to use?
(1)
Interesting Research Questions
October 3, 2002
© Massachusetts Institute of Technology
A Two-Level Optimization
Approach
2
Automotive Platforming Example:
Name
Compact Car
Medium Sedan
Luxury Sedan
Sports-Roadster
SUV
Truck
Van
October 3, 2002
16.8 M/year
Size
2,357,802
4,198,028
1,591,438
514,837
3,519,461
2,800,104
1,589,958
16,571,628
Mean Price
$13,427
$19,844
$34,238
$23,424
$25,146
$22,805
$24,986
What is the right strategy?
#vhc
30
33
65
34
56
51
24
© Massachusetts
Institute
of Technology
Source:
NIADA National
Market
Report - 2002
Total U.S. Market 2001 ca.
New Vehicle Sales
Symbol
LOWC
MDSD
LXSD
SPTR
SUVC
PUPT
MVAN
We are a (new) automotive manufacturer and want
to compete successfully in these market segments:
Problem Setup
October 3, 2002
© Massachusetts Institute of Technology
• The targeted market segments are known
• One product (vehicle) per market segment
• The basic vehicle architecture is given
• Market segments operate independently
• Competitors continue to offer the same
• The fixed operating cost per year is $B 4.0
• MSRP corresponds to actual sales price
• Offer at the same price as market leader
Simplifying Assumptions
October 3, 2002
Selected platform
vehicle design
variables
styling
i=1…N
market segments
# platforms (?)
vehicle architecture
Decision
Vector
© Massachusetts Institute of Technology
(Vehicle Design & Optimization)
Individual Vehicle
VARIANT LEVEL
(Platform Architecting)
FAMILY LEVEL
Vehicle Portfolio
Performance
Perceived Value
Cost
Relative Market
Position
Σi
Market share
Variable Cost
Profit
Objective
Vector
Two Level Approach
October 3, 2002
N=7
Sports
Sedans
Utility
VAN
TRCK
© Massachusetts Institute of Technology
7 Variants, but how many platforms?
SPTR
SUV
LOWC
MDSD
LXSD
Vehicle Segments
ED
ED
HT
October 3, 2002
WB
T
SF ]
WT
© Massachusetts Institute of Technology
Simplified BOF
Vehicle Architecture
engine
platform
r
x = [WB WT
body
SF
(phenotype)
Vehicle Architecture
HT
[in]
ED
2990
[ccm]
j
decode
1400 1.0 ]
Body
SF ]T
1.0 ]T
[-]
binary
October 3, 2002
© Massachusetts Institute of Technology
DV’’=[ 0 1 1 | 1 1 1 | 1 0 1 0 0 1 1 1 | 0 1 1 1 0 1 0 1 ]
encode
DV’ = [ k
HT
58
[in]
MDV(j) =[ ED ]
Platform Engine
DV=[ WB WT
DV=[ 108.2 61.3
[in]
PDV(k) =[ WB WT]T
Example:
Units
(genotype = vehicle “DNA”)
Vehicle Design Vector
0.15
0.1
0.2
0.4
0.15
0.15
0.15
0.05
0.45
0.2
0.4
0.3
0.05
0.2
0.05
0.1
0.25
0.05
0.3
0.3
Preference weight matrix
0.1
0.1
0.4
0.3
0.1
October 3, 2002
i =1
∑ wi, j ⋅
JˆML ,i
Ji
MSRPj
Relative
Price Prel , j = MSRP
ML , j
© Massachusetts Institute of Technology
Relative
J rel , j =
Performance
5
i =1
= 1.0
0.05
0.1
0.05
0.4
0.4
i, j
∑w
5
0.15
0.35
0.10
0.05
0.35
Perceived value = Aggregate performance relative
to the market segment leader
*TBD
AC - Acceleration
HP - HorsePower
FC - Fuel Efficiency
PV - Passenger Vol.
CV - Cargo Volume
SR - Styling Rating*
LOWC MDSD LXSD SPTR SUV TRCK VAN
Vehicle Price & Performance
$0
$5,000
$10,000
$15,000
$20,000
$25,000
0
100000
300000
Sales Volume
200000
Compact Cars -Sales vs MSRP
October 3, 2002
Source: AutoPro
400000
0
Honda
Civic
$10,000
$20,000
$30,000
$40,000
$50,000
$60,000
$70,000
$80,000
100000
300000
400000
500000
Ford Explorer
Sales Volume
200000
Sales Volume vs MSRP - SUVs
Sports Utility Vehicles - SUV
© Massachusetts Institute of Technology
Who are the leaders?
MSRP
Compact Cars - LOWC
Example Segments
MSRP
October 3, 2002
Relative Price
$0.00
$0.50
$1.00
$1.50
$2.00
$2.50
$3.00
$3.50
$4.00
0.800
1.100
1.200
Relative Performance
1.000
1.300
© Massachusetts Institute of Technology
0.900
1.500
Suggests two
“sub-segments”
1.400
Relative Position w.r.t. Leader - SUV
“Sweet Spot” Hypothesis
$0.60
$0.70
$0.80
$0.90
$1.00
$1.10
$1.20
$1.30
$1.40
$1.50
$1.60
October 3, 2002
Relative Price
0.800
1.000
J rel ,i
Prel ,i
rel ,i
− 1) + ( Prel ,i − 1)
© Massachusetts Institute of Technology
1.200
I - Contender
(J
1.100
⋅
2
II - Overscoped
Relative Performance
0.900
III - Underscoped
Dw =
III - Noncompetitive
Relative Position w.r.t Leader - LOWC
Competitive Position
2
0
0.00
50000
100000
150000
200000
250000
300000
350000
400000
450000
0.40
0.60
0.80
1.00
w ,i
(β D
α
+ 1)
© Massachusetts Institute of Technology
Establishes an estimate for sales volume
penalty due to platforming/commonality effects
Weighted Distance from Leader Dw
0.20
SVi = f ( Dw,i ) =
Sales Volume Sensitivity - MDSD
Sales Volume Sensitivity
October 3, 2002
Sales Volume
100
110
120
Wheelbase [in]
SUV
130
26
28
10
2000 2500 3000 3500 4000 4500 5000 5500 6000
Curb Weight [lbs]
12
14
16
18
20
22
24
SUV
© Massachusetts Institute of Technology
Instead of detailed CAD/CAE-simulation model,
use approximate scaling relationships for each segment.
2000
90
2500
3000
3500
4000
4500
5000
5500
6000
October 3, 2002
Curb Weight [lbs]
Engineering Model
Fuel Economy [mpg]
October 3, 2002
WB
WT
HP
HT
input
103.1
57.9
115
55.1
FC
PV
CV
CW
FC
PV
CV
CW
output
33.85824
87.707918
11.86767
2233.1235
© Massachusetts Institute of Technology
truth
36
85.9
12.9
2405
- Multivariable function approximator
- Trained with market segment data
WB
WT
HP
HT
FC
PV
CV
CW
error %
6.3257
2.0613
8.6987
7.6967
Neural Network Model
October 3, 2002
C fam =
i =1
∑
# platform
45%
- x%
∑
i =1
Body
30%
i =1
∑
# bodies
TFU b ,i N bB,i
B = 1 − ln(100 / S ) ln 2
© Massachusetts Institute of Technology
TFU p ,i N pB,i +
# engines
L
B
C = TFU ⋅ {
N
Engine
25%
Vehicle
Cost
TFU e ,i N eB,i +
Include Learning
Curve Effect
Platform
Total Product Family Cost:
Margins x:
LOWC 5%
MDSD 10%
LXSD 20%
SPTR 15%
SUV
15%
Truck 25%
Van
15%
Market Leader
MSRP
Cost Model
October 3, 2002
# of platforms
platform
nnet
opt 2
cost
© Massachusetts Institute of Technology
i=1,..,7
opt 1
mkt
Vehicle Level
Family Level
Simulation Framework
portfolio
profit
0
0.2
1
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
1
2
October 3, 2002
3
4
5
Number of Platforms
6
© Massachusetts Institute of Technology
Non-competitiveness
due to uniformity
“Optimal”
Platform
Strategy
2
7
7
cost of
variety
Null
Platform
Strategy
Optimal # of Platforms
Product Family Profit [B$]
1
October 3, 2002
0
0.02
0.04
0.06
0.08
0.1
0.12
2
3
4
# of platforms
6
7
© Massachusetts Institute of Technology
5
LOWC
MDSD
LXSD
SPTR
SUV
TRCK
VAN
105.1
105.1
114.8
105.1
114.8
137.0
114.8
WB
α β γ
Optimal - 3 Platforms
- results sensitive to learning curve S
Some observations
Performance
penalty goes down
with increasing
# of platforms
Weighted Avg Distance
October 3, 2002
α
β
χ
δ
ε
φ
γ
1
2
3
4
4
4
4
Sedans
LOWC
MDSD
LXSD
1
2
3
3
3
3
3
1
2
2
2
2
2
7
1
2
2
2
2
2
2
1
1
1
1
1
1
1
1
2
2
2
5
5
5
Utility
VAN
TRCK
Can see the
resulting strategy
as more platforms
are added.
Horizontal Levering
Beachhead strategy
© Massachusetts Institute of Technology
1
2
3
4
4
6
6
Sports
SPTR
SUV
Resulting Strategy
2200
10
2300
2400
2500
2600
2700
2800
2900
3000
Can one reverse engineer
a competitors platform strategy
based on:
- grid patterns in data?
- teardown inspection?
In design space one discovers
grid patterns for some design
variables … indicates commonality
Future Work
© Massachusetts Institute of Technology
Attempt to apply methodology to GM vehicle
family ( > 80 vehicles, >20 platforms)
15
20
Fuel Capacity (gal)
GM Vehicle Family
October 3, 2002
Wheelbase [mm]
October 3, 2002
© Massachusetts Institute of Technology
• Include components in platform that are:
- hidden from the customer
- insensitive to performance
- high value
- technologically stable
• Different strategies exist, must be carefully chosen
• Platform Architecture important for cost effectively
creating product variety with commonality savings
Summary