six sigma
TRANSCRIPT
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Module 1.1Introduction to Design for Six Sigma
Lear Corporation Confidential/Proprietary
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 2
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Lean Manufacturing and Assembly (DFMA)Design for Lean Manufacturing and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Introductions - Why DFSS?
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 4
Lear’s DFSSDFSS deployment is based on an EngineeringEngineering Leadership Leadership focused
Team TrainingTeam Training approach that will achieve improvements during Product DevelopmentProduct Development
in order to provide benefits during Product Launch.Lear Corporation Confidential/Proprietary
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 5
$ $ $ $
$ $ $ $
$ $ $ $
$ $ $ $
$ $ $ $
Product DevelopmentProduct Development
Why focus on Product Development activities?Why focus on Product Development activities?
Planning Prototype Pilot Launch Post Launch
Pgm Management
Marketing
Customer Reps
Supplier Mgmt
Quality Engineers
Manuf. Engineers
Design Engineers
Plan
Produc
t Deve
lopment
Problem
s
Result
In Los
t Reve
nue
Actuals
Lear Corporation Confidential/Proprietary
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 6
“Golden Opportunities”?
Contributors to Product Contributors to ProductCostCost
Application of Six Sigma Application of Six Sigma
Influences on Product CostInfluences on Product Cost
Synergy with Six SigmaSynergy with Six SigmaAdequate design margins
Stable parts and materials
Good process capabilities
0
10
20
30
40
50
60
70
Infl
uenc
e on
Pro
duct
Cos
t (%
)
Design Material Labor Overhead
Major Cost Contributors
Material OH Labor Engr
Product Cost Elements
50%
30%
15%5%
Follow the money:
Six Sigma
ConceptUtility
Inc
rea
sin
g E
mp
ha
sis
Program Lifecycle Phase
ConceptInitiation
Planning Prototype
Pilot
DFSSTraditionalSix Sigma
LaunchPost
Launch
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 7
What is “DFSS”?
Design for Six Sigma IsDesign for Six Sigma Is:– A philosophyphilosophy of how products should be developed– A processprocess that emphasizes designing in quality– A measuremeasure of the level of quality in a product’s performance– A methodmethod for:
analyzing and improving the robustness of a product’sperformance based on the statistical properties andcontribution of each input, component, or parameter.
Why Bother?Why Bother?– The fundamental premise is that customers care about quality
and companies care about profits.– Product Variation ⇒ Poor Quality ⇒ Higher Prices
⇒ Customer Dissatisfaction
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 8
Why Change?
Common Engineering Language.Statistical Engineering for Robust Product DesignManufacturing Process Capabilities are a Driving Force in ProductDesignPredictability of Product Performance, Manufacturability, ReliabilityReduced Number of Prototypes, ECN’s, Rework, ScrapFocus on Customer Requirements and Product Quality.Reduced Warranty Expenses.Increase Customer Satisfaction & Loyalty.Gain Market Share
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 9
We can’t get there without betterdesigns
The “Five Sigma” Wall
• 70-80% of all qualityproblems aredesigned in
• Manufacturing willnever be able tomake our currentproducts with SixSigma quality
6 Sigma
3 SigmaNew, inherently SixSigma designs are theonly way to reach the
corporate goal.
DFSS is needed
“5 Sigm a W all”
DM
AIC
DM
AIC
Design for Six Sigma
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 10
How Other Companies View DFSS
General ElectricGeneral Electric• “Every new GE product and service
will be ‘DFSS’ - Design for SixSigma. These new offerings willtruly take us to a new definition of‘World Class’.”
General Electric Company Annual Report, 1998, p 4
Fruedenberg-NOKFruedenberg-NOK• A systematic methodology, tools,
and techniques which enable us todesign products and processesthat can meet customerexpectations and can be producedat the 6 sigma level.
Drew Algase, “Successful Project Selection forDFSS Operations, IQPC DFSS Conference, Jul2001.
General Domestic Appliance Ltd.,General Domestic Appliance Ltd.,GE-MarconiGE-Marconi
• Vision of product and serviceexcellence.
• Way to manage technical risk.• Gives confidence to decision
making for leading-edge products.• Added value to our customers.• Key element for continued growth.
Phil Rowe, “Training Requirements for DFSS”,IQPCDFSS Conference, Aug 2000.
General MotorsGeneral Motors• "It is expected that our top Interior
Suppliers will be ready to supportthe GM DFSS Initiative for allInterior Integration programsbeginning in 2002!"
Tom Lewandowski,General Motors PurchasingMarch, 2001
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 11
What have other Companiesachieved with DFSS
Source: IQPC DFSS Conference Materials, 2000-2001.
General Domestic Appliance Ltd.General Domestic Appliance Ltd.• Revenue growth: volume, market
share, price.• Warranty and manufacturing cost
reductions.• Customers who are delighted with
their products.
A Driver for GrowthA Driver for Growth
Texas Instruments DSEGTexas Instruments DSEG• Capability & Statistical Analyses
resulted in Cost Avoidance• Pgm A - $2K per system• Pgm B - $16.4K per system• Pgm C - $1000 per lot
Lower Production CostsLower Production CostsGeneral Electric Medical SystemsGeneral Electric Medical Systems• Light-Speed Scanner System
• Chest scan from 180 to 17seconds
• Lower cost/scan• $69 million in orders in first
90 days
Customer SatisfactionCustomer Satisfaction
iomegaiomega• Reduced Development Cycle
Time from 12 to 3 months.• Reduced Critical Engineering
Changes from 37 days to 2 days• Reduced Tooling Lead Time from
12 to 6 weeks.• Reduced Component Lead Time
from 36 to 12 weeks.
Improved Time to MarketImproved Time to Market
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 12
What is DFSS for Lear?
DFSS….DFSS….
Is a standardized and regimented product developmentprocess that integrates with LPMPintegrates with LPMP
Incorporates process capabilityprocess capability, consumer inputconsumer input and customercustomerrequirementsrequirements into the design process
Provides statistical toolstools to analyze and optimize productdesigns
Is driven by the Engineering organization and requires fullfullproduct teamproduct team involvement
Is geared to Product Development, notnot Process Improvementon existing products (DMAIC)
Lear Corporation Confidential/Proprietary
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 13
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
DFSS Overview
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 15
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 16
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 17
The Language of DFSS
Fundamental metricFundamental metric is the probability of notmeeting a requirement:
the Probability of Non-ComplianceProbability of Non-Compliance ( (PNCPNC))
ComputeCompute PNCPNC by applying statistics to existingengineering analyses . . . during design!
“Non-compliant”
USLLSL T
“Non-compliant”
1 Sigma ⇒ PNC = 0.317
2 Sigma ⇒ PNC = 0.046
3 Sigma ⇒ PNC = 0.0027
4 Sigma ⇒ PNC = 6.33 ×10-5
5 Sigma ⇒ PNC = 5.74 ×10-7
6 Sigma ⇒ PNC = 1.78 ×10-9 (1.5σ Mean Shift not included)
“Compliant”
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 18
Fundamentals of DFSS
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
Y
C
E
A
B
D
ULLL T
PNCPNC
UnderstandingUnderstandingRequirements,Requirements,Specifications,Specifications,& Capabilities& Capabilities
ApplyingApplyingModels &Models &AnalysesAnalyses
PredictingPredictingProbability of Probability of
Non-ComplianceNon-Compliance
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 19
Tools for DFSS
Inventive Principle 1
Example 1
Example k
Definition j
Example 1
Example k
Inventive Principle i
Definition 1
Example 1
Example k
Definition j
Example 1
Example k
Flowdown forInventive Principles,Definitions andExamples
Contradiction
Definition 1
TRIZTRIZ
SensitivitySensitivityAnalysisAnalysis
x4 2 0 2 40.20.4
x0 1 2 3
0.5
1
x0 2 4 6 8
0.5
42 44 46 48 50 52
0.1
0.2
Monte Carlo AnalysisMonte Carlo Analysis
AllocationAllocation
High PerformanceHigh PerformanceLow CostLow Cost BalancedBalanced
Plus Outlet Air Req’tPlus Outlet Air Req’t
OptimizationOptimization
It’s all about ...It’s all about ...It’s all about ...LEVEL
1LEVEL
2
LEVEL3
QFDQFD
ScorecardScorecardEvents Good Units Bad Units Total
Unit Fails Test 0.01998 0.00099 0.02097
Unit Passes Test 0.97902 0.00001 0.97903
Total 0.99900 0.001 1.0000
… designing products that work as planned!… designing products that work as planned!… designing products that work as planned!
Test TestEffectivenessEffectiveness
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 20
Module 3.3DF Lean M/A (4)
Module 1.1Intro to DFSS (6)
Module 2.1QFD, Inventive Thinking, & Design Trade-off (6)
Module 3.4DFMEA (4)
Module 3.2Reliability, Gage R &
R, & TestEffectiveness (6)
Module 4.3Product Design Variation (4)
Module 4.4Design Optimization (4)
Module 4.2Modeling (TWO DAYS:4hrs then 6 hrs)
Module 4.1Excel / MiniTab Primer (optional) (4)
(Can be taken in any order)
( Shown in recommended order)
I
II
III
IV
Information coveredin Key Stake Holder
Level Course Flow
Module 3.1Basic Statistics(4)
Modular Training
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 21
Module 1.1 - Introduction toDesign for Six Sigma (DFSS)(6 Hours)– Why DFSS– DFSS Overview– Quality Function Deployment– Trade-Offs (Pugh, & SDI
Analysis)– SPC– Program Management– Benchmarking– FMEA– DFMA– Design for Reliability– Regression & DOE– Other DFSS Tools– Score Card
Module 2.1 - QFD, InventiveThinking & Design Trade-Off
(6 Hrs)– QFD– Examples and Exercises– Benchmarking– Pugh Techniques– TRIZ
Module 3.1 - Basic Statistics(4 Hrs)- Basic statistics- Probability, Sigma and PNC- Process capability- Scorecard
DFSS Training Modules
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 22
Module 3.2 - Reliability, Gage R& R, and Test Effectiveness
(6 Hrs)- Defining Reliability & Failures- Reliability Models- Weibull analysis- Gage R&R + TestingEffectiveness
Module 3.3 - Design for LearnManufacturing and Assembly(DFM/A) (4 Hrs)- Lean Principles- DFM/A Definitions- General Principles and Guidelines- Examples and Case Studies
Module 3.4 - Design FailureMode and Analysis (DFMEA)
(4 Hrs)– FMEA Introduction / Types of
FMEA– What and Why DFMEA– DFMEA Teams and Facilitation– Defining the Design– DFMEA Creation– Taking Action
Module 4.1 - Excel/MinitabPrimer (4 Hrs)– Minitab environment– Excel equations and linking– Discrete Modeling
DFSS Training Modules
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 23
Module 4.2 - Modeling(Day 1: 4 Hrs + Day 2: 6 Hrs)
– Fundamentals and principles– Analytical Equations– Regression– DOE
Module 4.3 - Product DesignVariation (4 Hrs)– Sensitivity– Monte Carlo Analysis– Allocation Analysis
Module 4.4 -DesignOptimization (4 Hrs)– What is Optimization– Single Objective– Multi-Objective
DFSS Training Modules
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 24
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Quality Function Deployment
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 26
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 27
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 28
Quality Function Deployment:Learning Objectives
At the end of this module, participants will understand the use of . . .
QFDQFD and Voice of the CustomerVoice of the Customer concepts that will be usedby project teams in the requirements development process
House of QualityHouse of Quality for requirements identification and flowdown
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 29
SalesQuoted
EngineeringSpecified
Design Modeled Plant Produced Repaired at Dealer
Customerwanted
Communication Breakdown
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 30
The House of Qualityprovides ...
… documentation of how the problem is viewed … fromdifferent perspectives
CustomerCustomerNeedNeed
DesignDesignRequirementsRequirements
CompetitiveCompetitiveAssessmentAssessment CorporateCorporate
CompetenciesCompetenciesandand
PlanningPlanning
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 31
Quality Function Deployment
Quality Function Deployment (QFD):Quality Function Deployment (QFD): a planningtool for translating the Voice of the Customer(VOC) into explicit design, production, andmanufacturing process requirements.
The QFD House of Quality transfers customers’needs to requirements based on the strengthof the inter-relationships.
HOW Vs. HOW
Relationships
HOW MUCH
HOW
WH
AT
House ofQuality
Requirements,SpecificationsProcesses
HOW
WHAT
QFDQFD
Fuzzy
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 32
Fewer Design Changes
21 12 33 ProductionbeginsMonths
Num
ber
of D
esig
n C
hang
es
90% of Totalchangescomplete
Company Using QFD
Company not using QFD
Adapted from L.P Sullivan. “Quality Function Deployment.” Quality Progress, June 1986
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 33
The House of Qualitycontains Primary andSecondary elements.Primary elements arerequired to determineCTCs.
Secondary elementsare informational.
House of Quality
CustomerRequirements
(WHATS)
RelationshipMatrix
PerformanceCharacteristics
(HOWS)
InteractionMatrix
Technical Weights
Target Values
CompetitiveBenchmark
CompetitiveBenchmark
Impo
rtan
ce
Com
plet
enes
s
Target Direction
Primary Element
Secondary Element
Req
. P
lann
ing
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 34
Customer Need &Design Requirements
95
24
76
63
Faster 5
Lighter 2
+ Operating Time 4
Voice Recognition 3
H
H
L
H
H
H
M
L
M H
H
72 41 18 64 27 3610 GHz 75 Wh 0.5 lb
0.01 C/W
1 kHz 5 W
CP
U S
pee
d
Lo
ng
-Lif
eB
atte
ryL
igh
twei
gh
tC
ase
Hea
t P
ipe
Co
olin
gL
ow
-Pas
sF
ilter
Lo
w-P
ow
erL
CD
Dis
pla
y
/\ /\ \/ \/ O \/
-- +
++ -++ +
1. What the customer wants1. What the customer wants
2. Ranking of customer needs2. Ranking of customer needs
3. Quality Characteristics (Hows)are quantitative and explicit
3. Quality Characteristics (Hows)are quantitative and explicit
4. Typical numerical valuesfor the relationships are:High H 9Medium M 3Low L 1
4. Typical numerical valuesfor the relationships are:High H 9Medium M 3Low L 1 5. Technical Weight5. Technical Weight
6. Completeness6. Completeness
7. Target Direction7. Target Direction
8. Interaction Matrix8. Interaction Matrix
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 35
Quality Function Flow
ProductionRequirements
Key
Pro
cess
Op
erat
ion
s
PRODUCTIONPRODUCTIONPLANNINGPLANNING
Key ProcessOperations
Par
t Q
ua
lity
Ch
arac
teri
stic
s
PROCESSPROCESSPLANNINGPLANNING
Part QualityCharacteristics
En
gin
eer
ing
Ch
arac
teri
stic
s
PARTSPARTSDEPLOYMENTDEPLOYMENT
Cu
sto
me
rA
ttri
bu
tes
EngineeringCharacteristics
ENGINEERINGENGINEERINGREQUIREM’TSREQUIREM’TS
Source : Hauser, J. R. and D. Clausing, “The House of Quality,”Harvard Business Review, May-June 1988
VOCVOC
Customer ExpectationsCustomer ExpectationsSolid
4 mm maxmovement Diameter
+/- .1mmCNC
MachiningSPC on
DiameterEverything relates back
to Voice Of the Customer
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 36
House of Quality:Tool for Implementation
Triptych - a toolsetthat includes a QFDimplementation tool
Level 1
Level 2
Level 3
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 37
Voice of the Customer
CollectCollect as much information as possible before starting toconstruct a QFD
Sources of Information for the Product TeamSources of Information for the Product Team– Requirement Documents and Specifications– Complaints– Recommendations– Data– Internal Information– External Information– Market Research
Customer interviews, questionnaires, surveys, test reports,competitive analyses
– Historical Warranty and Quality Information– Team Experience
One House, One day
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 38
Benefits of QFD Process
Dispels the “business as usual” approach to planning anddesign
Revises operational norms– Directors and managers become “team members”
– Functional managers are required to donate the services ofpersonnel
– Decision authority is with teams and not individuals
– Lines of communication are established between functions at theworking level
Provides a structured approachstructured approach for identifying anddocumentingdocumenting product requirements– ‘What’ is required
– ‘How’ can requirements be achieved
– Key relationships and comparative weightings
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 39
Quality FunctionDeployment Summary
QFD is a planning tool for translating the Voice of theCustomer (VOC) into explicit design, production, andmanufacturing process requirements.
QFD is a tool and process for listening to customers. QFD isnot a decision maker.
Each House of Quality is used to identify key CTC’s thatmaximize the chance of meeting customer needs at that level.
Information is used define the product design approach andproject development activities.
Team discussions and planning sessions are used to capturethe information.
Triptych QFD tool facilitates the process.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 40
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Benchmarking
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 42
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 43
Learning Objectives
At the end of this module, participants will understand the use of . . .
What is Benchmarking?Benchmarking?
DFSS Benchmarking ProcessDFSS Benchmarking Process
–– Activity-Type BenchmarkingActivity-Type Benchmarking
–– Gap Reduction BenchmarkingGap Reduction Benchmarking
BenchmarkingBenchmarking in the Critical-To-Customer (CTC) technicalrequirements development process
Phased Benchmarking approachPhased Benchmarking approach for collecting, prioritizing,and improving key CTC requirements
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 44
What is Benchmarking?
Benchmarking -Benchmarking -– performance measuring tool– can be used to measure comparative performance– aids in identifying “best-in-class” practices– supports achievement of major improvements
Benchmarking creates value -Benchmarking creates value -– through improvements in performance, reliability, cost and
revenues generated in product developments efforts: - Provides a way to improve customer satisfaction - Identifies key Critical-to-Customer (CTC) performance gaps - Assists in eliminating the “not-invented-here” syndrome - Establishes the state-of-the-art performance in the market place - Allows for making better business decisions based on a larger data base of customer-driven inputs
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 45
Benchmarking Types
VisionWorks Benchmarking
– Corporate research group that supports and provides services forLear globally
Project Team Benchmarking
– Activity-Type
For understanding the details of the customer needs and wants andfor developing QFDs for CTC Technical Weight information
– Gap Reduction
For developing and executing a gap reduction plan based on identifiedCTC Technical Weights, gap analysis, and project goals
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 46
Activity-TypeBenchmarking Steps
Step 1Step 1
– Use the Quality Functional Deployment (QFD) tool todevelop and prioritize Critical-to-Customer (CTC) technicalweights
Step 2Step 2
– Develop and prioritize key competitors CTC technicalweights (TW) using existing knowledge and same QFDtemplate
Step 3Step 3
– Use TRIZ/Design Trade-off tools to develop new designconcepts and to prioritize new design concepts CTC TWs
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 47
Gap-ReductionBenchmarking Steps
Step 1: PlanStep 1: Plan– Identify which CTCs will be
benchmarked– Select product team members
to conduct benchmarking– Determine companies to be
benchmarked
Step 2: PerformStep 2: Perform– Establish benchmarking
agreements with candidatecompanies/organizations insame or other industries.
– Determine key personnel pointsof contact.
– Organize team and conductBenchmarking session usingappropriate means ofcommunication.
Step 3: AnalyzeStep 3: Analyze– Organize collected data– Conduct QFD session using
collected data.– Perform CTC gap analysis
Step 4: DevelopStep 4: Develop– Develop gap-reduction
implementation plan– Communicate plan with
team/management– Obtain agreement for ‘‘go
ahead”
Step 5: ImplementStep 5: Implement– Execute plan– Update progress with reviews– Finalize benchmarking
documentation
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 48
Full Interior and Component Teardown
– Feature, Component and Assembly Analysis
– Customized study per each customer request
– ‘A’ side feature & data measurements (non-destructive)
– ‘B’ side data measurements
With components disassembled from the vehicle.
– Digital photos of components
In-vehicle condition
‘A’ & ‘B’ side views of each component disassembled from the
vehicle
VisionWorks Services
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 49
Benchmarkingdefined in LPMP
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 50
Lear Benchmarking Info
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 51
Learnet Benchmarking info
Pending
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 52
Learnet Benchmarking info
Pending
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 53
VisionWorksBenchmarking Request
Have a list of deliverables and / or a project scope to review with the VW group
List of deliverables should consist ofTarget vehiclesTarget market segmentsOEM recommended vehiclesList of components targetedAny specific issues targeted per vehicle or component (“the why’s”)Program timing dates
Benchmarking study completion timing requirements
Benchmarking output/results format requirementsPresentationHard copyBurned copy on CDElectronic copy
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 54
Benchmarking Summary
BenchmarkingBenchmarking– CTCs can be used to identify need for Benchmarking.– QFD tool supports creation of gap analysisgap analysis
with respect to design goals/targetswith respect to competition
– DFSS benchmarking is cost effective and supports productdevelopment activities.
– Gap-reduction implementation is a technique for makingsignificant customer-focusedcustomer-focused improvements
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 55
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
TRIZ andTrade-Offs
• Theory of Inventive ProblemSolving
• Pugh Analysis
• SDI Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 57
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 58
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 59
TRIZ and Trade-Off:Learning Objectives
At the end of this module, participants will understand the use of . . .
MethodsMethods such as TRIZ to promote inventive problem solving
TRIZ methodologyTRIZ methodology as a tool for project teams to solveconflicting requirements
Design Trade-OffDesign Trade-Off approaches to select the best idea orapproach
QualitativeQualitative and QuantitativeQuantitative approaches to design trade-off
Each method and when it should be applied.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 60
Conflicting Requirements
QFDQFD translates the voice of the customer into engineeringrequirements.
Engineering requirements are often conflicting.– Example: The product gets stronger, but the weight increases.
Generating solutions that satisfy conflicting requirements isusually performed using various methods such as:– Free association brainstorming
– Cross fertilization (drawing analogies from other disciplines)
– Futuring (thinking to the future when these constraints may nolonger exist)
The above methods occasionally generate a solution toconflicting requirements, but is there a reliable method forgenerating solutions? YESYES
TRIZTRIZ is a method for generating Inventive Solutions toconflicting engineering requirements that is based on patternsof innovation identified in patents.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 61
Methodology for SolvingTechnical Contradictions
Methodology for resolving Technical Contradictions:1. Identify the Technical Contradiction.2. Determine which Feature in the contradiction is to be Improved
and which one is Degraded (or worsened).3. Choose the Design Parameters that most closely match the
Improving Feature and Degraded Feature in the contradiction.4. Examine the proposed Inventive Principles.5. Select the best Inventive Principle.6. Apply the Inventive Principle to the Technical Contradiction.
SDI’s TRIZ tool (part of the Triptych tool suite) performsSteps 3 and 4 by providing a simple interface to thecontradictions matrix with 130 definitions and 167 examples.
IdentifyContradiction
Choose Best TRIZ General Solution
Translate to TRIZ General Contradiction
Apply Solution to Problem
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 62
TRIZ:Tool for Implementation
Select ImprovingFeature from the list of39 Design Parameters
Idea Button is usedto store ideas on aTRIZ worksheet
Select DegradedFeature from the list of39 Design Parameters
Up to four InventivePrinciples are listed here,along with their rank (interms of frequency of use)
One or more definitionsof the selected InventivePrinciple are shown
One or more examplesof the selected InventivePrinciple definition areshown
Arrows cycle throughdefinitions andexamples
Save Button is usedto store selectedFeatures, InventivePrinciple, Definition,and Example on aTRIZ worksheet
Definition Buttonshows definitions forselected DesignParameters
Done Button closesTRIZ Tool
Help Button launchesTRIZ Tool Help
Enter your Idea here.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Trade-Off
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 64
Multiple Approacheshave been identified …
If multiple approaches or design alternatives are identified …how do you determine the ‘best’ choice?
Moving a Heavy ObjectPower Assist
Vs...Composite Materials
Vs..Balanced Center of Gravity
Moving a Heavy ObjectPower Assist
Vs...Composite Materials
Vs..Balanced Center of Gravity
Electrical Part CoolingAluminum Heat Sink
Vs..Copper Heat Sink
Vs..Cool Plate
Vs..Heat Pipe
Electrical Part CoolingAluminum Heat Sink
Vs..Copper Heat Sink
Vs..Cool Plate
Vs..Heat PipeSeat Recliner Mechanism
Pawl & Sector, Single SidedVs..
Pawl & Sector, Dual SidedVs..
Integrated AdjusterVs..
Cam Adjuster
Seat Recliner MechanismPawl & Sector, Single Sided
Vs..Pawl & Sector, Dual Sided
Vs..Integrated Adjuster
Vs..Cam Adjuster
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 65
Introduction toDesign Trade-Off
Design Trade-OffDesign Trade-Off is a decision-making process that is usedto select a design (among many alternate designs) that mostclosely meets multiple, conflicting CTCs.There are many approaches to performing Design Trade Off,but all have the following characteristics:– Gather Design Options
brainstorming, TRIZ, patent searches– Gather Design Criteria (CTCs)
benchmarking, QFD, surveys– Compare each Design Option against each Design Criterion
qualitative or quantitative scoring– Compare each Design Option against each other
qualitative or quantitative scoring
Triptych uses two approaches:–– Pugh MatrixPugh Matrix for preliminary design phase (Qualitative)–– SDI Trade-Off ToolSDI Trade-Off Tool for detailed design (Quantitative)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 66
Pugh Trade-Off Matrix(Qualitative)
Pugh Matrix is used to make fast qualitative comparisons whenDesign Options and Design Criteria are known.
1. Create matrix with Design Optionsand Criteria
2. Assign Importance to Criteria3. Choose Datum4. Compare Options with Datum
Better (+)Same (S)Worse (-)
5. Add the Weighted Sumsfor Better, Same, andWorse
6. Choose the new datum (thebest design Option for thecurrent iteration cycle)
7. Repeat until best conceptis found.
Weighted Sum of +
Weighted Sum of S
Weighted Sum of -
Cost 2
Heat Transfer 3
Complexity 1
Junction Temp. 2
-
S
S
+
-
+
S
+
S
+
S
+
-
+
-
+
2 5 5 5
4 1 3 0
2 2 0 3
Alu
min
um
Hea
t S
ink
Co
pp
erH
eat
Sin
k
Co
ld P
late
Hea
t P
ipe
Sp
ray
Co
olin
g
Design Options
Crit
eria
Impo
rtan
ce
1st Datum
Next Datum
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 67
SDI Trade-Off Matrix(Quantitative)
SDI’s approach to design trade-off is to quantitatively evaluatehow well the Design Option meets each Design Criterion.Specification limits (LSL and/or USL) may be specified alongwith a Target for each Design Criterion.Design Option Scores are calculated for each CTC:Linear variation inside the specificationQuadratic variation inside the specificationInside/Outside of specificationMaximizeMinimize
LSL Target USL
LSL Target USL
LSL Target USL
Min Max
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 68
TRIZ and Design Trade Off:Summary
TRIZ - aids in generation of alternativesTRIZ - aids in generation of alternatives– The key concept behind TRIZ is that there is a 95% probability
that someone else, in some other industry has solved the samefundamental technical contradiction.
– TRIZ is a method for creating innovative solutions to engineeringcontradictions.
– TRIZ can be used to generate new ideas.
Design Trade-Off - aids in comparison of alternativesDesign Trade-Off - aids in comparison of alternatives– Design Trade-Off is used to selected Design Options that best
satisfy multiple Design Criteria.
– Pugh Matrix is a qualitative comparison of Design Options againstmultiple Design Criteria
Can be used at any part of the design cycle.
– SDI Trade-Off is a quantitative comparison of Design Optionsagainst multiple Design Criteria
Requires detailed knowledge of the Design Criteria (CTCs).
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 69
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Lear’s Program Management Process(LPMP)
• Introduction
• Scope
• Phases
• Deliverables
• Integration of DFSS into LPMP
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 71
LPMP
Lear Program Management Processalso
The Process for Product Developmentalso
The Process for Product Launchcould be called
The Cross Functional Team Process
Does Not Work Withoutthe Cross Functional TEAM
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 72
Introduction
The Lear Program Management Process is intended toprovide a structured yet flexible methodologyprovide a structured yet flexible methodology to manageproduct development from pre-award of business throughlaunch of production.
The Lear Program Management Process is designed to give athorough overview and template for developing, executingdeveloping, executingand managingand managing a program from early conception throughprogram launch.
This process serves as a planning and managementplanning and management tool toensure that specific activitiesspecific activities occur at the proper time inthe product development process to create products, includingsupportive services that meet and exceed (internal andexternal) customer expectations.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 73
Scope
It provides common strategy, framework, and controls to beutilized through-out Lear Corporation.
Execution of the process is led by a program manager butrequires significant support requires significant support from a cross-functional cross-functionalprogram team.program team.
LPMP is the backbone structurebackbone structure for key deliverables withinDFSS.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 74
LPMP on the LearNet
Click Here
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 75
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 76
Phases
There are five phases in the LPMP process: Planning,Planning,Prototype, Pilot, Launch, and Post Launch.Prototype, Pilot, Launch, and Post Launch.
– For each phase there is a phase overview on the web site thatprovides a descriptive summary of the scope of activity within therespective phase and highlights key elements of the phase.key elements of the phase.
Each phase overview has a link to a relationship diagramrelationship diagram thatshows the general order and relationship of all therelationship of all thedeliverablesdeliverables within the respective phase.
– The relationship diagramsrelationship diagrams are provided as a tool to help teamsbetter understand how deliverables are related to one another.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 77
LPMP Deliverables
Each phase of LPMP has an associated set of deliverablesset of deliverablesidentified.identified.
For each deliverable there is an associated deliverableinformation sheet that provides the following information:
– Definition: A concise description of the task
– Clarification / Training Point: An expanded explanationexpanded explanation of thedeliverable. May provide more detail on what should or shouldnot be considered in accomplishing a deliverable.
– LPMP Phase: Listing of phases that the respective deliverable isaccomplished.
– Deliverables: A clear statement of the expected outputexpected outputproduced in accomplishing the task.produced in accomplishing the task.
In some cases there is no output such as when the deliverable is amilestone.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 78
LPMP Deliverables (Cont’d)
– Requirements: This section will identify any related/requiredidentify any related/requiredforms or proceduresforms or procedures that are considered mandatory within thecorporation or specified segment of the corporation.
– Best Practices: Best Practices will often provide documents,provide documents,spreadsheets, and formsspreadsheets, and forms that various teams have successfullyused on their program and are considered best practices.
– Reference: Where formal reference documents are available,known, and considered useful, this field will identify them.
At a minimum, there are references to the appropriate sections of theAIAG APQP manualAIAG APQP manual where applicable.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 79
Engineering Deliverables
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 80
Integration of DFSS Into LPMP
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 81
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 82
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 83
LPMP Summary
Gate / Phase SystemFlexibleAll ProductsAll CustomersSeries of ReviewsEncompasses APQP
Product Development and LaunchWeb-Based ManualIdentifies and tracks key DFSS deliverables
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 84
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Design Failure Mode and EffectsAnalysis (DFMEA)
• History
• Types of FMEA
• DFMEA Defined
• DFMEA Supports the designprocess
• Role of DFMEA
• The Form
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 86
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 87
Learning Objectives
At the end of this module, participants will understand . . .
DFMEA and it’s link to the design processDFMEA and it’s link to the design process
Sources of risk
DFMEA concepts that will be used by project teams in thedevelopment process
Why we use DFMEA, and when to use it.Why we use DFMEA, and when to use it.
The DFMEA form.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 88
History
First used in the 1960’s in the Aerospace industry during theApollo missionsIn 1974 the Navy developed MIL-STD-1629 regarding the useof FMEAIn the early to mid 1970’s, automotive applications driven byliability costsCurrent Standard is SAE J-1739
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 89
Where Does Risk Come From?
VagueWorkmanship
Standards
Cumulative Risk
Poor controlplans & SOP’s
Raw MaterialVariation
Poorly developedSpecification
LimitsMeasurement
Variation(Online and QC)
MachineReliability
PotentialSafety
HazardsUnclear CustomerExpectations
D. H. Stamatis, FMEA:FMEA from Theory to Practice, Quality Press, 1995
Poor ProcessCapability
Job Assignment
Variation
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 90
Type of FMEA’s
System:– Analyzes systems and sub-systems in the early concept and
design stages. Focuses on functions and interactions amongsystems.
Design:– Analyzes product designs before they are released to production.– A DFMEA should always be completed well in advance of a
prototype build.– Focuses on potential failure modes of products due to design
deficiencies or errors.Process:– Analyzes production or administrative processes.– Focuses on potential failure modes of the output caused by
process deficiencies.Machine:– Analyzes a piece of manufacturing equipment prior to its
construction.– Focuses on potential failure modes of the manufacturing
equipment due to design deficiencies or errors.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 91
Type of FMEA’s
Equipment:– Analyzes production or administrative process.– Focuses on potential failure modes of the output caused by
process deficiencies of the manufacturing equipment only.
Change:– Analyzes a design or process change not covered in existing
FMEAs production.
Containment:– Analyzes a containment screen.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 92
DFMEA Defined
A Design FMEA is an analytical technique utilizedanalytical technique utilized primarilyby a Design Responsible Engineer/Team as a means to assurethat, potential failure modes and their associatedcauses/mechanisms have been considered and addressed.
An DFMEA is a summary of an engineer’s and the team’sis a summary of an engineer’s and the team’sthoughtsthoughts (including an analysis of items that could go wrongbased on experience and past concerns) as a component,subsystem or system is designed.
This systematic approach parallels, formalizes and documentsthe mental disciplines that an engineer normally goes throughin any design process.
Source: FMEA manual; Automotive Industry Action Group (AIAG)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 93
DFMEA Supports the DesignProcess
The Design FMEA supports the design processsupports the design process in reducingthe risk failures by:– Aiding in the objective evaluation of design requirementsevaluation of design requirements and
design alternatives.– Aiding the initial design for manufacturing and assemblymanufacturing and assembly
requirementsrequirements–– Increasing the probabilityIncreasing the probability that potential failure modes and
their effects on system and vehicle operation have beenconsidered in the design/development process.
–– Providing potential informationProviding potential information to aid in the planning ofthorough and efficient design test and development programs
– Developing a list of potential failure modespotential failure modes ranked accordingto their effect on the “customer”, thus establishing a prioritysystem for design improvements and development testing.
–– Providing an open issueProviding an open issue format to recommending and trackingrisk reducing actions.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 94
Role of Design FMEA
Key tool for the design team to improve the designimprove the design in a
preemptivepreemptive manner (before failures occur)
Used to prioritize resourcesprioritize resources to insure design improvement
efforts are beneficial to customer
Used to document completiondocument completion of projects
Should be a dynamicdynamic document, continually reviewed,
amended, updated
Used to analyze new design conceptsanalyze new design concepts
EvaluatesEvaluates the risk of design changes
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 95
DFMEA Team
Team approach is necessaryResponsible Engineer typically leads the teamDocument should be owned by the process ownerRecommended representatives:– Design– Manufacturing operators / supervisors– Quality– Reliability– Materials– Testing– Supplier
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 96
The Form
POTENTIALFAILURE MODE AND EFFECTS ANALYSIS
Print # NUMBER Rev. ECL (DESIGN FMEA) FMEA Number: FILE.XLS
System/Subsystem/Component: Design Responsibility: Prepared by:
Model Year(s)/Vehicle(s) APPLICATION Key Date Page: of
Team: FMEA Date (Orig.) (Rev.)
C Potential O Current Current DItem Potential Potential S l Cause(s)/ c Design Design e R. Recommended Responsibility Action Results
Failure Effect(s) of e a Mechanism(s) c Controls Controls t P. Actions & Target Actions S O D R.Mode Failure v s of Failure u - Prevention - Detection e N. Date Taken e c e P.
Function s r c v c t N.
SUPPLIER
SEVERITY SCALE OCCURENCE SCALE DETECTION SCALE
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 97
Summary
DFMEA is a technique utilized primarily by a DesignDesignResponsible Engineer/TeamResponsible Engineer/Team as a means to assure that, tothe extent possible, potential failure modes and theirassociated causes have been considered and addressed.DFMEA:–– Identifies potential failure modesIdentifies potential failure modes early in development.–– Assists in evaluation of product requirementsAssists in evaluation of product requirements and
alternatives.–– Identifies special characteristicsIdentifies special characteristics–– Prioritizes design improvements.Prioritizes design improvements.
Using the team approachteam approach recognizes that a group workingtogether can accomplish more than individuals workingseparately.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 98
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Design for Manufacturing and Assembly(DFMA)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 100
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 101
DFMA:Learning Objectives
At the end of this module, participants will understand the use of . . .
Simultaneous Engineering and Cross Functional Teams
Benefits of DFMA, Benefits of DFMA, and its importance in the DFSS process
Methods such as Ten Guidelines for Design for Assembly
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 102
Design for Manufacturing andAssembly
The DFMA process answers the question: Is the designIs the designoptimum for manufacturing and assembly?optimum for manufacturing and assembly? DFMA isdefined as follows:
– Is any procedure or design process that considers theconsiders theproduction factorsproduction factors from the beginning of the product design.
– Every design activity, from conceptualization to evaluation, mustfocus on generation of a design that meets marketmeets marketexpectations and can be manufactured successfully.expectations and can be manufactured successfully.
Source: Computer-Integrated Manufacturing; James A. Rehg
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 103
Simultaneous Engineering /Cross Functional Teams
Simultaneously design the product and the process
Prevents over-the-wall design
Cross-functional teams continually evaluate each others workand have input on the whole product/process design
Simultaneous decision-making by design teams
Integrates product design & process planning
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 104
Breaking Down Barriers
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 105
DFMA Benefits
Shorter time to bring the product to marketMajor Concurrent Engineering DriverSmoother transition into productionOptimized manufacturing/assembly methods and processesBetter informed tooling and capital equipment decisionsFewer components in the final productEasier assemblyShorter assembly timeLower costs of productionMajor cost savings (parts & labor)Reduced defectsIncreased product quality and reliability.Greater customer satisfaction
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 106
Ten Guidelines forDesign for Assembly
1. Minimize the number of parts: Combine or eliminate parts whenever possible.2. Minimize assembly surfaces.3. Design for top-down assembly.4. Improve assembly access.5. Maximize part compliance.6. Maximize part symmetry.7. Optimize part handling.8. Avoid separate fasteners.9. Provide parts with integral self-locking features.10. Focus on modular design.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 107
Design Simplification (DFMAExample)
Design for push-and-snap assembly
(b) Revised design
One-piece base & elimination of fasteners
(a) The original design
Assembly using common fasteners
(c) Final design
Source: Boothroyd/Dewhurst, “ Design for Manufacturing and Assembly,” April 1988, Society of Manufacturing Engineers
Spindle/Housing Assembly
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 108
Design a product for easy & economical production (8 month)Baseline: 105 separate parts (reduced to 9 parts)Total calculated time: 1440 seconds (reduced to 258 seconds)Easy to fabricate & assemble componentsIntegrated product design with process planning
DFMA - Case Study
Source: Boothroyd Dewhurst
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 109
DFMA Summary
DFMA is recognized as the key to simultaneouslysimultaneouslyminimizing manufacturing cost, assuring productminimizing manufacturing cost, assuring productquality, and increasing productivity.quality, and increasing productivity.
It encourages teamworkencourages teamwork and a dialogue between designersand manufacturing engineers, and any other individual whoplay a part in determining the product costs during the earlystages of design.
The DFMA procedure often produces a considerableconsiderablereduction in part countreduction in part count, resulting in simpler and morereliable products which are less expensive to assemble andmanufacture.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 110
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Design for Reliability
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 112
Learning Objectives
At the end of this module, participants will understand the use of . . .
Definition of Reliability
Importance of Reliability
Assessment of Reliability
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 113
Definition of Reliability
ReliabilityReliability is a performance attribute of the product orsystem
It is also based on a Probability of SuccessProbability of Success, but is primarilyfocuses on Frequency of FailuresFrequency of Failures
Definition:Definition:
– The probability that an item will perform itsintended function under stated conditions, foreither a specified interval or over its useful life.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 114
Importance of Reliability
Typical performance measures are irrelevant if the productfails
– speed, capacity, range, and other “normal” performancemeasures
Reliability is critical to safety and can be a liability
Reliability is a primary factor in determining operating, repair,and warranty costs
Reliability determines whether or not a product is capable toperform its function
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 115
Occurrence of Failures
The “Bathtub Curve”– typically used for electronic equipment– applicable to mechanical equipment with less pronounced
constant failure rate regionFa
ilure
Rat
e ( λ
)
Infant Mortality Useful Life Wearout
Reliability Measure
Durability Measure(Life)
Design RelatedFailures
Quality RelatedFailures
Time
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 116
Everything Effects Reliability
PlanningPlanning– Market Survey– Benchmarking– QFD
DesignDesign– Critical Item ID &
Control– Derating– Design Reviews– Environment
Characterization– Fault Tolerance– Parts Application &
Selection– Supplier Control– Thermal Design
AnalysisAnalysis– Allocations– DOE– Dormancy Analysis– Durability Assessment– FMEA– Fault Tree Analysis– Finite Element Analysis– Life Cycle Planning– Modeling & Simulation– Part Obsolescence– Predictions– Repair Strategies– Sneak Circuit Analysis– Thermal Analysis– Translations– Worst Case Analysis– Statistical Analysis
TestTest– Accelerated Life
Test– Environmental
Stress Screening– Reliability Growth
Testing
ManufacturingManufacturing– SPC– Inspection
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 117
Prediction withReliability Analysis
Reliability Analysis can not tell you how to FIX the failure.The focus is prediction. Prediction of a potential problem.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 118
Summary
Reliability has a significant impact on our future costs.
Reliability should be considered with all engineering and
design decisions.
Reliability of the design must be predicted and verified.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 119
AgendaWhy DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Regression and Design of Experiments
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 121
Regression & DOE:Learning Objectives
At the end of this module, participants will understand the use of . . .
Models Models in the application of DFSS
Analytical, empirical , and semi-empirical equations asmodels.
DOEDOE and RegressionRegression analysis as tools for developing modelsfrom hardware and simulations
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 122
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 123
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 124
Fundamentals of DFSS
Product ModelProduct Model((equationequation, , simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
Y
C
E
A
B
D
ULLL T
PNCPNC
UnderstandingUnderstandingRequirements,Requirements,Specifications,Specifications,& Capabilities& Capabilities
ApplyingApplyingModels &Models &AnalysesAnalyses
PredictingPredictingProbability of Probability of
Non-ComplianceNon-Compliance
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 125
Regression Models:A Case for Action
ModelingModeling is the cornerstone of engineering design. There isno lack of models in any engineering organization.
However, most of these models can be in forms that are verycumbersomecumbersome, slowslow, or expensiveexpensive to use.
A critical enabler for DFSS is having fast, accurate modelsfast, accurate modelsfor analysisfor analysis (Sensitivity Analysis, Monte Carlo), allocationand optimization.
RegressionRegression techniques allow these fast, accurate modelsfast, accurate modelsto be created by approximating the original model.
Multiple types of models can be used in DFSS
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 126
The original model may be:The original model may be:
an Equation expert opinion, Engineering textbooks, basic physics
embedded in Data repository of historical data, ongoing data collection
a computer Simulation pSpice, ProEngineer, Matlab, Ansys, etc.
Prototypes physical models or mock-ups, pilot production lines
the Actual System the actual product or process being designed
?X2
X1
Xn
Y
Initial StateInitial State
Original ModelOriginal Model
Y=f(X)X2
X1
Xn
Y
Desired StateDesired State
Fast, Accurate ApproximationFast, Accurate Approximation
StatisticalModelingProcess
Increasing accuracyand expense
Types of Models
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 127
Types of Equations
Equations can come from many sources:– Textbooks– Basic Physics– Expert Opinion– Curve-Fits
Equations can be analytical, empirical, or semi-empirical.–– Analytical Equations:Analytical Equations: Relying on or derived from
analysis of elemental parts or basic principles–– Empirical Equations:Empirical Equations: Relying on or derived from
observation or experiment–– Semi-Empirical Equations:Semi-Empirical Equations: The format of the equation
has a basis in physics or first-principles and the coefficientsand/or exponents are fit to experimental data.
All equations are models.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 128
Empirical Equations
Empirical equationsEmpirical equations are typically the result of a regressionprocess where data is fit to a equationExample:Example: An equation was fit to 239 data points taken forthe thermal coefficient of expansion of copper as a function oftemperature.
0
5
10
15
20
25
0 200 400 600 800 1000
Temperature (K)
Co
eff
icie
nt
of
Th
erm
al E
xp
an
sio
n
Empirical Equation
Experimental Data
CTE = (1.08 –0.123 T + 4.09x10-6 T2 – 1.43x10-6 T3) / (1 – 0.00576 T + 2.41x10-4 T2 – 1.23x10-7 T3)
Data Source: NIST standard referencedataset Hahn11.dat
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 129
Semi-Empirical Equations
Semi-empirical equationsSemi-empirical equations combine experimental data withphysics-based analytical equations.Example:Example: The heat transferred from the hot walls of a pipe toto a cold fluid flowing through the pipe has the followingphysics-based, analytical relationship:
where a, b, and c are unknown coefficients.Experiments were performed at various flow rates for threefluids:The unknown coefficientsa,b, and c were determinedby fitting the experimentaldata to the equation.
Heat Transferred = a • (Flow Rate)b • (Fluid Property)c
3
6
6
0.7
0.7
Fluid Property
22.2
44.0
31.8
18.6
12.4
Heat Transferred
3400
6000
4000
5000
3000
Flow
Heat Transferred = 0.023 • (Flow Rate)0.8 • (Fluid Property)0.33
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 130
Say we have a system, and we don’t have a good understandingof how its outputs are affected by its attributes.
One way to gain understanding is to vary the system attributes andcollect data on the system outputs.DOE is a structured method, based in statistics, for collecting andanalyzing this data.
Design of Experiments (DOE)
SystemSystemOutputsOutputs
(Responses)(Responses)
•• PerformancePerformanceParametersParameters
•• YieldYield•• CostCost•• ScheduleSchedule•• QualityQuality
SYSTEM
• Design Prototypes• Manufacturing Process• Computer Simulation • Purchased Components• Etc.
SystemSystemAttributesAttributes(Factors)(Factors)
•• DimensionsDimensions•• TolerancesTolerances•• ComponentComponent
ValuesValues•• ProcessProcess
ParametersParameters•• MaterialsMaterials
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 131
Role of DOE in DFSS
Primary role of DOE is for model creation:model creation:
Results in a fast approximate model that mimicsthe data set.These models are then used in analyzing,allocating and optimizing variability.
Results in a fast approximate model that mimicsthe data set.These models are then used in analyzing,allocating and optimizing variability.
Original(slow,
complex)System
Regression
RepeatedRuns
Data Set
Y = f(x)
Math Model
DOE
Inputs, X Output, Y
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 132
Design of Experiments (DOE)
DOE is a structured method, based in statistics, forDOE is a structured method, based in statistics, forrunning a set of tests or analyses on a system, product,running a set of tests or analyses on a system, product,or process.or process.
Two basic applications:Two basic applications: screening and modeling
–– ScreeningScreening experiments help identify the most significantfactors from a larger initial set.
–– ModelingModeling experiments yield equations that approximate thesystem’s behavior.
–– BothBoth applications can indicate directions for applications can indicate directions forimprovement.improvement.
Compared with random or one-factor-at-a-time testing, muchmore information can be extracted from fewer runs.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 133
What Does a RegressionModel Look Like?
Only some model terms aresignificant (implied by P < 0.05)
The output model is:Y = 97.74 - 8.00 X1 + 8.40 X2
- 26.80 X3 – 9.50 X1 X3
Output is from Minitab
Standard Form: Y = β0 + β1x1 + β2x2 + … + β12x1x2 + … + β11x12 +
…
Inputs, X
Interaction between inputs, X1 and X3
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 134
Inputs[X’s,Y’s]
Excel-CompatibleModel
Polynomial EquationY=a + b • X1 + c•X1
2
Physics-Based Equatione.g. Y=a • X1
b X2c
Lookup Table[X1 X2 Y1] . . . [Xi Xj Yk]
Linear Regression
Nonlinear Regression
Nonparametric Regression
Regression Techniques
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 135
Regression and DOE:Summary
IntroducedIntroduced models as a necessary part of DFSS
HighlightedHighlighted types of equations and compared them
– Analytical, Empirical, and Semi-Empricial
Introduced Introduced how Regression and DOE can be used toproduce models
Discussed Discussed how models are used in design analysis
Opened the door for using Models in DFSSusing Models in DFSS
Key Message … Key Message … Models are a must for DFSSModels are a must for DFSS
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 136
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
DFSS Tool Box
• Sensitivity Analysis
• Monte Carlo Analysis
• Statistical Allocation
• Optimization
• Other Enablers to DFSS
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 138
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 139
Sensitivity Analysis:Learning Objectives
At the end of this module, participants will understand the use of . . .
Sensitivity Analysis Sensitivity Analysis as a means of generating µ, σ, and PNCfor a response given:
– an equation or Excel workbook relating a responsevariable to a set of input factors,
– mean and standard deviation (µ and σ) for each inputfactor, and
– upper and lower limits on the response
Sensitivity Analysis WorksheetsWorksheets in design and analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 140
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 141
Purpose: Compute variability of an output from a deterministicmodel (i.e. equation)
Information Needed:
Means and standard deviations (µ and σ) for the input factorsA model that relates a response to a set of input factorsSpecification limits (LSL and USL) on the response
Results Obtained:
Response mean and standard deviation (calculation for µ and σ)Relative importance of the factors’ contributions to the responseThe expected Probability of Non-Compliance
Sensitivity AnalysisMethod Overview
Existing ModelExisting Model(equation, (equation, simulation,simulation,
workbook, etc.)workbook, etc.)
Y
USLLSL T
PNCPNC
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
A
B
D
C
E
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 142
Applying Sensitivity Analysis
We have an existing heat sink design that, using a 25 mmfan, looks feasible using nominal analysis.
Will this existing heat sink perform well consideringmanufacturing and processor variation?
If not, what is the probability of non-compliance?
Will this existing heat sink perform well consideringmanufacturing and processor variation?
If not, what is the probability of non-compliance?
Length = 40 mmWidth = 40 mm
Inlet Temperature = 44 °CHeat Rejected = 34 W
Number of Fins = 16Fin Thickness = 0.5 mm
Fin Height = 15 mmNumber of Fans = 1
Fan Location = 0 (front)
Base Temperature= 99.1 °C
(USL = 100 °C)
Inputs, X Output, YProduct
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 143
39.0 39.5 40.0 40.5 41.0
Gather Information for Inputs
µ = 40σ = 0.333
Heat sink fin thickness: ± 0.13 mm
Processor heat rejected: ± 0.5 WHeat sink length & width: ± 1 mm
Heat sink fin height: ± 0.5 mm
14.50 14.75 15.00 15.25 15.50
Consulting supplier data sheets and process data,we identify manufacturing tolerances and variation.We assume tolerances are 3 Sigma.
µ = 15σ = 0.167
0.37 0.44 0.50 0.56 0.63
33.50 33.75 34.00 34.25 34.50
µ = 0.5σ = 0.043
µ = 34σ = 0.289
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 144
Sensitivity Analysis Example
Outputs:Outputs: Y Mean Y Mean Y Std Deviation Y Std Deviation PNC PNC
… and … and Contribution Contribution of X’s to of X’s to Variation Variation
Inputs:Inputs: Equation Equation X Values X Values X Variation X Variation
Conclusion: PNC = 16.4%Conclusion: PNC = 16.4%Conclusion: PNC = 16.4%
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 145
Sensitivity Analysis:Summary
Quick and easy to use for most functions using theSensitivity Analysis tool in Excel
Allows response variation to be predictedAllows response variation to be predicted fromknowledge of the equation, nominal values, and assumedtolerances.
Provides contribution to varianceProvides contribution to variance for each parameter tosupport cost, process capability, and tolerance decisions.
Accurate for functions with a linear response within therange of variability (≈ ±3σ) for each factor
– non-linear responses can be handled with Monte Carlo analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 146
Monte Carlo Analysis:Learning Objectives
At the end of this module, participants will understand the use of . . .
Monte Carlo analysis Monte Carlo analysis as a means to generate a probabilitydistribution and PNC for a response given:
– an equation or Excel workbook relating a response variable to aset of input factors,
– a probability distribution for each input factor, and
– upper and lower limits on the response
the Crystal BallCrystal Ball Monte Carlo tool in design and analysis
Monte Carlo techniquesMonte Carlo techniques and when to use them
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 147
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 148
Purpose: Compute variability of outputs from a deterministicmodel (i.e. equations or sets of equations)
Information Needed:
A model that relates a response to a set of input factorsProbability distributions for the input factorsSpecification limits (LSL and USL) on the response
Results Obtained:
A probability distribution for the response (statistical data analysis)Relative importance of the factors’ contributions to the responseThe expected Probability of Non-Compliance
Monte Carlo Method Overview
Existing ModelExisting Model(equation, (equation, simulation,simulation,
workbook, etc.)workbook, etc.)
Y
A
B
D
C
EUSLLSL T
PNCPNC
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 149
What isMonte Carlo Analysis?
A number of Monte Carlo trials,A number of Monte Carlo trials, n,n, is chosen up front.is chosen up front.
For each trial,For each trial,
– A random value is generated for each input factor that followsits specified distribution.
– A response value is computed using these input factor values.
What results is a samplesample of nn response values.
Sampling statistics are used to compute response mean andstandard deviation, and can also be used to fit a probabilitydistribution to the data.
Proportions and Confidence Intervals are used for computingPNC.
In Summary - Monte Carlo is solving the equations over andover again using randomly selected values and collecting theresults for statistical analysis.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 150
Applying Monte Carlo Analysis
We have an existing heat sink design that, using a 25 mmfan, looks feasible using nominal analysis.
Will this existing heat sink perform well consideringmanufacturing and processor variation?
If not, what is the probability of non-compliance?
Will this existing heat sink perform well consideringmanufacturing and processor variation?
If not, what is the probability of non-compliance?
Length = 40 mmWidth = 40 mm
Inlet Temperature = 44 °CHeat Rejected = 34 W
Number of Fins = 16Fin Thickness = 0.5 mm
Fin Height = 15 mmNumber of Fans = 1
Fan Location = 0 (front)
Base Temperature= 99.1 °C
(USL = 100 °C)
Inputs, X Output, YProduct
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 151
Perform Monte CarloAnalysis with Crystal Ball
5. Run Simulation and analyzeresults from multiple trials.
2. Set each input factoras an “assumption” cell.
4. Set the responseas a “forecast” cell.
39.0 40.0 41.0
14.50 15.00 15.50
0.37 0.50 0.63
33.50 34.00 34.50
3. Define statisticalproperties for the
input X’s
1. Define ExcelTM
Spreadsheet Model
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 152
Monte Carlo Analysis:Summary
Allows Allows a probability distribution and PNC for a responseto be computed given:
– an equation or Excel workbook relating a response variable to aset of input factors,
– a probability distribution for each input factor, and
– upper and lower limits on the response
Crystal Ball Crystal Ball (Decisioneering) is a Monte Carlo tool that canbe used with any Excel spreadsheet
Monte Carlo techniquesMonte Carlo techniques can be used for any analysis orsimulation
Monte Carlo AnalysisMonte Carlo Analysis provides a “virtual productmanufacturing” process
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 153
Statistical Allocation:Learning Objectives
Statistical allocationStatistical allocation as a means to meet a desired level ofresponse PNC given:
– an equation or Excel workbook relating a response variable to aset of input factors,
– upper and lower limits on the response, and
– a target value for response standard deviation or PNC
Allocation Worksheets Allocation Worksheets for assigning tolerances based oncontribution to variation, process capability, or cost.
At the end of this module, participants will understand the use of . . .
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 154
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Best PracticeBest Practice
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 155
Why is Allocation Needed?
Most Product requirementProduct requirementdefinitiondefinition begins at the TOP
– It must do this …
– We must produce a quantity of …
– The cost must not exceed …
– The production yield must be greater than …
– The profit goal is …
Assembly and ComponentAssembly and Componentrequirementsrequirements flowdown from theProduct requirements
How well the TOP level requirement ismet is dependent upon how well theAssembly and Componentrequirements are met.
Once the Product design is verified,manufacturing variation is usually theproblem
ProductRequirement
Assembly ARequirement
Assembly BRequirement
Comp A
Reqmt
Comp B
Reqmt
Comp C
Reqmt
Comp D
Reqmt
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 156
Why is Allocation Needed?
If Assembly and Componentvariation is known, thenANAYLSISANAYLSIS can be used todetermine Product variation
ProductRequirement
Assembly ARequirement
Assembly BRequirement
Comp A
Reqmt
Comp B
Reqmt
Comp C
Reqmt
Comp D
Reqmt
AnalysisAnalysis ProductRequirement
Assembly ARequirement
Assembly BRequirement
Comp A
Reqmt
Comp B
Reqmt
Comp C
Reqmt
Comp D
Reqmt
AllocationAllocation
If Product variation is known,then ALLOCATIONALLOCATION can be usedto determine Assembly andComponent variation
Use Sensitivity Analysis Use Sensitivity Analysis and Monte Carlo to do thisand Monte Carlo to do this
Use Statistical AllocationUse Statistical Allocationto do thisto do this
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 157
Purpose: Allocate variability to inputs of a deterministic model
Information Needed:A model that relates a response to a set of input factors
Input factor means µµii
Specification limits (LSL and USL) on the response
The desired Probability of Non-Compliance
Results Obtained:Factor standard deviations σσii
Relative importance of the factors’ contributions to the response
Allocation Method Overview
Existing ModelExisting Model(equation, (equation, simulation,simulation,
workbook, etc.)workbook, etc.)
Y
USLLSL T
PNCPNC
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
(µ,σ)
A
B
D
C
E
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 158
Statistical Allocation:Typical Results
Outputs:Outputs: X Std Deviation X Std Deviation PNC PNC
Contribution ofContribution ofX’s to VariationX’s to Variation
Inputs:Inputs: Equation Equation X Values X Values
Inputs:Inputs: Y Std Deviation Y Std Deviation Allocation Drivers Allocation Drivers
TolerancesTolerances
Heat Sink SA Demo
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 159
Statistical Allocation:Summary
Quick and easy to performQuick and easy to perform for most functions using theStatistical Allocation tool in Excel.
Uses budgeted response variationUses budgeted response variation as a requirement fordetermining parameter variations or tolerances.
AllowsAllows the engineer to make component tolerance decisionsbased on contribution to variation, process capability, or cost.
ReplacesReplaces iterative, trial & error methods typically used forassigning tolerances.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 160
Optimization:Learning Objectives
At the end of this module, participants will understand the use of . . .
OptimizationOptimization in engineering design:
–– Single-Objective:Single-Objective: automated search of alternatives
–– Multi-Objective:Multi-Objective: trade study tool
Trade Studies Trade Studies to identify designs that meet multiplerequirements and multiple PNC goals simultaneously
ApogeeApogee, a tool for statistical multi-objective optimization
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 161
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Good PracticesGood Practices
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 162
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 163
Traditional View– Maximize (or minimize) a singlesingle objective function f(x)
– Subject to constraints gi(x) < ci , i = 1, 2, ... NDesignvariables {x} are real-valued (continuous or discrete)
– Design variables {x} are real-valued (continuous or discrete)
– Output is a set of values {xi} that is the “optimum”
Working Definition
– Rapid, automated search and evaluation of design alternatives
Implementation– Employ software packages to find “optimum” values for {x}
– Each package uses different algorithms (gradient based, heuristic, etc.)
– Problem formulation is fed to package as input
What is Optimization?
Need modelsfor f(x) and gi(x)
Need modelsfor f(x) and gi(x)
Specificinstances {xi}
Specificinstances {xi}
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 164
Single-Objective Vs..Multi-Objective
BenefitsBenefitsBenefits
MathFormulation
MathMathFormulationFormulation
SolutionApproach
SolutionSolutionApproachApproach
Set f1(x) = T1
f2(x) = T2
fk(x) = Tk
Subject to g1(x) < c1
g2(x) < c2
gn(x) < cn
Maximize f(x)
Subject to g1(x) < c1
g2(x) < c2
gn(x) < cn
Find “optimum” values for{x} that maximize f(x)
Find best values for {x} thatbring each fi(x) onto its target Ti
Rapid, automated searchof design space
Trade studies.Trade studies. Exploredifferent priorities forbringing each goal on target
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 165
Y=f(xY=f(x11,x,x33,x,x55))PerformancePerformance
Target Target
Yperf 1
Y=g(xY=g(x11,x,x33,x,x22) ) CostCost
Y=h(xY=h(x33,x,x22,x,x55))PerformancePerformance
TargetTarget
Multi - ObjectiveOptimization
USLLSL T
PNCPNC
USLLSL T
PNCPNC
Yperf 2
YCost
X1 (µ,σ)
X3(µ,σ)
X3(µ,σ)
X3(µ,σ)
X1 (µ,σ)
X5 (µ,σ)
X2 (µ,σ)
X2 (µ,σ)
X5 (µ,σ)
USLT
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 166
Optimization Example:Heat Sink Design
We discovered that the existing heat sink design hasunacceptable performance when manufacturing variation isconsidered.
Length = 40 to 60 mmWidth = 40 to 60 mm
Inlet Temperature = 44 °CHeat Rejected = 34 W
Number of Fins = 10 to 22Fin Thickness = 0.25 to 1.25 mm
Fin Height = 15 to 25 mmNumber of Fans = 1 or 2
Fan Location = 0 (front) or 1 (back) or 2 (both)
Base Temp < 99.1 °CBase Temp PNC < 0.001
Cost = 0Base Temp USL = 100 °C
Current PNC = 0.16
Can we find a different nominal design that will improve PNCwithout having to tighten tolerances?Can we find a different nominal design that will improve PNCwithout having to tighten tolerances?
Input Factor Ranges(Xmin to Xmax)
Models for CostPerformance,
etc. (Y)Constraints,
Goals,Priorities Length
WidthNo. of Fins
Fin Thickness Fin Height
No. of FansFan Location
SelectedValues (X)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 167
Optimization Formulationusing Apogee
Define inputs, variation, and models that capture system behavior.Define inputs, variation, and models that capture system behavior.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 168
Optimization Formulation
Link to customer requirements.Define constraints, goals, and priorities.
Link to customer requirements.Define constraints, goals, and priorities.
RunOptimizationand ReviewResults
RunOptimizationand ReviewResults
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 169
Optimization Results
By exercising the formulation, a familyfamily of solutions is created:
High PerformanceHigh Performance
Low CostLow Cost BalancedBalanced
Plus Outlet Air Req’tPlus Outlet Air Req’t
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 170
EXPLORE MORE DESIGN ALTERNATIVESEXPLORE MORE DESIGN ALTERNATIVES– Evaluate hundreds or thousands of options in a few hours– Orders-of-magnitude improvement on standard process
CONSIDER MORE DESIGN ATTRIBUTES: CONSIDER MORE DESIGN ATTRIBUTES: Trade StudiesTrade Studies– Easy to add cost models, quality models, cycle time models– Trade off performance with cost, quality, and schedule
GENERAGENERATE BETTER DESITE BETTER DESIGNSGNS– More designs evaluated across more attributes in less time.
Improved designs are clearly a probable outcome.
IIMPROVE CUSTOMER RELATIONSMPROVE CUSTOMER RELATIONS– Present customer with options
“Your costs can be cut by a third if you’d relax this one requirement...”
– Empower customer to make choices
Benefits of Optimization
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 171
Other Enablers to DFSS:Learning Objectives
At the end of this module, participants will understand the use of . . .
Process capability data Process capability data in the design process
Gage R & RGage R & R to verify that performance is properly assessed
Test EffectivenessTest Effectiveness calculations to verify that test results arereliable and correct
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 172
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Good PracticesGood Practices
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 173
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 174
15.19 14.79 14.77 15.01 14.9215.09 14.76 14.79 15.14 14.9914.84 15.2 14.99 15.09 15.2515.21 15.25 14.79 14.92 14.8114.83 15.14 14.94 15.13 14.8514.84 14.82 14.97 14.98 14.9515.21 15.01 14.79 14.89 14.8214.77 14.77 14.81 15.27 15.0714.97 14.75 14.99 15.16 15.0115.08 15.03 14.98 14.88 14.89
Use Process Capability Data
Product ModelProduct Model
Y = f(X)Y = f(X)YC
E
A
B
D
ULLL T
PNCPNC
14 15 16
0.05
0.1
CatalogSpecs or Supplier
Estim ates
Sim ilar ProductM easured Data
Process andM aterialDatabase
Perform anceEstim ates
LS US
Potential,Expensive
Catalog,Inexpensive
LS US
LS US
LS US
OneSupplier
with ShiftedM ean
M ultipleSupplierswith ShiftedM eans
ProcessOut ofControl Statistical Methods must be used
to analyze data and test methodsto prevent part and materialrelated contributions to PNC.
Statistical Methods must be usedto analyze data and test methodsto prevent part and materialrelated contributions to PNC.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 175
MeasurementError
Spec Width
Actualoperatingpoint
The Effect ofMeasurement Error
In a perfectperfect world, a test would measure the actualoperating point of a parameter and all tests would beaccurateHowever, in the real world,– If we test a given unit many times, we will probably get a
range of measurements
Testing is a process, just like anyother
All measurements have someerrorDue to measurement error, wecould pass a bad unit or fail agood unit
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 176
Measurement system studies assess how much variation isassociated with the measurement system as compared to themanufacturing process variationGage studies assess:
Defining Measurement SystemAccuracy with Gage R & R
RepeatabilityRepeatabilityAccuracyAccuracy
Accuracy
How close is the measurement tothe true value?
True value
ReproducibilityReproducibilityHow close are series of measurementsby several people on the same part on
the same equipment?
Repeatability
True value
How close is a series of measurementson one part by one person?
Reproducibility
True value
Operator A
Operator C
Operator B
Total measurement variance is the sum of the repeatabilityand reproducibility variances
We can estimate these variances using Gage R & R
222& ityrepeatabililityReproducibrR σσσ +=
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 177
Understanding Test Effectiveness
In terms of products to be tested, there are two types–– Good UnitsGood Units and Bad UnitsBad Units
In terms of product test results, there are two outcomes–– Passed UnitsPassed Units and Failed UnitsFailed Units
Of the four possible test result outcomes, only two areacceptable– P(Good+Passes) and P(Bad+Fails)
EventsGoodUnits
BadUnits Total
Unit FailsTest
P(G+F) P(B+F) P(F)
Unit PassesTest
P(G+P) P(B+P) P(P)
Total P(G) P(B) P(total)=1
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 178
Test EffectivenessAnalysis
USLXbar
ProductVariation
ProductVariation
MeasurementVariation
MeasurementVariation
Probability of Probability of Passing a Bad UnitPassing a Bad Unit
Probability of Probability of Failing a Good UnitFailing a Good Unit
LowerUSLXbar
Integrating the concepts of product and measurementvariability allows us to calculate the true test efficiencytrue test efficiency
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 179
We can make thetest effectivenesscalculation using– Excel– Minitab– or a simple
Windows program
Inputs:– Process mean– Process Std
Deviation– Test Error
Outputs:– Overall test
probabilities– Conditional
Pass/FailProbabilities
Test EffectivenessCalculations
Test Effectiveness
Bivariate example
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 180
Other Enablers to DFSS:Summary
All measurements include both the real value process resultplus error
Measurement error can have a large influence on the results
We can predict the measured results during the design stageand use this information to further optimize productperformance
Measurement error is obtained using Gage R & R studies
Test Effectiveness Calculations tell us the consequences ofprocess variation and test variation
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 181
Agenda
Why DFSSWhy DFSSDFSS OverviewDFSS OverviewQuality Function Deployment (QFD)Quality Function Deployment (QFD)BenchmarkingBenchmarkingTRIZ / Trade-Offs (Pugh, and SDI Analysis)TRIZ / Trade-Offs (Pugh, and SDI Analysis)Program Management (LPMP)Program Management (LPMP)Design Failure Mode and Effects Analysis (DFMEA)Design Failure Mode and Effects Analysis (DFMEA)Design for Manufacturability and Assembly (DFMA)Design for Manufacturability and Assembly (DFMA)Design for Reliability (DFR)Design for Reliability (DFR)Regression and Design of Experiments (DOE)Regression and Design of Experiments (DOE)DFSS Tool BoxDFSS Tool Box
–– Sensitivity AnalysisSensitivity Analysis–– Monte CarloMonte Carlo–– AllocationAllocation–– OptimizationOptimization–– Other EnablersOther Enablers
Statistical Roll Up and Score CardStatistical Roll Up and Score CardConclusion / Wrap-Up and DiscussionConclusion / Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved.
Revised 17 Dec 03
Statistical Roll-up and Score Card
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 183
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Good PracticesGood Practices
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 184
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 185
Statistical Roll-up & Scorecard:Learning Objectives
At the end of this module, participants will understand the use of . . .
Model-based approaches to rollup results and generatescorecards:
– Roll up statistical information (µ, σ) to the system level usingexisting engineering models
– Use PNC as the fundamental metric for each requirement
– Use scorecards to evaluate multiple requirements at any level inthe system
Scorecards Scorecards as a method to drive the right DFSS behavior ona program
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 186
Metric Simplification
Fundamental metric is the probability of not meeting arequirement: the Probability of Non-ComplianceProbability of Non-Compliance (PNC) (PNC)
ComputeCompute PNC by applying statistics to existing engineeringanalyses.
IncludeInclude all known sources of variation in the distributionabove. Thus avoid the categorization of “short term” and“long term”, and avoid enforcing any arbitrary sigma shifts.
For external reportingexternal reporting, can convert PNC into Z, Sigma, ordpmo values (if truly required).
“Non-compliant”
USLLSL T
“Non-compliant”“Compliant”
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 187
Statistical Rollup
Use Sensitivity Analysisor Monte Carlo tocompute statistics forsystem parameters.
Use output of eachanalysis as input to thenext higher level in thesystem.
Compute system PNC bycomparing top-levelsystem parameters withcustomer specifications.
Customer Specifications (LSL, USL, Target)Customer Specifications (LSL, USL, Target)
Manufacturing CapabilitiesManufacturing Capabilities
System Parameter Y1System Parameter Y1
YY11 = f(X) = f(X)
XX11 = f(P) = f(P) XX22 = f(P) = f(P) XXnn = f(P) = f(P)
System ModelSystem Model
SubsystemSubsystemModelsModels
Subsystem Parameters {XSubsystem Parameters {X11, X, X22, ... X, ... Xnn}}
Component Parameters {PComponent Parameters {P11, P, P22, ... P, ... Pmm}}
(µ1, σ1) (µ2, σ2) (µm, σm)
(µ, σ)
(µ, σ)(µ, σ) (µ, σ)
(µ, σ) (µ, σ) (µ, σ)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 188
Implementation ofStatistical Rollup
For any requirement, apply Sensitivity Analysis or Monte Carloto each model in the hierarchy, propagating the outputsupward.
USLLSL T
PNCPNC
A (µ,σ)
B (µ,σ)
D (µ,σ)
C (µ,σ)
E (µ,σ)
(µ,σ)
(µ,σ)
Y(µ,σ)
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 189
Program Name:Prepared By: PNC
Date: Identified Goal Met:
Notes: Addressed Goal Unmet:
Allocated # Reqts: 0Enter phase names and dates here: Analyzed
Program Phase: Planning Prototype Pilot Launch Post Launch Measured
Planned Completion: 1-Nov-03 1-Jan-04 1-Mar-04 1-Aug-04 1-Oct-04 Controlled
Name Units Target LSL USL PNC Goal State Distribution Mean StDev Gage R&R PNC123456
2001 Statistical Design Institute, LLC. All Rights Reserved.
Current State
SummaryState Completion Status
Critical Requirements
24-Oct-03
Insert Reqts
Delete Reqts
Use a Scorecard toTrack Multiple Requirements
Gather key requirements at the systemlevel.
Track CTC state and PNC goals.
Use as input for project management.
Gather key requirements at the systemlevel.
Track CTC state and PNC goals.
Use as input for project management.
From StatisticalAnalysis
From StatisticalFrom StatisticalAnalysisAnalysis
From Voice of the Customer
From Voice of From Voice of the Customerthe Customer
From ProgramGoals
From ProgramFrom ProgramGoalsGoals
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 190
Driving Program BehaviorWith DFSS Scorecards
Red Red flags:flags:– CTC’s that are lagging behind the
program schedule or are not meetingPNC goals
Actions:Actions:– Escalate issues– Remove barriers
to execution– Manage risks
Program Phase: Planning Prototype Pilot Launch Post LaunchIdentif ied
Addressed
Allocated
Analyzed
Measured
Controlled
Ratings: Requirements in this state at phase completion are acceptable Requirements in this state at phase completion should be examined Requirements in this state at phase completion are lagging
State Definitions: Identified -
Addressed -Allocated -
Analyzed -Measured -Controlled -
2001 S ta tis tica l Des ign Ins titute , LLC. All R ights Res erved.
A requirement that has a PNC computed from measured data.A requirement that has been transitioned to production processes that are in control.
A requirement that doesn't have specifications defined.A requirement that has a specified Target, LSL, and USLA requirement that has an allocated mean and standard deviation to meet a specified PNC Goal. A requirement that has a PNC estimated from statistical analysis.
Expected Statesfor Critical Product Requirements at Phase Completion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 191
Statistical Roll-up & Scorecard:Summary
DiscussedDiscussed statistical rollup using existing SensitivityAnalysis and Monte Carlo techniques
IntroducedIntroduced a scorecard for tracking multiple requirements
OutlinedOutlined how to use scorecards to drive DFSS behavior
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 192
DFSS Deployment DFSS Deployment
Things to look for -Things to look for -
– Full team participation in the development of newproducts.
– Activities focused on identifying customer requirements.
– An emphasis on the variation of a Product’s Performance inaddition to the nominal response.
– Attention to Cost Avoidance in addition to Performance andReliability.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 193
DFSS Deployment DFSS Deployment
Questions to ask -Questions to ask -
– How were the requirements for this Product determined?– Have we prioritized the requirements that are Critical to
the Customer?– Have the Performance Parameters for this Product been
Analyzed?– Do we have statistical data for the Product Parameters?– What type of Product Model did you use in the Performance
Analysis?– What other approaches did we consider?– Do we understand the factors that will cause variation in
the Product Response?– Which Parameters are expected to contribute the greatest
to the Response variation?
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 194
DFSS DeploymentDFSS Deployment
More Questions to Ask -More Questions to Ask -
– Have Parameter Tolerances been assigned? Are theParameter Tolerances based on Process Capabilities? Priorexperience? Statistical Analysis?
– Have opportunities to relax Parameter Tolerances (andCost) been investigated?
– Which Parts will be measured prior to Product assembly?– Is the quality of the Product ‘designed-in’ or ‘tested-in’?– What is the ‘Probability of Non-Compliance’ for each
requirement?– Have trade studies been performed to optimize
performance, cost, reliability, …?– Have defects due to test variation been considered?– Have FMEA methods been used to identify and reduce
defects?
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 195
Concepts forDesign for Six Sigma
DFSS ProcessDFSS Process
Good PracticesGood Practices
Build Models
Voice of the Customer
Design that best meetsall requirements
DFSS Enablers & ToolsDFSS Enablers & Tools
DFSSDFSS
DevelopmentDevelopmentProcessProcessConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
PostLaunchPostPost
LaunchLaunch
DOE and RegressionDOE and Regression
Monte Carlo AnalysisMonte Carlo Analysis
Statistical AllocationStatistical Allocation
Multi-Objective OptimizationMulti-Objective Optimization
TRIZ & Design Trade-offTRIZ & Design Trade-off
Quality Function DeploymentQuality Function Deployment
ScorecardsScorecards
Test Effectiveness AnalysisTest Effectiveness Analysis
Optimize the Design• Analyze Variability• Allocate Variability• Optimize Variability
Identify CriticalRequirements
Define Alternatives
Verify & Validate
Sensitivity AnalysisSensitivity Analysis
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 196
DFSS ToolApplication Roadmap
Programidentified
Prototypes Simulation/Computer Models
Monte CarloAnalysis
µx, σx
µy, σy, PNC
σx
DescriptiveStatistics
Process data, Supplier data,and Reliability Data[GR&R and Test Effectiveness must be performed]
*
Scorecard
RegressionAnalysis
EquationsEquations
Requirements&
Specifications(LL, T, UL)
Analytical Models
µy, σy, PNC
Design ofExperiments
Historical Data
µy, σy, PNCY
ULLL T
PNCPNC
Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
C
E
A
B
D
SensitivityAnalysis
ToleranceAllocation
MonteCarlo
Multi-ObjectiveOptimization
Equations
*
ConceptDesign
QFD
Whats/Hows
TRIZExperience &Brainstorming
BenchmarkingTrade-Off
***
DFM/A
DFMEA
DFR
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 197
DFSSApplication Assessment
What
do y
ou k
now
?Customer Needs (What)Priority of Needs (Importance)Technical/Capability GapContradictionDesign AlternativesKey CharacteristicsCharacteristic DegradationMeasured DataProduct/Process FactorsExperiment MatrixProduct/Process ResponseProduct Model, Equation, or SimulationInput Parameter StatisticsInput Parameter PDFInput Parameter MeanInput Parameter Std DevResponse ToleranceResponse Std Dev or PNCResponse ToleranceInput Parameter Design SpaceResponse or Sub-Assembly Constraints Response GoalsGoal PrioritiesParameter ConceptsParameter Failure RatesAssembly ConfigurationsFunctional Block DiagramSample Test Results for Units, Operators, EquipTest Result Mean & Std DevTest Upper & Lower Spec LimitsTest ErrorMean ShiftKey Product RequirementsLPMP Phase Completion DatesRequirement Definition & Analysis StateEvaluated Response Mean, Std Dev, PDF
What
do y
ou n
eed t
o k
now
?
Product Requirements (How)Priority of Needs (Tech. Weight)What-How Relationships What-What CorrelationComparative AnalysisImprovement PlanInventive PrinciplesInventive Principle ExamplesRanking of Design AlternativesStatistics (mean, std dev, skew, kurtosis)HistogramProbability Density Function (PDF)Parameter or Factor SignificanceParameter or Factor InteractionsResponse Mean Prediction EquationResponse Std Dev Prediction EquationResponse Sensitivity to Input VariationResponse PNCResponse MeanResponse Std DevResponse Mean Shift due to Input VariationGoal & Constraint Evaluation versus
Mean, Std Dev, or P(fail)Design AlternativesTrade Study ResultsProduct ReliabilityReliability Block DiagramTest RepeatabilityTest ReproducibilityProbability Good Unit Tests BadProbability Bad Unit Tests GoodRequirement PNC & SigmaLPMP Requirement Analysis StatusNumber of Requirement PNCs MetNumber of Requirement PNCs UnMet
What
DFS
S T
ool ca
n b
e use
d?
QFD
Benchmarking
TRIZ
Trade-Off
Descriptive Statistics
DOE
Monte Carlo
Sensitivity Analysis
Tolerance Allocation
Multi-Obj Optimization
Reliability Analysis
Gage R&R
Test Effectiveness
Scorecard
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 198
DFSS Method & ToolApplication Guide
Method Inputs OutputsQuality Function Deployment
BenchmarkingTRIZ
Trade-Off
Descriptive Statistics
Design of Experiments
Monte Carlo
Sensitivity Analysis
Tolerance Allocation
Customer Needs (What)Priority of Needs (Importance)
Technical/Capability Gap
Contradiction
Design AlternativesKey CharacteristicsCharacteristic DegradationMeasured Data
Product/Process FactorsExperiment MatrixProduct/Process Response
Product Model, Equation, or SimulationInput Parameter StatisticsInput Parameter PDFProduct EquationInput Parameter MeanInput Parameter Std DevResponse Tolerance
Product EquationInput Parameter MeanResponse Std Dev or PNCResponse Tolerance
Product Requirements (How)Priority of Needs (Tech. Weight)What-How Relationships What-What CorrelationComparative AnalysisImprovement PlanInventive PrinciplesInventive Principle ExamplesRanking of Design Alternatives
Statistics (mean, std dev, skew, kurtosis)HistogramProbability Density Function (PDF)Parameter or Factor SignificanceParameter or Factor InteractionsResponse Mean Prediction EquationResponse Std Dev Prediction EquationResponse StatisticsResponse PDFResponse Sensitivity to Input VariationResponse PNCResponse MeanResponse Std DevResponse Sensitivity to Input VariationResponse Mean Shift due to Input Variation Response PNCResponse MeanInput Parameter Std DevResponse Sensitivity to Input VariationResponse Mean Shift due to Input Variation
Iden
tify
Req
mts
an
d C
on
cep
ts
Mo
del
Data
&
Pro
du
cts
An
aly
ze a
nd
All
oca
te V
ari
ati
on
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 199
DFSS Method & ToolApplication Guide
Method Inputs Outputs
Multi-Objective Optimization
Reliability Analysis
Gage R&R
Test Effectiveness
Scorecard
Input Parameter Design SpaceInput Parameter Std DevProduct Model, Equation, or SimulationResponse or Sub-Assembly
ConstraintsResponse GoalsGoal PrioritiesParameter ConceptsParameter Failure RatesAssembly ConfigurationsSample Test Results for Units, Operators, Equipment,
Test Result Mean & Std DevTest Upper & Lower Spec LimitsTest ErrorMean ShiftKey Product RequirementsRequirement Target, LL, ULLPMP Phase Completion DatesRequirement Definition &
Analysis StateEvaluated Response Mean, Std
Dev, PDFEstimated Sigma Shift
Goal & Constraint Evaluation versusMean, Std Dev, or P(fail)
Design AlternativesTrade Study Results
Product ReliabilityReliability Block Diagram
Test RepeatabilityTest Reproducibility
Probability Good Unit Tests BadProbability Bad Unit Tests Good
Requirement PNC & SigmaLPMP Requirement Analysis StatusNumber of Requirement PNCs MetNumber of Requirement PNCs UnMet
Op
tim
ize
Para
mete
rs
& V
ari
ati
on
An
aly
zeTest
Meth
od
s
Rep
ort
P
roje
ctR
esu
lts
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 200
Conclusion
DFSS -DFSS -
– Starts with the Voice of the Customer.
– Considers ‘options’ for providing a solution to the Customerneed.
– Makes sure that the solution is represented in ‘engineeringterms’ as a model, equation, or simulation, beforehardware is built.
– Allows analysis of multiple trade-offs to ensure customerrequirements are met while minimizing cost.
– Analyzes capabilities and potential for ‘variation’ in themanufacturing and operational scenarios.
– Seeks to eliminate Launch and Deployment problemsbefore they occur.
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 201
This is“Design for Six Sigma”!
Identify CriticalRequirements
Build Models
Optimizethe Design
Verify &Validate
Voice of the Customer
Design that best meetsall requirements
DFSSDFSS
ConceptInitiationConceptConceptInitiationInitiation
PrototypePrototypePrototype
PilotPilotPilot
LaunchLaunchLaunch
PlanningPlanningPlanning
AnalyzeVariability
AllocateVariability
OptimizeVariability
PostLaunchPostPost
LaunchLaunch Product ModelProduct Model(equation, (equation, simulation,simulation,workbook,workbook,
hardware, etc.)hardware, etc.)
Y
C
E
A
B
D
ULLL T
PNCPNC
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 202
Wrap-Up and Discussion
2002 Lear Corporation. Unpublished Material © 2002 Statistical Design Institute, LLC. All Rights Reserved. Module 1.1 Intro to DFSS - 203