Design For Robustness - PowerPoint PPT Presentation

Loading...

PPT – Design For Robustness PowerPoint presentation | free to download - id: 3bdd1b-ZTViM



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Design For Robustness

Description:

Design For Robustness MPD575 DFX Jonathan Weaver Development History Originally developed by Cohort 1 students: Steve Borkes, Larry Liotino, Dennis Person (Fall 2000 ... – PowerPoint PPT presentation

Number of Views:603
Avg rating:3.0/5.0
Slides: 139
Provided by: weaverjmF
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Design For Robustness


1
Design For Robustness
  • MPD575 DFX
  • Jonathan Weaver

2
Development History
  • Originally developed by Cohort 1 students Steve
    Borkes, Larry Liotino, Dennis Person (Fall 2000)
  • Revised by Cohort 1 students Tom Boettcher, Al
    Figlioli, and John Rinke (Fall 2000)

3
(No Transcript)
4
Design for Robustness (DFR)
  • Introduction to DFR
  • DFR Timing
  • DFR Key Principles and Procedures
  • Robustness Case Studies
  • DFR Software
  • Conclusions
  • References

5
Introductionto DFR
6
Elements of Quality
7
Introduction to DFR
  • DFR is a key enabler for
  • Design for Quality
  • Design for Reliability
  • Robust System Design
  • Design for Six Sigma

8
Robustness Defined
  • Robustness is
  • The capability of a product or process to perform
    its intended function consistently in the
    presence of noise during its expected life
  • That is, the performance of the product or
    process is insensitive to sources of variability

9
Robustness Defined
  • Sources of variability, or noises, can be
    grouped in the following categories
  • Manufacturing Variation
  • Wear-out/Fatigue
  • Customer Usage
  • External Environment
  • System Interactions

10
Robustness Defined
  • The premise of robust design is that a product
    that performs its function "on target" will
    generate the smallest loss to the customer and
    the producer (e.g., cost to repair or replace,
    cost of customer dissatisfaction). As a product
    deviates from its target response or expected
    output, cost will begin to accrue in one way or
    another
  • Customers tend to be more satisfied with their
    purchases if the product is robust and reliable

11
Robustness Defined
  • Robustness requires a shift from "What is wrong
    and how do we fix it?" way of thinking to "What
    is the intent and how do we maximize it?" Dr.
    Genichi Taguchi
  • Instead of trying to control, compensate, or
    eliminate usage variations, robustness attempts
    to make the system insensitive to usage
    variations. Dr. Genichi Taguchi

12
Introduction to DFR
  • Designing for Robustness is not newit is just
    good engineering
  • The principles of robustness are not new
    concepts, but the DFR process facilitates the
    engineer in applying an engineering process to
    obtain a robust product and a means for
    quantifying the benefits

13
Why To Apply DFR?
  • Develop products or processes that are robust
    against different sources of variation
  • Promote proactive problem solving techniques
  • Avoid build-test-fix iterations, especially
    when combined with CAE methods
  • Achieving proactive robust designs at the
    subsystem level will help to better achieve
    robust system level performance once integrated

14
Why To Apply DFR?
  • Address new technology or existing technology in
    new environments
  • Address a component or sub-system that is part of
    a complex system
  • Gain an understanding of a systems sensitivity
    to noises
  • Meet the global challenge of producing products
    that perform well in all markets

15
Why To Apply DFR?
  • Customer Benefits
  • Increased customer satisfaction
  • Added value for the customer
  • High product quality and reliability
  • Long term durability.
  • Reduced cost of ownership
  • Lower maintenance cost
  • Higher residual values
  • Corporate Benefits
  • Lower warranty
  • Lower cost components
  • Lower manufacturing costs
  • Higher productivity
  • Higher profit
  • Improved customer loyalty
  • Low cost producer
  • Increased Shareholder Value

16
Introduction to DFR
  • DFR is not intended only for designing components
    or products
  • The techniques can be used for other systems,
    such as manufacturing processes, organizational
    structures, and product distribution networks

17
Terminology Concepts
  • Reliability the probability that a component or
    system will achieve a certain defined life
  • Concept Design involves selecting the
    appropriate level of technology required to
    provide a particular function.
  • Parameter Design involves using the DOE method
    to study the effect of control factors on a
    response in the presence of noise.
  • Tolerance Design shares many of the same
    principles as Parameter Design the focus is on
    determining different tolerances of control
    factors necessary for a product to deliver its
    intended function cost effectively.

18
Terminology Concepts
  • A p-diagram is a common way to represent a
    system

19
Parameter(P)-Diagram
Parameters which influence system
variability, but are difficult, expensive or
impossible to control Noise Factors
System Function
Input Signal
Output Response
Error States
Control Factors Parameters whose nominal values
can be adjusted by the engineer, ideally with
minimal impact on cost
20
Terminology Concepts
  • Ideal Function is the primary intended function
    of the design (often energy related, because
    mechanical engineering is about making things
    move or stopping things from moving).
  • Signal Factor is the energy which is put into
    the engineering system (either by the customer,
    or by a neighboring system) to make it work.

21
Terminology Concepts
  • Control Factors are features of the design that
    can be changed by the engineer (e.g. dimensions,
    shapes, materials, positions, locations etc).
  • Noise Factors are sources of disturbing
    influences that can disrupt ideal function ,
    causing error states which lead to quality
    problems.
  • Error state is an undesirable output of the
    engineering system (we can also call these
    failure modes), characterized by - variation in
    ideal function (soft failure), - degradation in
    ideal function (soft failure), - or loss of ideal
    function (hard failure).

22
DFR Timing
23
When To Apply DFR?
  • DFR should be conducted throughout the entire PD
    cycle.
  • DFR should be initiated EARLY to
  • Gain the maximum benefit from the entire process.
  • Define intended functions, requirements, and
    noises to support the cascade process.

24
DFR and the System V
Phase I
Phase I
Phase III
Phase III
Define
Define
Total System
Total System
Customer/System Requirements
Customer System Requirements
Confirmation
Confirmation
Integrate and
Integrate and
Verify Designs
Verify Designs
Cascade and
Cascade and
Balance Targets
Balance Targets
ROBUST
ROBUST
DESIGNS
DESIGNS
Phase II
Phase II
25
Principles of DFR
26
Principles of DFR
  • DFR IS ABOUT...
  • Engineering in ideal function and avoiding
    failure modes.
  • Measuring technical performance with objective
    measures.
  • Anticipating the effects of noise factors
    up-front in the design planning process.
  • Developing a Noise Factor Management Strategy.
  • Lowest possible cost solutions.

27
Principles of DFR
  • DFR IS NOT ABOUT...
  • Measuring predicting symptoms of poor quality
    (e.g. TGW, repair cost per unit (CPU), R/1000).
  • Vehicle Evaluation Ratings (VER).
  • Ignoring noise factors until their effects are
    highlighted in the DVP or worse, with the
    customer.
  • Just running orthogonal array experiments (The
    Lets do a DOE/Taguchi mentality).
  • Adding to design cost.

28
Principles of DFR
  • Conceptual Blockbusting
  • Typical Thinking Errors
  • A Guide to Better Thinking
  • Traditional Philosophy of Quality
  • Meet the Target
  • Quality Loss Function Philosophy
  • Illustrations of DFR Benefits

29
Conceptual Blockbusting
  • Typical Thinking Errors
  • Partialism
  • Time Scale
  • Polarization

30
Conceptual Blockbusting Typical Thinking Error
  • Partialism
  • Focusing on only a limited aspect of the data and
    constructing a logical argument to prove a
    particular point of view.
  • Politicians are expert at this, particularly when
    they use one economic indicator at the expense of
    others
  • Component optimization which leads to system
    sub-optimization

31
Conceptual Blockbusting Typical Thinking Error
  • Time Scale
  • Looking only at the short term implications
    without considering a longer time scale

32
Conceptual Blockbusting Typical Thinking Error
  • Polarization
  • Oversimplifying the choices by considering
    everything as black or white
  • If you cant control a particular factor then you
    cant control any factors. Or, if you compromise
    on one factor youll have to compromise on all
    factors
  • Sometime the best solution is a shade of gray.
    Be careful of being forced into an either/or
    decision.

33
Conceptual Blockbusting
  • A Guide to Better Thinking
  • Identify your current paradigm(Avoid jumping
    ahead based on what you think you know)
  • Avoid closure(Prevent premature closure)
  • Change your thinking
  • Differentiate

34
Conceptual Blockbusting A Guide to Better
Thinking
  • Change your thinking by
  • Moving your entry point.
  • Reversing your usual assumptions.
  • Magnifying, minimizing, substituting or
    exaggeration.
  • Changing the context.

35
Conceptual Blockbusting A Guide to Better
Thinking
  • Differentiate
  • Not every new idea is a good idea
  • All good ideas need to be logical in hindsight
  • The skill lies in knowing when to draw a line,
    and when to push the line!
  • Determine whats radical and whats ridiculous

36
Traditional Philosophy of Quality
  • Target value is ideal
  • Due to manufacturing variation, spec limits are
    established
  • Tradition thinking goal post mentality
  • Bad assumption financial costs are only incurred
    when scrap is made or if an out-of-spec product
    reaches the customer

37
Meet the Target
  • The farther away from the target value, the lower
    the quality and the less satisfied the customer.
  • The goal is to meet the target, not the
    specification limits.

MEET THE TARGET
USL
LSL
Target Value
38
Quality Loss Function Philosophy
  • As performance varies from target, financial
    losses due to customer dissatisfaction increases.
  • Total loss may include costs related to assembly
    difficulty, degraded performance, customer
    dissatisfaction, warranty cost, etc.
  • The closer to the target value, the more reliable
    the product will be.

Poor
Poor
Fair
Fair
Loss ()
Good
Good
Best
y Response
Target Value
39
Quality Loss Function Philosophy
  • If this distribution were produced, the producer
    and the customer would experience some loss

LSL
USL
Target Value
40
Quality Loss Function Philosophy
  • Producing this distribution, which has less
    variance from target, the loss would be lower.
    This represents a nominal-is-best loss function.
  • Other loss functions include smaller-is-better
    and larger-is-better.

LSL
USL
Target Value
41
Principles of DFRIllustration of DFR Benefits
  • To appreciate the fundamental concept that robust
    design is trying to accomplish, consider the
    following simple illustration
  • Let x be a control factor
  • Let n be a noise factor

42
Principles of DFRIllustration of DFR Benefits
  • The relationship between these factors and the
    response z (product performance) is
  • z f(n,x)
  • Variability due to the uncontrollable noise
    factor n generates variability in product
    performance z.
  • Assume that as the product performance deviates
    from the target (T), the loss is represented by a
    quadratic function
  • L(z) k (z - T)2 k (f(n,x) T)2
  • where k is the quality loss coefficient.

43
Principles of DFRIllustration of DFR Benefits
  • Assuming a target (T0) and that z follows a
    standard normal distribution, the effect of the
    control factor setting on the loss function is
    illustrated.
  • By selecting a proper setting of control factor
    x, the loss function with respect to the noise
    factor can be "flattened." The effect of noise
    is minimized by a flatter loss function. In the
    illustration, x 1 has a flatter loss function
    than that of x 2.

z
Performance Distribution
Performance, z
44
Principles of DFRIllustration of DFR Benefits
  • In terms of product performance z, the effect of
    choosing the proper setting of control factor x
    results in different robustness levels of z.
  • The performance variability when x 1 is less
    than the performance variability when x 2.
    That is, the design is more robust when x 1
    than when x 2.

45
Principles of DFRIllustration of DFR Benefits
  • In terms of product performance, robustness can
    be measured by
  • Expected loss EL(z)
  • Deviation from performance target, mz - T, and
    performance variability sz
  • A combined measure of the performance mean and
    standard deviation is called the Signal-to-Noise
    (S/N) Ratio.

46
Principles of DFRIllustration of DFR Benefits
  • The objective is to find x that minimizes g(x),
    where g(x) represents a performance loss
    function.
  • Without considering variability (noise), design
    point A is a better choice than design point B.
  • However, when the same amount of variability is
    considered, B is a better choice than A because
    the performance loss variation of B is much
    smaller than that of A.

Performance Loss
47
Principles of DFRIllustration of DFR Benefits
Stress and Strength Interference
  • The region of interference between stress and
    strength defines the probability of failure. This
    is the region of interest in robustness analysis.
  • Same (or larger) design margin may have larger
    probability of failure depending upon the
    variabilities of the loading and strength.

Design Margin
Stress
Strength
Design Margin
Stress
Strength
48
Robustness Implementation
  • In principle, robustness should focus on the
    intended function rather than error states. The
    reasoning is that improving the intended function
    will reduce the error states.
  • However, eliminating an error state does not
    guarantee improvement of the intended function,
    because the elimination of one error state may
    create other error states.

49
Robustness Implementation
  • In any given problem, there is not a single
    robust solution, but a family of solutions that
    depend on both subjective and objective
    assessment
  • Shift from measuring the symptoms of poor quality
    to measuring the transformation of energy Dr.
    Genichi Taguchi
  • This is facilitated using a tool called the
    Parameter(P)-diagram

50
Parameter(P)-Diagram
Parameters which influence system
variability, but are difficult, expensive or
impossible to control Noise Factors
System Function
Input Signal
Output Response
Error States
Control Factors Parameters whose nominal values
can be adjusted by the engineer, ideally with
minimal impact on cost
51
Types of Noise Factors(That Disrupt Ideal
Function)
1) Piece-to-piece variation of part dimensions
Inner Noises
2) Changes over time/mileage in dimensions or
strength (such as wear out or fatigue)
Outer Noises
Conditions of use
3) Customer usage and duty cycle
Operating Environment
4) External (climatic and road conditions)
NOTE effects of noise factors can sometimes be
represented or captured by others
5) Internal , due to a) error states being
received as noise factors from
neighboring sub-systems. b) unresolved design
issues related to neighboring
sub-systems.
52
Fuel Delivery SystemInterface Diagram
Back pressure
Injector flow
Valve
Fuel Pump
Rail pressure
Flow
Voltage
EEC
Measurement
Manifold pressure from accelerator pedal
Other inputs
53
Fuel Delivery SystemIdeal Function
  • The ideal function is a description of how the
    system should perform its intended function
    perfectly

Translation of customers perceived result to
measurable engineering terms
Intent (What customer wants)
Output Response
FUEL PUMP
Signal
Translation of customers intent to measurable
engineering terms
Perceived Result (What customer gets)
  • Example fuel pump must provide sufficient liquid
    fuel to the injectors as requested by engine
    under all conditions

54
Fuel Delivery SystemSignal Factor
  • Energy which is put into the engineering system
    (either by the customer, or by a neighboring
    system) to make it work.
  • Other examples not related to fuel systems
  • Temperature setting on car heater
  • Steering wheel angle
  • Force on brake pedal

55
Fuel Delivery SystemInput Signal Vs. Output
Response
Signal Factor
Flow rate
FUEL PUMP
  • VI
  • P
  • V Voltage
  • I Current
  • P Back pressure

56
Fuel Delivery SystemNoise Factors
  • Factors which influence system variability that
    an engineer decides not to control because they
    are difficult, expensive or impossible to
    control.
  • Noise factors disrupt system function and cause
    error states, which will degrade the intended
    functional performance and cause failures under
    customer usage, operating, or environmental
    conditions.

57
Fuel Delivery SystemNoise Factors
  • Noise Factors
  • Fuel volatility
  • Fuel temperature
  • Tank vapor pressure

FUEL PUMP
  • VI/P
  • V Voltage
  • I Current
  • P Back pressure

Flow rate
Error States
58
Fuel Delivery SystemControl Factors
  • Parameters whose nominal values can be adjusted
    by the engineer, ideally with minimal impact on
    cost.
  • The factors are usually chosen based on cost,
    resource, and technology considerations.
  • Examples
  • Material type
  • Dimension

59
Fuel Delivery SystemControl Factors
  • Noise Factors
  • Fuel volatility
  • Fuel temperature
  • Tank vapor pressure

FUEL PUMP
Flow rate
  • VI/P
  • V Voltage
  • I Current
  • P Back pressure

Control Factors A. Fuel Pump Type B. Assembly
Type C. Mounting Angle (from vertical) D. Rated
Pump Flow (lph) E. Modulation Frequency (kHz)
60
Fuel Delivery SystemP-Diagram
  • Noise Factors
  • Fuel volatility
  • Fuel temperature
  • Tank vapor pressure
  • VI/P
  • V Voltage
  • I Current
  • P Back pressure

FUEL PUMP
Flow rate
Error States
Control Factors A. Fuel Pump Type B. Assembly
Type C. Mounting Angle (from vertical) D. Rated
Pump Flow (lph) E. Modulation Frequency (kHz)
61
Parameter Design
62
Parameter Design
  • Robust designs can be achieved through "brute
    force" techniques of added design margin or
    tighter tolerances
  • Better yet is through "intelligent design" by
    understanding which product and process design
    parameters are critical to the achievement of a
    performance characteristic and what are the
    optimum values to both achieve the performance
    characteristic and minimize variation.

63
Parameter Design
  • Parameter Design is one commonly used design
    method in which the optimum parameters of key
    product and process characteristics are
    determined such that the product is least
    sensitive to "noise" factors.

64
Parameter Design
  • Achievement of robust systems require optimal
    selection of design parameters and their
    settings. This is accomplished using
    statistically designed experiments (DOE).
  • DOE establishes the settings of the design
    parameters that desensitize the system to
    variation due to noise.
  • Application of the DOE techniques results in
    products of superior quality and can help achieve
    significant cost reductions.
  • Reducing variability in key characteristics
    always results in a relative benefit.

65
Parameter Design
  • Three Step Optimization
  • Reduce Variability by identifying the factors
    which interact with noise and use them to reduce
    variability.
  • Adjust the Mean to Target by identifying the
    factors with location effects, that do not
    interact with noise and use these to adjust the
    mean.
  • Select the low cost settings for parameters with
    minimal effect on the response.

66
Parameter Design
Reduce Variation
Performance Distribution after reducing variation
Performance Distribution for current design
Note Example shows a nominal-is-best static
quality characteristic
Adjust the Mean to Target
Performance Distribution after adjusting mean
Target
67
Fuel Delivery SystemOptimal Design Parameters
and Confirmation
o
o
68
Fuel Delivery SystemVehicle Evaluation in
Corporate Test
Original
Robust Design
P
P
Fuel requested by engine
Fuel delivery by pump
25 of test procedure
Optimum Parameter Design
Best Guess Design by Team
69
Ways To Measure Robustness
70
Ways To Measure Robustness
1) Classical Measure S/N Ratio (NTB1 shown)
Less Robust
More Robust
Due to Noise
Ideal
Ideal
Due to Noise
Due to Noise
Ideal Function
Ideal Function
Due to Noise
Signal
Signal
71
Ways To Measure Robustness
  • S Energy producing the intended results
  • N Energy wasted on unintended results
  • Measures performance in the presence of noise
  • Evaluates the interaction between control and
    noise factors.
  • Always maximize the S/N Ratio
  • S/N , Quality , Loss


72
Ways To Measure Robustness
  • Three Types of Static Robustness Metrics
  • Smaller-the-better S/N Ratio
  • Example Wear, Leakage
  • Larger-the-better S/N Ratio
  • Example Fuel Economy, Stock Market Return
  • Nominal-the-best S/N Ratio
  • Type 1 Mean and standard deviation scale
    together
  • Example Glass Run Thickness
  • Type 2 Standard deviation is independent of the
    mean
  • Example machining errors

73
Ways To Measure Robustness
2) Variation in Output
Less Robust
More Robust
Noises
Noises
Ideal
Ideal
74
Ways To Measure Robustness
3) Degradation of Function
Less Robust
More Robust
Ideal
Ideal
Functional Attribute
Functional Attribute
Degraded
Degraded
Target
Target
75
Ways To Measure Robustness
4) Time to Failure (hard failures)
Robust Design
Failure Rate
Non-Robust Design
Failure Percent
Ideal
Wearout
Expected Life
Infant Mortality
Less Robust
More Robust
Time/Mileage
Time/Mileage
76
Strategies forImproving Robustness
77
Strategies for Improving Robustness
  • Change the design concept
  • Change the technology
  • Apply principles of Axiomatic Design
  • Make basic design assumptions insensitive to the
    noises
  • Through Parameter Design (may need the technique
    of DOE)
  • By beefing up design (upgrading specifications)
  • Redundancy (not usually possible in automotive
    engineering)

78
Strategies for Improving Robustness
  • Reduce or remove the noise factor(s)
  • e.g. reducing variation of critical dimensions
    in the manufacturing process (may need the
    technique of DOE), upgrade materials, control the
    environment
  • Insert a compensation device
  • e.g. heat shield, mass damper, use feedback
    control or feed forward control
  • Send the error state/noise somewhere else where
    it will do less harm (disguise the effect)
  • e.g. smoked headlamp glass to hide H20 ingress

79
Robust SystemDesign
80
Robust System Design Steps
  • Step 1 Summary
  • Narrow system down to critical area of
    understanding
  • Form cross function team of experts
  • Clearly define project objective
  • Define roles and responsibilities to team members
  • Translate customer intent non-technical terms
    into technical terms
  • Identify product quality issues
  • Isolate the boundary conditions and describe the
    system in terms of its inputs and outputs

Step 1 Identify Project and Team
81
Robust System Design Steps
  • Step 2 Summary
  • Select a response function(s)
  • Select a signal parameter(s)
  • Determine if problem is static or dynamic
    parameter
  • Determine the S/N function. See section Step2
    e.g. Nominal-the-best, etc.

Step 2 Formulate Engineered System Ideal
Function / Quality Characteristic(s)
82
Robust System Design Steps
  • Step 3 Summary
  • Select control factor(s)
  • Rank control factors
  • Select noise factors(s)

Step 3 Formulate Engineered System Parameters
83
Robust System Design Steps
  • Step 4 Summary
  • Determine control factor levels
  • Determine if there are any interactions between
    control factors
  • Calculate the DOF
  • Select the appropriate Orthogonal Array

Step 4 Assign Control Factors to Inner Array
84
Robust System Design Steps
  • Step 5 Summary
  • Determine noise factors and levels
  • Determine noise strategy
  • Surrogate Noise Strategy
  • Compound Noise
  • Treat Noise Individually
  • Establish outer noise array matrix

Step 5 Assign Noise Factors to Outer Array
85
Robust System Design Steps
  • Step 6 Summary
  • Cross functional team will need to develop a
    step-by-step plan to carry out the logistical
    activities necessary for successful completion of
    the data collection phase of the optimization
    experiment
  • Identify Facility Constraints
  • Determine logistical/run order

Step 6 Conduct Experiment and Collect Data
86
Robust System Design Steps
  • Step 7 Summary
  • Calculate the Following Values
  • Mean, Variance, S/N Ratio, ANOVA, Factor Effects
  • Interpret the Results
  • Select Factor Levels providing the largest
    improvement in S/N
  • Use the mean effect values to predict the new
    mean of the improved system
  • If needed, select a value for the scaling factor

Step 7 Analyze Data and Select Optimal Design
87
Robust System Design Steps
  • Step 8 Summary
  • Make prediction in response for optimized system
  • Confirm prediction by running test on optimized
    system

Step 8 Predict and Confirm
88
Benefits to Robust Design
  • Testing Efficiency
  • Properties of the Orthogonal Array result in more
    accurate information than one-factor-at-a-time
    experimentation
  • OA's better replicate the dynamic nature of "real
    world conditions by pair-wise balancing factors
  • The number of tests required will be much less
    than full-factorial testing
  • Predictive Model
  • Useful for evaluating design concepts without
    committing resources to building prototypes

89
Benefits to Robust Design
  • The RD technique can increase quality (or value)
    in a design in a quantifiable manner. It can also
    identify areas in which there are over-designs
    which may be incurred at unnecessary expense.
  • Robust Design can be used proactively to avoid
    the build-test-fix scenario.
  • By applying Robust Design to system components,
    the system quality after integration will
    typically be improved, but not necessarily
    optimized.
  • A properly designed Robust Design experiment will
    allow the exploration of factor interactions.

90
Benefits to Robust Design
  • Through the Robust Design process, knowledge of
    individual factors with the design will be
    captured. This supports the book-shelving of
    the underlying principles of the overall design.
  • Robust design can be applied very early in the
    design cycle, and the earlier it is applied the
    more impact it can have on the final design.

91
Limitations to Robust Design
  • The robust design technique makes it difficult to
    achieve results with "partial experiments.
  • Significant effort must be spent on
    understanding the system.

92
Limitations to Robust Design
  • Some of the noise factors need to be controllable
    for experimentation purposes.
  • The RD technique requires a linear factor
    separable relationship, which relates the factors
    in the design that can be controlled with the
    intended response of the system.

93
DFR Case Studies
  • Automotive Window System
  • Door Sealing System
  • Body Side Molding

94
DFR Case Studies
  • Automotive Window System
  • Door Sealing System
  • Body Side Molding

95
Automotive Window System
Concerns high window effort, handle broken,
high TGW and warranty
96
Automotive Window SystemEngineered System
P-Diagram
  • Noise Factors
  • Glass Run Thickness Variability
  • Customer Usage / Duty Cycle
  • Wear or Cycles
  • Temperature / Environment / System Interfaces

Signal M Motor Voltage
Response y Velocity or Torque
Door Glass Lifting System
  • Control Factors
  • Clearance of Glass to A-Pillar Channel
  • Clearance of Glass to B-Pillar Channel
  • Glass Run Shape
  • Glass Run Thickness
  • Regulator Angle

97
Automotive Window System Improvement Actions
Use Design of Experiment (DOE) and
Signal-to-Noise Ratio to Optimize Design for
Improvement. Energy producing the
intended result (moving window) S/N
Energy wasted on unintended results (friction or
drag)
  • Actions
  • Move A/Pillar bottom 2 mm forward
  • Reduce variation in glass run thickness

98
Automotive Window SystemResults - Robust Design
Better
25
24
S/N Ratio
23
6
22
21
20
19
18
17
Old Design
Robust Design
Window Effort Torque in-lbs
25
20
20
15
25
15
10
5
Better
0
Old Design
Robust Design
99
Automotive Window System Robustness Reduces
Warranty Costs
150 TGWs
150
Customer Complaints TGW/1000
80
100
30 TGWs
50
0
Old Design
Robust Design
100
80
Warranty First Month-In-Service ()
50
60
40
20
0
Old Design
Robust Design
100
DFR Case Studies
  • Automotive Window System
  • Door Sealing System
  • Body Side Molding

101
Door Sealing SystemRobustness Assessment
  • From sealing quality history, wind noise is the
    number 1 TGW issue for the door sealing system.
  • The objective
  • Nominal-is-best sealing force per length
    preliminary target 16 N / 200 mm seal
  • Larger-is-better pressure differential (inside
    vs. outside at least 0.006 N/mm2) that the
    primary door seal can withstand without loosing
    contact.

102
Door Sealing SystemP-Diagram
SOURCES OF NOISE Piece-to-piece variation gap
variability Changes in dimension due to wear
fatigue ageing of seals Customer Usage/Duty
cycle stepping on seal, etc. External
Environment (climate, roads etc.) dust Internal
Environment due to interfaces/neighboring
systems door variability
IDEAL FUNTION Bulb pressure TARGET Pressure
differential (larger-is-better) 0.0006 N/mm
2 Force (nominal-is-best) 16 N / 200 mm seal
Input Signal Static load
Primary Door Seal
CONTROL FACTORS profile shape (Hu. Hl,
W) material (M) sheet metal contact surface (S)

ERROR STATE Loss of contact (potential energy)
ERROR STATE Noise transmit (kinetic energy)
103
Door Sealing SystemControl Factors for CAE
testing
104
Door Sealing SystemNoise Factors for CAE Testing
105
Door Sealing SystemTest Method and Matrix
  • Experimental work done by CAE analysis.
  • Natural variability excluded, S/N ratios could
    not be established out of the CAE data
  • An inner L8 (2(5-2)) control factor array was
    combined with an outer L2 noise factor array.
  • The empty columns e1 and e2 contain interactions.
  • For each CAE run the closing travel was increased
    in incremental steps of 0.4 mm, and the response
    force (exerted per seal length) was recorded for
    the minimum seal gap and maximum seal gap.
  • Pressure differential was increased in
    incremental steps of 0.0001 N/mm2, response
    force recorded until it became zero or negative
    yielding the maximum pressure differential
    (inside to outside the primary seal) that the
    respective setting can withstand.

106
Door Sealing SystemTest Matrix
107
Door Sealing SystemConclusion
  • The key outcome of the study is that the current
    design is the best within the existing
    alternatives no improvement of the overall
    situation could be reached by alternative
    settings.
  • Sealing material, wall thickness, contact surface
    have got significant effects.
  • However, due to a lack of accurate CAE analysis
    tools, a final assessment of the door closing
    efforts has to be made through hardware testing.
  • The results show that the current setting could
    not be significantly optimized further within the
    given parameter space, if both responses have to
    be considered simultaneously.

108
DFR Case Studies
  • Automotive Window System
  • Door Sealing System
  • Body Side Molding

109
Body Side Molding Study
  • Background
  • In 1994 Ford Motor Company had experienced high
    warranty cost and customer dissatisfaction with
    body side moldings becoming detached on some of
    its vehicles.
  • The Robust Design Methodology was viewed as an
    ideal tool for providing a solution to the
    problem.
  • The case study focused on initial quality
    application (initial installation), since this
    had been determined, through several extensive
    quality and durability tests over time, to be the
    root cause of the majority of warranty failures.
  • These prior tests also concluded that if the part
    were applied to specification it would be capable
    of 10 years/150,000 miles performance.

110
Body Side Molding Study
  • Background (contd)
  • Most of the body side moldings, and the ones
    studied in this case are made of a flexible
    polymer with metal inserts for foam attaching
    tape adhesion. This is in turn applied to the
    vehicle during final trim of the painted body
    side. When a molding comes loose, it is normally
    in the initial months. This is an " infant
    mortality" failure that focuses our team on the
    installation method.

111
Body Side Molding Study
  • Identifying a project candidate
  • High warranty cost and customer dissatisfaction
    were the main motivation for this project. This
    information was obtained through the Ford
    internal quality indicator systems and five Truck
    Assembly Plant reports.
  • The cross-functional team was formed for this
    project, including members from Body Engineering
    (BE) Quality, BE Reliability Technology, supplier
    Quality Engineering (SQE), Alpha, BAO
    Processing, Suppliers, Materials Engineering,
    Testing (Detroit Lab Testing) and personnel from
    5 Truck Assembly Plants.

112
Body Side Molding Study
  • Building the team
  • The objective was to analyze the current body
    side molding assembly process and develop a
    capable robust process which would be able to
    produce reliable side moldings.
  • Team members visited assembly plants and
    suppliers facilities for extensive research about
    the Competition's and Ford's current body side
    molding assembly process and failure/problem
    investigation.

113
Body Side Molding Study
  • The objective of this project was to focus on XYZ
    vehicle's body side molding assembly process and
    develop a capable and robust process which would
    be able to produce reliable adhering body side
    moldings.
  • Four different adhesives (tapes) were
    benchmarked.
  • Tape 1 4248 Neoprene (AR7/AR7), current adhesive
    used for all Ford vehicle's nameplates.
  • Tape 2 4220 Acrylic (AR7/DS4), currently used in
    Ford passenger cars body side Molding.
  • Tape 3 5380 Acrylic(RC6/DS4), currently used in
    Ford passenger cars body side Molding.
  • Tape 4 Z545 Urethane, currently used by other
    OEM's. Considered as the benchmark adhesive based
    on reliability, quality, cost savings, and test
    data from suppliers.

114
Body Side Molding Study
  • The adhesives were selected for this parameter
    design case study based on the following data
    sources of performance history
  • Inspection studies,
  • Quality data (3MIS and 48MIS)
  • Durability data (1-10years)
  • Montreal surveys
  • Florida surveys
  • First run capability at Ford, Mazda, GM,
    Chryslers' assembly plants.
  • This assumes that the data reflects the relative
    performance of the tape, considering all have
    gone through the same installation processes.

115
Body Side Molding StudyResponse Factors
  • Six response factors were established by the
    engineering team. Note More than one response
    factor can be measured
  • Total removal energy
  • Break-away peak energy (energy required when
    adhesive fails)
  • Break-away peel load (maximum load)
  • Continuous peel energy
  • Continuous peel load
  • Area of tape not wet-out (adhesion did not take
    place)

116
Body Side Molding Study
  • Control factors selected for the body side
    molding experiment include
  • Alcohol wipe
  • Body molding temperature
  • S/Metal Temperature
  • Roller Pressure
  • Stationary Roller Pressure
  • Dwell time
  • S/Metal Water/Residue
  • A single noise factor with two levels was chosen
  • Sheet metal variation (hem flange roll off. This
    is considered a control factor rather than a
    noise, because we can choose to not put the
    molding out to the roll-off region. This can
    actually be treated as an imbedded noise factor
    in the inner array)
  • Level 1 was a flat panel
  • Level 2 was "roll off" test panel (3mm)

117
Body Side Molding StudyControl Factors and Levels
  • The following control factor levels were
    established

Based on the control factor table, an appropriate
orthogonal array was determined (L18).
118
Body Side Molding StudyNoise Strategy
  • A compound noise strategy was selected to combine
    the noises so one factor could be used and
    varied. This can only be done because the
    directionality of the effect of each noise were
    known. The compounding dramatically reduces the
    number of tests required.
  • In this experiment a surrogate noise factor was
    chosen, sheet metal variation (hem flange
    roll-off). There were two levels to this noise
    factor, flat test panel and roll-off test panel.

Roll-off
119
Body Side Molding Study
  • L18 Orthogonal Array

120
Body Side Molding Study
  • Collect data.
  • Perform ANOVA.
  • Analyze S/N ratio effects.
  • Choose appropriate settings for the control
    factors.

Factor S/N Effect Plot
121
Body Side Molding Study
122
Body Side Molding StudyConclusion
  • The entire study was completed in 8 months. The
    results provided engineering with a final
    adhesive candidate which reduced warranty from 2
    R/1000 to 0.8 R/1000, and had a warranty cost
    saved of 32,700 / year.
  • The big bonus, however, showed up in the
    resulting material change of the foam tape
    material that saved Ford Motor Company over
    1,000,000 annually (0.25 / vehicle time 5
    million vehicles a year).

123
Conclusions
124
Robustness Rules of Engagement
  • Concentrate on Ideal Function and establish a way
    to measure it do not use symptoms of poor
    quality
  • Identify sources of the five types of noises an
    expected magnitude remember system interactions
  • Concentrate on the effects of the noises maybe
    one noise can be used to represent others
  • Understand how error states and noise factors
    cross system interfaces and boundaries establish
    contracts with neighboring systems

125
Robustness Rules of Engagement
  • Develop a noise factor management strategy
    Removing the noise might be easier than becoming
    robust to it. The laws of physics are strict.
  • Work out how to include remaining Noise Factors
    in all tests in the DVP.
  • Plan a robustness assessment of current design to
    compare against ideal performance.
  • Where robustness improvement strategy is obvious
    from knowledge of physics, DO IT!
  • Where robustness improvement is not obvious, plan
    parameter design studies (using DOE if necessary)
    to discover the improvement.

126
DFR Checklist
  • Has the desired level of robustness been achieved
    for all prioritized systems and/or requirements?
  • Have all the robustness models (P-diagram)/
    Engineering System been identified?
  • noise factors
  • control factors
  • error states
  • ideal function
  • robustness / reliability metrics

127
DFR Checklist (contd)
  • Have all the analysis tools been simultaneously
    updated with the latest data?
  • FMEA
  • VDS, SDS, CDS
  • Real World Usage Profiles
  • Robustness Models (P-diagrams)

128
Institutionalizing DFR
  • SDS should include
  • An interface diagram.
  • All of the results from the p-diagram development
    including
  • The p-diagram.
  • All information on the noise factors.
  • All information on ideal function, input signal,
    response and control factors.
  • Specification dimensions over useful life of
    product.
  • Specification dimensions correlated to customer
    satisfaction

129
Institutionalizing DFR
  • FMEA should include
  • Sensitivity to noise factors
  • The principles of robustness as part of the
    development process
  • ADVP should include
  • All noise factors from developed P-diagram
  • Make use of the robustness tools available

130
Institutionalizing DFR
  • Bookshelf all robust design results for future
    reference and use
  • Practice Conceptual Blockbusting

131
Challenges to DFR
  • There are many challenges in applying reliability
    and robust design methods to analytic models.
    These include
  • In practice, many analytical (CAE) models are
    focused on the error states (NVH, fatigue, etc.).
    It is important to be cautious that reducing one
    error state does not generate other error states.
  • Many CAE models have limited capability to
    represent real-world noise therefore, surrogate
    noise based on engineering knowledge will be
    required.

132
Challenges to DFR (contd)
  • Precise reliability estimates require precise
    knowledge of statistical distributions of noise
    factors (As a contrast, comparative reliability
    assessments and robust design require only
    approximate knowledge of statistical
    distributions.)
  • When statistical distributions of noise factors
    are assumed known, the statistical description of
    noise factors needs to be fully specified in the
    analysis without introducing an unnecessary
    computational burden.

133
Challenges to DFR (contd)
  • Many CAE models are computationally expensive
    (both preparation time to set up the model and
    computing time)
  • It is often desirable to study a large number of
    design variables within a large design space in
    this situation, nonlinear relationships between
    input (design variables) and output (performance)
    will be common

134
Challenges to DFR (contd)
  • Many CAE models focus on error states (e.g.,
    fatigue, vibration, noise) therefore, a
    multi-objective optimization is often needed.
  • In early product development, when the impact of
    robust design can be greatest, design objectives
    and constraints are still imprecise.

135
References
  • FRG, Module 18, Ford Automotive Operations
    Quality, 1999.
  • Robustness Thinking Robust Engineering Design,
    Ford Motor Company Quality Office, Quality
    Reliability Implementation Group, August 2000
    Edition 7.01.
  • FAO Reliability Guide PD Useful Life Reliability
    Commitment Edition 4, Ford Automotive Operations
    Quality, 1996,1997,1998.
  • Ford Design Institute, The Robustness Imperative,
    Ford Motor Company, 1993

136
References
  • (RPDP) Robust Product Development Process, One of
    Six Powertrain Breakthrough Initiatives Ford
    Motor Company, Release 1.0, June 1996
  • Ford Design Institute, Robustness Parameter
    Design, Ford Motor Company, January 1998.
  • Don Clausing, Total Quality Development, 1994.
  • ARR http//www-c3s.pd9.ford.com/vehs/arr/index
  • Nam P. Suh, The Principles of Design, Oxford
    University Press, Inc., New York, 1990

137
References
  • http//www.ctis.ford.com/ekb/whatis.html
  • http//www.op.dlr.de/FF-DR-ER/research/flight/gart
    eur.html
  • http//www.analogy.com/Milaero/designflow/Default.
    htm
  • http//www.spacefuture.com/archive/rlv_design_opti
    mization_for_human_presence_in_space.shtml
  • http//www.princeton.edu/stengel/FTFCS.html
  • http//www.promodel.com/products/supplychainguru/
  • http//www.hq.nasa.gov/office/codeq/aqshp/jacg01.h
    tm
  • http//members.aol.com/drmassoc/glossary.html
  • http//members.aol.com/drmassoc/robust.html
  • http//www.princeton.edu/stengel/FTFCS.html
  • http//www.princeton.edu/stengel/FTFCS.html
  • http//www.promodel.com/products/supplychainguru/
  • http//www.dearborn2.ford.com/avtqual2/fmea/index.
    htm

138
QUESTIONS ????
About PowerShow.com