Objectives
? Review the format and expectations for the
final exam
? Review material needed to succeed on the
final exam
? Set the material from the course in the
context of the product realization process
? Answer questions
16.881 MIT
Importance of the Final
? The course grade is determined by
– 40% term project
Balance?
– 30% final exam
– 20% homework
– 10% quizzes
Dominates
? The final shows how you’ve learned over
time
– Did the home works and quizzes stick?
16.881 MIT
Format of the Final Exam
? 4 essay questions (10% each)
? 20 short answer questions (3% each)
? Which means its
– 40% Essay
– 60% Short answer
? 2 Hours -- I’d suggest
– 10 minutes per essay
– 3 minutes per short answer
– 20 minute buffer / review
MIT16.881
Final Exam Rules
? Open book
? Open notes
? Solutions to homework, quizzes -- OK
? Calculators -- probably helpful
? Laptop computers -- fine, but not needed
16.881 MIT
Essay Questions
? Emphasize the big picture and concepts
? Composed of several inter-related questions
?Example
– What is a scaling factor?
– What are the properties of a good scaling factor?
– Provide an example of a scaling factor
– If you found that there was no control factor with the
desired properties, what would you do?
– Tell me anything you know about scaling factors that
you consider essential to practicing robust design.
16.881 MIT
Expectations on Essay Questions
? Answer the questions!
? Make your responses concise
– About 3 sentences per question if possible
? Make the answer complete but avoid a shot gun
approach
– Points will be deducted for imprecise statements
? Examples should have engineering relevance
? Examples should preferably be from some area
you know from experience rather than from a text
16.881 MIT
Short Answers
? Fairly similar to quizzes in format,
difficulty, and sometimes in content
? No multiple choice or true / false
? Usually come in clusters of 3-5
? Relate to a scenario, data table, graphs …
? Usually have a “right answer”
? Often require estimation
16.881 MIT
Expectations on Short Answers
? Right answer ± 10% gets you full credit
– So simplify and estimate when appropriate
? Right procedure gets you 2/3 credit
– So show your work if you have time
? A reasonable attempt gets you 1/3 credit
– So explain your assumptions
16.881 MIT
Short Answer -- Example
? The data below represent the results from an L
8
(2
7
).
Control Factors
Exp no. A B C D e e F
η
1 1 1 1 14
2 1 1 2 18
3 1 2 2 19
4 1 2 1 10
5 2 1 2 18
6 2 1 1 12
7 2 2 1 13
8 2 2 2 19
The fifth and sixth columns were left
1 1 1 1
2 2 2 1
2 1 1 2
1 2 2 2
1 2 1 2
2 1 2 2
2 2 1 1
1 1 2 1
unassigned.
? What is the factor effect f
1
?
? Estimate the sum of squares due to the mean.
? You wish to study interaction between control
factors A and D and also between factors F and B.
Will this design allow you to determine the effect
16.881
of these two interactions? MIT
What is Fair Game
? Any concept or technique described in
Phadke
? Any material in the lecture notes
? Any material in quizzes & home works
? Questions requiring original thought on
subtle topics not explicitly discussed in
class
16.881 MIT
High Probability Topics
? Ideal function
? Noise factors, control factors, signal factors,
responses
? Design for additivity
? Interaction plots
? Selection of appropriate OAs
? Dummy levels
16.881 MIT
High Probability Topics
? Orthogonality
? The balancing property
? Estimating variance of responses
? Quality loss functions
? ANOVA (Taguchi style)
?ANOM
? Design of dynamic systems
16.881 MIT
High Probability Topics
? Compounding factors
?Noise strategies
? System integration & RD
? Counting DOF of a system
? Selecting an OA to suit a scenario
? Studying interactions in OAs
? Tolerance design (insofar Phadke covers it)
16.881 MIT
High Probability Topics
? Failure modes & RD
? Confirmation experiments
? Column merging
? Factor effect plots
? The additive model
? Prediction based on the additive model
? Sliding levels
16.881 MIT
High Probability Topics
? Pooling and F ratios
? Choosing a proper S/N ratio
? Interpreting S/N ratios
? Making engineering and economic
judgements based on data
? Selecting quality characteristics
? Selecting control factors
16.881 MIT
Proactive Problem Solving
Example Essay
? You are the manager of a new product
development program. 75% of the technology in
the product is established and 25% is being fielded
for the first time.
? What techniques from this class would you apply?
? At what stages would you apply them?
? How would your efforts differ between the new
technology and the established technology?
16.881 MIT
Robust Design and Failure Modes
Example Essay
? When variance in a quality characteristic is too large,
describe how adjustment of the mean can lead to chasing
the problem from one failure mode to another (and often
back again).
? Give an example of this phenomenon from an engineering
context.
? If your product has multiple quality characteristics, how
does this impact this phenomenon?
? How can the architecture of the system aggravate or
ameliorate this problem?
16.881 MIT
MIT16.881
Noise Factors
Example Short Answer Group
? Air Shock Absorber -- h and D vary by 1%
? Estimate the ratio of the contribution of h and D
2
to variance in t
? Estimate the ratio of σ
t
to t
F
D
hD
t
π
ρπ 2
2
2
3
1
=
D
2
D
1
F
h
ρ
System Integration
Sample Essay
? Describe how lack of robustness in subsystems can lead to
difficulties in system integration.
? Give an example of a system integration problem due to lack
of robustness.
? If a robust design effort reduces the variance in all the
subsystems, how will this effect the variance of the system?
? How is this effect a function of system scale and system
architecture?
16.881 MIT
Noise Strategy
Sample Essay
? What is a compound noise factor?
? When would you use a compound noise factor?
? What is an outer orthogonal array?
? Compare the strategies of compounding noise factors with
employing an outer array of noise with regard to:
? Its effect on selection of control factor levels
? Tolerance design decisions
? Decision to field or not to field the system
? Discuss any alternate noise strategies you might consider
16.881 MIT
Parameter Design
Example Problem
?Given
– Description of engineering scenario
– Control factors and levels
? Questions
? Which signal-to-noise ratio would you use?
? How many experiments are required?
? What is the smallest experiment that will allow you to resolve
the main effects?
? What is the gain in experimental efficiency by switching from
one-factor-at-a-time to orthogonal array based experiments?
? It is likely that there is a significant interaction between A and
B. How will you ensure that your experimental plan can
resolve this interaction effect?
16.881 MIT
Next Steps
? Final exam
– 8AM (Sharp!) -10AM
? First off-campus session
– 3:03-4:55
? Each student may resubmit up to three quizzes
and/or home works by Monday 13 July (grades
will be averaged with the original grades)
16.881 MIT