Question 6
First we should define the benchmark that we are going to compare with. Though
different programs have different courses and requirements we can still find benchmarks
from other programs. S
1
INTRODUCTION TO
DESIGNED EXPERIMENTS
Outline for Introduction
2
Experimentation vs. Testing
Why perform designed Experiments
Basic Components
performance measure, factors, levels
Treatments, replica
MADHURI
Quality Assurance
SUMMARY
Over 7+ years of experience in QA testing methodologies and automated testing tools.
Technical expertise in Automation / Test Management tools like Quick Test Pro, Wi
1
ACCEPTANCE SAMPLING
Accept or reject a lot?
What is a lot?
Goal: decision making,
not parameter estimation
Also called lot sentencing
Basic Idea of Acceptance
Sampling
3
Consumer receives a large lo
1
PROCESS CAPABILITY,
STRATIFICATION, MIXING
Process Capability
2
Can the process do the job the customer
wants?
Is the process a good match for the
customer?
Process capability compares consumer
spec
1
CONTROL CHART FOR
INDIVIDUAL
MEASUREMENTS
Sometimes a sample of 5 measurements
in a row doesnt make sense
2
Oven temperature
Inventory level
Pressures
Accounts payable
If variance in sample reflects
1
CONSTRUCTING X-BAR
AND R CONTROL CHARTS
i>clicker
2
1.
2.
3.
4.
Which of these sample statistics belong on
variables control charts?
Sample
Sample
Sample
Sample
average amount filled by filler
range
1
SIGNIFICANCE WITH
ONLY 1 REPLICATE
Finding stat sig effects in
2k Experiment with rep
2
Model
Finding significant effects rests on
estimate of
How to estimate with just 1 replicate?
Fishing Experime
1
SEQUENTIAL EXPERIMENTS,
RANDOMIZATION,
BLOCKING
Do a sequence of experiments
2
Screening
Start with many
possible factors and
identify important
ones with big main
effects
Modeling
Create a refined
1
ATTRIBUTE CONTROL
CHARTS
Attribute Control Charts for sample
statistics based on discrete r.v.
2
Count Defectives Binomial 0,1,2,n
Count Defects Poisson 0,1,2,
lower case
n is
sample
size
i>clicker
1
SIGNIFICANCE WITH
REPLICATES
Which effects are significant in 2k
Factorial Experiment with 2 reps
2
Model
Assumption:
same sd for
all
treatments
Which effects (and coefficients) stat
significantly d
1
C AND U CHARTS FOR
NUMBER OF DEFECTS
C Chart for Number of Defects
2
Examples:
contaminants per sample of recycled plastic
defects (all types) per sample of cars
defects (scratches, bubbles, etc.) p
Name:_
540:433 Quality Engineering and Statistics Midterm I
1.
2016 SOLUTION
(10 points) Give the reduced model that relates the outcome Y to the factors including error
term.
Performance Measure, fo
1
STATISTICAL PROCESS
CONTROL INTRODUCTION
There is variation in ALL processes!
2
Two types
Chance Causes - many unavoidable
relatively small causes of variation
Assignable Causes- unusual things that
Name_
540:433 Quality Engineering - Examination II - Fall 2016 SOLUTIONS
1. (10 points) To monitor the dimension of a part, the following
X
and R charts with a sample
size of 4 are in use. (NOT NECESS
1
AN EXPERIMENT
REVEALS A MODEL
When are factorial experiments
appropriate?
2
2 levels means linear model
MRE
beef
yield
MRE
beef
yield
20 mins
25 mins
30 mins
20 mins
25 mins
30 mins
Linear models of
1
ESTIMATING & GRAPHING
MAIN & INTERACTION
EFFECTS
Question what is the best way to
make microwave popcorn?
2
Performance measure poprate ?
Popcorn Experiment with 5 replicates
3
Factors
Time in micro
1
QUALITY
MANAGEMENT
2
This course is about
Quality
Products
Services
Processes
Shorthand: one or more of the above
interchangeably
i>clicker
3
It costs more to
provide better quality
services or prod
1
IS DATA FROM A
NORMAL DISTRIBUTION?
What you need to know from
probability
2
8 Observations of weights of
newborns. Normal Dist?
3
11.0500
.9121
9.4
9.545
.07
9.021
5.21
7.592
Weights, X
Histogram w
Question 2
1. Overall sigma level:
DPMO=(6/5000)*1000000/5=240
So its approximately 5 sigma with off-centering of 1.5 sigma
2. For the characteristic that showed two defects:
DPMO=(2/5000)*1000000=400
Question 1
=0.0009 5
0.00095 t
R ( t ) =e
Probability of failure within 5000 hours=1R ( 5000 )=0.991348
Question 2
a)
Reliability=0.98 0.97 [ 1( 10.91 ) ( 10.91 ) ] 0.99=0.933471
b) Configuration:
2
2
Table 1 - Raw Data
Defect Category
Get up
before
10am
Tooth
brushing
Having
breakfast
Take
everything
Pefect
parking
Taking
exercise
Keep in
touch with
parents
Total
Day #1
Day #2
Day #3
Day #4
Day #5
1
STATISTICAL PROCESS
CONTROL
INTRODUCTION
There is variation in ALL processes!
2
Two types
Chance Causes - many unavoidable
relatively small causes of variation
Assignable Causes- unusual things that
1
PROCESS CAPABILITY,
STRATIFICATION, MIXING
Process Capability
2
Can the process do the job the customer
wants?
Is the process a good match for the
customer?
Some examples
3
Patient has complex
combi
1
FRACTIONAL
FACTORIAL DESIGNS
Problem: 7 Factor experiment has 128
runs - too many!
2
Solution: fractional factorial design with
7 factors and only 16 runs
Is this possible?
Comparison of Full and Fa
1
CONTROL CHART FOR
INDIVIDUAL
MEASUREMENTS
Sometimes a sample of 5 measurements
in a row doesnt make sense
2
Oven temperature
Inventory level
Pressures
Accounts payable
If variance in sample reflects