Vous êtes sur la page 1sur 34

ME-5102

Research Methodology &


Design of Experiments
Lect. # 7
DESIGN OF EXPERIMENTS
DOE Models and
Strategy of Experimentations
Dr. Nazeer Ahmad Anjum
Mechanical Engineering Program
Engineering University Taxila
DOE Terminology 3
Fixed Effects Model
• If the treatment levels are specifically chosen
by the experimenter, then conclusions reached will
only apply to those levels.
• Use fixed-effects (FE) whenever you are only interested
in analyzing the impact of variables that vary over
time.
• The fixed-effects model controls for all time-invariant
differences between the individuals, so the estimated
coefficients of the fixed-effects models cannot be biased
because of omitted time-invariant characteristics…[like culture,
religion, gender, race, etc]

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
DOE Terminology 4
Fixed Effects Model
• FE explore the relationship between predictor and
outcome variables within an entity (country,
person, company, etc.). Each entity has its own
individual characteristics that may or may not
influence the predictor variables. FE models are
designed to study the causes of changes within a
material, process, person [or entity]
Examples:, being a male or female could influence
the opinion toward certain issue, or
The political system of a particular country could have
some effect on trade or GDP; or
The business practices of a company may influence its
stock price Dr. Nazeer Ahmad Anjum, UET, Taxila
4/27/2019
DOE Terminology 5
Random Effects Model
 If the treatment levels are randomly chosen from a
population of many possible treatment levels, then
conclusions reached can be extended to all treatment
levels in the population.
Textile Company uses large number of looms to produce
fabric. The looms function in uniform manner so fabric
produced by company is of uniform strength.
To investigate whether there are any difference between
looms, 4 looms will be selected at random from all those
owned by company.
Four samples will be obtained from each loom.
Fixed or Random Effects? Since looms are randomly
selected, a random effects model must be used. Results
obtained from these 4 looms can be generalized to the set
of all looms owned by company. Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Planning a DOE 6

Everyone involved in the experiment should


have a clear idea in advance of exactly what is
to be studied, the objectives of the
experiment, the questions one hopes to
answer and the results anticipated

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Planning a DOE 7

Select a response/dependent variable(s) that


will provide information about the problem
under study and the proposed measurement
method for this response variable, including an
understanding of the measurement system
variability

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Planning a DOE 8
Choose an appropriate experimental design
(relatively simple design and analysis methods are
almost always best).
It will allow your experimental questions to be
answered once the data is collected and analyzed.
But keeping in mind tradeoffs between statistical
power and economic efficiency.
At this point in time, it is generally useful to simulate
the study by generating and analyzing artificial data
to insure that experimental questions can be
answered as a result of conducting your experiment.
Dr. Nazeer Ahmad Anjum, UET, Taxila
4/27/2019
Planning a DOE 9

Select the independent variables/factors


(quantitative or qualitative) to be investigated
in the experiment, the number of levels for
each factor, and the levels of each factor
chosen either specifically (fixed effects model)
or randomly (random effects model).

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Planning a DOE 10

Perform the experiment (collect data) paying


particular attention such things as
randomization and measurement system
accuracy, while maintaining as uniform an
experimental environment as possible.
How the data are to be collected is a critical
stage in DOE

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Planning a DOE 11

Analyze the data using the appropriate


statistical model insuring that attention is paid
to check the model accuracy by validating
underlying assumptions associated with the
model.
Be liberal in the utilization of all tools, including
graphical techniques, available in the
statistical software package to insure that a
maximum amount of information is
generated

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Planning a DOE 12

Based on the results of the analysis, draw


conclusions/inferences about the results.
Interpret the physical meaning of these results.
Determine the practical significance of the
findings, and
Make recommendations for a course of
action including further experiments

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Strategy of Experimentation 13
• “Best-guess” experiments
– Used a lot by engineers and scientists; based on technical or
theoretical knowledge
– More successful than you might suspect, but there are
disadvantages…May continue for a long time or may stop
• One-factor-at-a-time (OFAT) experiments
– Sometimes associated with the “scientific” or
“engineering” method
– Devastated by interaction, also very inefficient
• Statistically designed experiments
– Based on Fisher’s factorial concept

4/27/2019
Dr. Nazeer Ahmad Anjum, UET, Taxila
Single-Factor Experiment 14
• Only one factor is investigated. The factor may
be either qualitative or quantitative.
• If the levels of a factor are qualitative (type of
tool, type of material, etc.), it is called
qualitative factor.
• If the levels of a factor are quantitative
(temperature, pressure, velocity, etc.), it is
called quantitative factor.
• The levels of a factor can be fixed (selecting
specific levels) or random (selecting randomly).

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Example of Single Factor 15
• Studying the effect of type of tool on surface
finish of a machined part.
• Effect of type of soil on yield.
• Effect of type of training program on the
performance of participants.
• Effect of temperature on the process yield.
• Effect of speed on the surface finish of a
machined part.
• If the levels are fixed, the associated statistical
model is called fixed effects model. Each level of
the factor considered to be treatment.
Dr. Nazeer Ahmad Anjum, UET, Taxila
4/27/2019
Completely Randomized Design 16
• In a single-factor experiment if the order of
experimentation as well as allocation of
experimental units (samples) is completely
random, it is called completely randomized
design.
• The statistical Model

(1)

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Completely Randomized Design 17
• Equation ‘1’ is a linear statistical model often called the effects
model. Also it is referred as one-way or single-factor Analysis of
Variance (ANOVA) model.
• The objective here is to test the appropriate hypotheses about the
treatment means and estimate them.
• For hypothesis testing, the model errors are assumed to be
normally independently distributed random variables with
mean zero and variance 2. And 2 is assumed as constant for
all levels of the factor.
• The appropriate hypotheses are as follows:

• Here we are testing the equality of treatment means or testing that


the treatment effects are zero.
• The appropriate procedure for testing a treatment means is
4/27/2019
ANOVA. Dr. Nazeer Ahmad Anjum, UET, Taxila
Completely Randomized Design 18
Typical Data for Single-Factor Experiment
• The data collected from a single-factor experiment can be shown as in Table
below.

Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019


Completely Randomized Design 19
Typical Data for Single-Factor Experiment
(2)

(3)

(4)

(5)

where N = a * n, the total number of observations.

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Completely Randomized Design 20
Analysis of Variance
The total corrected Sum of Squares (SS)

(6)

Equation (6) is used as a measure of overall variability in the


data. It is partitioned into two components.
Total variation = variation between treatments + variation within
the treatment or error
SSTotal = SS due to treatments + SS due to error
i.e., SSTotal = SST + SSe

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Completely Randomized Design 21
Analysis of Variance
A typical format used for ANOVA computations is shown in
Table 2
Table 2: ANOVA: Single-Factor Experiment

Mean square is also called variance.


Generally, we use 5% level of significance ( = 5%) for testing
the hypothesis in ANOVA. P-value can also be used.
Dr. Nazeer Ahmad Anjum, UET, Taxila
4/27/2019
Computation of Sum of Squares 22
Let T.. = grand total of all observations/response (Y..)
N = total number of observations
n = number of replications/number of observations under
the ith treatment..
SS = sum of squares
CF = correction factor (9)
Ti. = ith treatment total
(10)
(7)

(8)
Note that the factor level (treatment) totals are used to compute
the treatment (factor) sum of squares.
Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Completely Randomized Design 23
Example: A manufacturing engineer wants to investigate the
effect of feed rate (mm/min) on the surface finish of a milling
operation. He has selected three different feed rates, i.e., 2, 4 and 6
mm/min for study and decided to obtain four observations at each
feed rate.
Thus, this study consists of 12 experiments (3 levels * 4
observations). Since the order of experimentation should be random,
a test sheet has to be prepared as explained below.

Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019


Completely Randomized Design 24
Test Sheet
Order of Experiment Run number Feed Rate
(time sequence)
1 7 4
2 3 2
3 11 6
4 2 2
5 6 4
6 1 2
7 5 4
8 9 6
9 4 2
10 10 6
11 8 4
12 12 6
Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Completely Randomized Design 25
Data from Surface Finish Experiment
Feed Rate Observation (Surface roughness) Ti. i
(mm/min)
2 7.0 7.5 7.8 8.3 30.6 7.65
4 5.8 4.6 4.8 6.2 21.4 5.35
6 9.2 9.6 8.2 8.5 35.5 8.88

The data can be analyzed through analysis of variance. The


hypothesis to be tested is
H0: 1 = 2 = 3 against H1: at least one mean is different.

As shown in above table


Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Completely Randomized Design 26
Computation of sum of squares

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Completely Randomized Design 27
ANOVA for the Surface Finish Experiment

The inference is that the treatment means are significantly


different at 5% level of significance. That is, the treatment (feed
rate) has significant effect on the surface finish.
Dr. Nazeer Ahmad Anjum, UET, Taxila
4/27/2019
Factorial Designs 28
The general approach to plan and conducting the experiment is
called the strategy of experimentation. There are several strategies
that an experimenter could have
• In a factorial experiment, all possible
combinations of factor levels are
tested
• The golf experiment:
– Type of driver (over or regular)
– Type of ball (balata or 3-piece)
– Walking vs. riding a cart
– Type of beverage (Water vs Bear)
– Time of round (am or pm)
– Weather
– Type of golf spike
– Etc, etc, etc…
Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Factorial Designs 29
Driver Effect

Driver Effect = 92 + 94 + 93 + 91 - 88 + 91 + 88 + 90
4 4
= 3.25
Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Factorial Designs 30
Ball Effect

Ball Effect = 88 + 91 + 92 + 94 - 88 + 90 + 93 + 91
4 4
= 0.75 Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Factorial Designs 31
Ball-Driver Effect

Ball-Driver Interaction Effect = 92+94+88+90 _ 88+91+93+91


4 4
= 0.25
Dr. Nazeer Ahmad Anjum, UET, Taxila 4/27/2019
Factorial Designs with Several 32
Factors
Three Factor Factorial
Driver, Ball Type, Beverage Type

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019
Factorial Designs with Several 33
Factors
A Fractional Factorial
Four Factor Factorial
Driver, Ball Type, Beverage Type, Mode of Travel

Dr. Nazeer Ahmad Anjum, UET, Taxila


4/27/2019

Vous aimerez peut-être aussi