Spring 2024

ME 794 Statistical Design of Experiments

Chapter 0

Introduction
Prof. Soham Mujumdar
Email: sohammujumdar@iitb.ac.in

NOTE: NOTE: You

Some of thedocourse
NOT have permission
material to share
is adopted fromthis file or and
‘Design any of its contents
Analysis with anyone
of Experiments’ byelse,
D. C.and/or upload itand
Montgomery, on internet or any of
similar courses the platforms
taught where
by Prof. S. it can be
G. Kapoor accessed
at the by others.
University of Illinois at
Urbana-Champaign and Prof. S. S. Joshi at IIT Bombay. You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on the internet or any of the
platforms where it can be accessed by others.
ME 794 Schedule
Spring 2024, IIT Bombay

Time: Tuesdays and Fridays, 5.30 pm – 6.55 pm

LC 301

Instructor:
Prof. Soham Mujumdar
S23, Mechanical Engineering,
sohammujumdar@iitb.ac.in

Office hours: By appointment

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments

Grading Policy
Lecture Notes:
● Take your own notes
● Lecture notes/slides and reference material will be provided as and when required via MS Teams

Grading:
● Quiz 50% (Best 4/5, preannounced, after every 4 lectures, NO MAKE-UP)
● Assignments -- (For your own practice, no grading)
● Project 20% (Team project, actual experiments are expected)
● Final Exam 30% (Comprehensive)

Any student/group found to have committed or aided and abetted the offence of plagiarism will receive
ZERO marks for the relevant assignment/quiz without any exceptions

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 3

What is the NEED for experiments?
● Why do we perform experiments? (scientific or otherwise)

● Share what experiments you have performed or are currently planning

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments

Example Experiment
https://www.youtube.com/watch?v=XtcbRuuxXpA

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 5

Example Experiment

Ref: www.morestream.com

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 6

Need for Experiments
● To understand how a product functions or a process behaves
● To discover the direction of changes that may lead to improvements in both quality and productivity.

A fundamental task in designing a (scientific) experiment is to select an appropriate arrangement of test

points within the space defined by the independent variables and develop a mathematical model
● For example, if a quadratic relationship between two variables is suspected, an experiment that studies the process
at only two levels of these variables will be inadequate.
● Similarly, an experiment using four levels would be unnecessary and inefficient if the true relationship were linear.

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 7

Why Statistical Methods?
● The world around us is not deterministic! – Variability is part of the natural order of things
● Variation in data is neither totally chaotic nor small enough to be ignored
● It is real, identifiable, and predictable statistically

What is variability?

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 8

Example

How long will the tool last? I

would like to know the tool life
of this tool.

Turning operation under following conditions

Speed: 170 FPM
Feed: 0.017 IPR
Depth of Cut: 0.07 IN
Workpiece: 1018 Steel of 6 IN Dia, 24 IN Length
NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 9

 I will go down to
the lab and run a
test to determine
the tool life Okay!

Speed
Tool Life
Feed
Depth of Cut

Material

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 10

 I have found the tool
That’s fine, but why don’t you I already told you
life to be 15 minutes
go back and run another test. I that the tool life is 15
for the conditions
just want to be sure that the minutes, but I will
you specified. Okay!
tool life is 15 minutes. run another test if
you insist!

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 11

 After running another test …
Professor, this time I got tool life of 16.2 minutes.
But that’s probably because of fluctuations in the
machine power. If we install a power regulator, we
can eliminate this source of variation, and get a
true tool life value.
Power
Sure. Do whatever Regulator
you need to do and Speed
re-run the test. Tool Life
Feed
Depth of Cut

Material

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 12

 But what about machine
vibrations? Was it vibrating
After installing the power during the test? Should you not
regulator, I now get 15.6 minutes. make the machine more rigid Oh yes! You are right.
Okay!
This is the true tool life! and re-run the test? I better make the
machine more rigid.

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 13

 How can you be sure you have
eliminated ALL variation? What
about difference in materials,
Now I get 16.2 minutes. I am sure
tools, operator inconsistencies, Hmm. I suppose they
that THIS is the true tool life! Okay!
etc.? could cause some
variation,

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 14

 What I really want is an average tool life for
an average tool cutting a material over
average environmental conditions. There Okay, I will run more tests
will always be some variation in the using several tools, material
process. We do the best we can! pieces, etc. and find out what
the average tool life is.

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 15

 I have run tests over random sample
of tools, workpieces, time of day, etc.
Here are the results. I find the
average tool life of 16.1 minutes
Test # Tool Life Test # Tool Life
1 15 7 15
2 15.6 8 16.7
3 16.2 9 16
4 16.5 10 16
5 16.2 11 16.7
6 16.5 12 16.8

Average of 12 tests = 16.1 minutes

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 16

 That’s not good
enough. You better
tell me something
That’s fine, but how like “I am x% It certainly seems
confident are you that Pretty darn confident that true logical. But I don’t
16.1 min is the true confident! tool life is within know the answer!
tool life? certain range”. Isn’t Go and learn
that logical? Statistical DOE
and Data Analysis

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 17

Why Statistical Methods?
● The world around us is not deterministic! – Variability is part of the natural order of things
● It is real, identifiable, and predictable statistically
● We can easily define many factors of potential significance but only a few accounts for the vast majority
of the structure/variation in the data.
● But, the problem is to screen from a large group of potentially important factors those few, which are
worthy of continuing study.
● All processes are subject to identifiable and unidentifiable disturbances which can totally invalidate
results.
● The world around us is non-linear.

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 18

Experimental Difficulties
Experimental Error
● Composed of many minute disturbances which individually have little effect on the outcome of the
experiment.
● Collectively these small chance occurrences may increase the dispersion or spread of the results to the
point where real variable effects are masked
● Composed of more than errors of measurement – not all instrumentation oriented. A good measurement
system accounts for no more than 10-15% of the total error.
● Can be a function of both unknown and known sources.

Here, statistical methods can help. They will tell us,

● Can the results be explained solely by chance causes?
● How much data is required to reveal the existence of true effects in light of chance error?

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 19

Experimental Difficulties
Correlation vs. Causation
● Two factors are highly related only because they are related to a third common (often unidentified)
factor. Deliberate change in one may not lead to a change in the other- The concept of planned versus
passively observed
● To really find out how changes in some factor affect the output, you have to change that factor
deliberately and observe the change.

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 20

Experimental Difficulties
Complexity of Variable Effects
● Ideal scenario: variable effects are linear and additive! Generally not the case!
● Example: Effects of aspirin and coffee on driving reaction time: aspirin – increase Δ, coffee – reduce 2Δ.
● Will one aspirin and one cup of coffee reduce reaction time by 1Δ? Additive: Will 10 aspirins and 5 cups of coffee keep
reaction time constant? Linear?
● Need to plan experiments to reveal variable interactions and nonlinear variable effects

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 21

Experimental Difficulties
Example of Variable Interactions:
Temperature and Pressure are thought to affect the chemical reaction rate/time

Mr. X’s
Experiment

Mr. Y’s
Experiment

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 22

Why Statistical Methods?
● We know much less about what makes things work than we think.
● Experimentation is a costly and time-consuming business. We better do it right.

Statistical tools can help us navigate through experimental problems

A myth: “Everything he is saying sounds logical and probably works fantastically for some people – but my
specific problem simply doesn’t lend itself to this approach and/or simply doesn’t need it anymore.”

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 23

Course Objectives
● To learn DOE techniques based on probability theory and statistical tools
● Understand proper methods for data collection, defining quality parameters, diagnostic tools, quality
analysis, and interpretation for process/product improvement.
● Statistical DOE enables understanding of the relationship between multiple input variables/factors and
the key responses, or product/process performance.

In-depth knowledge of these tools is quintessential while conducting scientific experiments

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 24

List of Topics*
Fundamental Concepts and Methods Response Surface Methodology
● Quality philosophy and conceptual framework ● First and second-order models and surfaces
● Statistical Methods and Probability Concepts for
● Central composite designs
Data Characterization
● Multiple response analysis, Design rotatability, Box-Behnken design

Classical Design of Experiments Robust Design Method

● Nature of variability, probability distributions ● Quality loss function, signal, noise and control factors, product life cycle
● Empirical models (regression, hypothesis testing,
● Matrix experiments using orthogonal arrays, analysis of means and
confidence intervals, applications)
variance, error prediction
● Two-level factorial designs (factor effects, ANOVA,
residual analysis, interactions) ● Steps in robust design: noise factors and testing, signal to noise ratio,
degrees of freedom, selection of orthogonal arrays
● General 2^k factorial designs
● Conducting matrix experiments: randomization, confounding, result
● Two-level fractional factorial designs
interpretation, verification
● Factor interactions, dynamic problems

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 25

 Spring 2024

ME 794 Statistical Design of Experiments

Chapter 1

Fundamental Concepts
Prof. Soham Mujumdar
Email: sohammujumdar@iitb.ac.in

NOTE: NOTE: You

Some of thedocourse
NOT have permission
material to share
is adopted fromthis file or and
‘Design any of its contents
Analysis with anyone
of Experiments’ byelse,
D. C.and/or upload itand
Montgomery, on internet or any of
similar courses the platforms
taught where
by Prof. S. it can be
G. Kapoor accessed
at the by others.
University of Illinois at
Urbana-Champaign and Prof. S. S. Joshi at IIT Bombay. You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on the internet or any of the
platforms where it can be accessed by others.
Nature of Data
Question: Is Sachin Tendulkar a ‘good’ batsman?

How would you answer that?

● You need some data to decide ..

● In general, when we collect data, we are interested in,
● How the process/product is behaving in terms of an output quality characteristic(s)
● Perhaps, in terms of (a) average value of the characteristic, (b) the variation in individual measurements of the
characteristic

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 27

Nature of Data
How would you collect the data?

In general, when we collect data, we have three choices,

● Observe the process ONCE, and use that observation as the
absolute reflection of the process behavior
‘Fatal Error!’

● Observe ALL of the output of the process to get a true reflection of

its behavior
Observing entire population/universe, i.e. all possible realizations of
the process – a very large number. Neither practical nor necessary

● Observe PART OF the output of the process and use it to infer

something about the process behavior
More ‘practical’! We are ‘sampling’ the process (population)
NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 28

Nature of Data
We need an efficient and adequate method to collect and analyze the
data

Critical Issues in Sampling

● How much do we sample?
● How do we sample?
● When do we sample?
● How/what can we say something about the process from the
information contained in the sample?

We will keep exploring these answers throughout the course…

For now, it is important to understand

the difference between a Sample vs. the Population

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 29

Nature of Data CEP2022_Notebook (1.1)

What do you notice?

Variability!

Ref: http://www.cricketweb.net

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 30

Sources of Variability
Taguchi suggests that variation in product and process function arises from three basic sources:

● Outer Noise: Sources of noise which influence performance as measured during field use under actual
operating conditions, e.g., temp, humidity, supply voltage, vibration

● Inner Noise: Internal change in product characteristics such as drift from the nominal over time due to
deterioration, e.g., mechanical wear, aging

● Variational Noise: Variation in the product parameters from one unit to another as a result of the
manufacturing process, e.g., manufacturing imperfection

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 31

Example 1: Refrigerator
● Outer Noise [Operating Conditions]
• The number of times the door is opened and closed
• The amount of food kept and the initial temperature of the food
• Variation in the ambient temperature
• Supply voltage variation

● Inner Noise [Deterioration]

• The leakage of Refrigerant
• Mechanical Wear of Compressor parts

● Variational Noise [Mfg/Use Imperfection]

• The tightness of door closure
• The amount of refrigerant used

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 32

Example 2: Braking Distance of a Car
● Outer Noise [Operating Conditions]
• Wet or dry road
• Concrete or Asphalt pavement
• Number of passengers in the car

● Inner Noise [Deterioration]

• The leakage of brake fluid
• Wear of brake drums and brake pads

● Variational Noise [Mfg/Use Imperfection]

• Variation in friction coefficient of pads and drums
• The amount of brake fluid

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 33

Data Characterization
THREE Important Characteristics of Data
● Central Tendency
● Variability or Dispersion
● Shape of Frequency Distribution

NOTE: You do NOT have permission to share this file or any of its contents with anyone else, and/or upload it on internet or any of the platforms where it can be accessed by others.

ME 794 Statistical Design of Experiments 34