0 valutazioniIl 0% ha trovato utile questo documento (0 voti)

4 visualizzazioni55 pagineDesign of experiments

Jan 18, 2019

© © All Rights Reserved

0 valutazioniIl 0% ha trovato utile questo documento (0 voti)

4 visualizzazioni55 pagineSei sulla pagina 1di 55

Faculty, Department of Production Engineering ,

PSG College of Technology

Coimbatore 641004

An Experimental

Design is the laying

out of a detailed

experimental plan in

advance of doing the

experiment.

Design of Experiment Types

One Factor Designs

These are the designs where only one factor is under investigation, and

the objective is to determine whether the response is significantly

different at different factor levels

Factorial Designs

In factorial designs, multiple factors are investigated simultaneously

during the test.

As in one factor designs, qualitative and/or quantitative factors can be

considered.

The objective of these designs is to identify the factors that have a

significant effect on the response, as well as investigate the effect of

Two-level full factorial designs

Consider the two-level, full factorial design for three factors,

namely the 23 design.

A 2 two-level, full factorial design

3

Order'

run X1 X2 X3

1 -1 -1 -1

2 1 -1 -1

3 -1 1 -1

4 1 1 -1

5 -1 -1 1

A 23 two-level, full factorial 6 1 -1 1

design; factors X1, X2, X3 7 -1 1 1

8 1 1 1

Simple example

Suppose that we wish to

improve the yield of a Metal

Cutting of a composite

material.

are considered important to

the operation are

• Speed (X1)

• Feed (X2)

• Depth (X3).

relative importance of each of

these factors on Material

Removal Rate(Y)

High (+1), Low (-1), and Standard (0) Settings

Low (-1) Standard High (+1) Units

(0)

Speed 16 20 24 rpm

Feed 0.001 0.003 0.005 cm/sec

Depth 0.01 0.015 0.02 cm/sec

A 23 Two-level, Full Factorial

Design; Factors X1, X2, X3.

(The arrows show the

direction of increase of the

factors.)

There are three main effects, three two-factor interactions, and a

three-factor interaction, all of which appear in the full model as follows

Y=β0+β1X1+β2X2+β3X3+β12X1X2+β13X1X3+β23X2X3+β123X1X2X3+ϵ

Standard order

refers to a standard way of writing down the

settings of an experiment called `standard

order'

Full Factorial Design Table

Showing Runs in `Standard

Order'

X1 X2 X3

1 -1 -1 -1

2 +1 -1 -1

3 -1 +1 -1

4 +1 +1 -1

5 -1 -1 +1

6 +1 -1 +1

7 -1 +1 +1

8 +1 +1 +1

Replication

Running the entire The 23 Full Factorial Replicated

design more than Twice and Presented in Standard

once makes for Order Speed, X1 Feed, X2 Depth, X3

1 16, -1 .001, -1 .01, -1

easier data 2 24, +1 .001, -1 .01, -1

analysis because, 3 16, -1 .005, +1 .01, -1

4 24, +1 .005, +1 .01, -1

•each run (i.e., 5 16, -1 .001, -1 .02, +1

`corner of the 6 24, +1 .001, -1 .02, +1

design box') 7 16, -1 .005, +1 .02, +1

obtain average 8 24, +1 .005, +1 .02, +1

9 16, -1 .001, -1 .01, -1

value of the 10 24, +1 .001, -1 .01, -1

response 11 16, -1 .005, +1 .01, -1

•dispersion 12 24, +1 .005, +1 .01, -1

13 16, -1 .001, -1 .02, +1

(variability, 14 24, +1 .001, -1 .02, +1

consistency) of 15 16, -1 .005, +1 .02, +1

the response at 16 24, +1 .005, +1 .02, +1

that setting

Randomization

Suppose now that four settings are run in the day and four at night,

and that (unknown to the experimenter) ambient temperature in the

machining shop affects Yield.

The 23 Full Factorial Replicated Twice

The 2 Full Factorial Replicated Twice with Random Run

3

with Random Run Order Indicated

Rand Stand X1 X2 X3 Random Standard X1 X2 X3

Order Order

om ard

Order Order 1 0 0 0

1 5 -1 -1 +1 2 5 -1 -1 +1

2 15 -1 +1 +1 3 15 -1 +1 +1

3 9 -1 -1 -1 4 9 -1 -1 -1

4 7 -1 +1 +1 5 7 -1 +1 +1

5 3 -1 +1 -1 6 3 -1 +1 -1

6 12 +1 +1 -1 7 12 +1 +1 -1

7 6 +1 -1 +1 8 6 +1 -1 +1

8 4 +1 +1 -1 9 0 0 0

9 2 +1 -1 -1 10 4 +1 +1 -1

10 13 -1 -1 +1 11 2 +1 -1 -1

11 8 +1 +1 +1 12 13 -1 -1 +1

12 16 +1 +1 +1 13 8 +1 +1 +1

13 1 -1 -1 -1 14 16 +1 +1 +1

14 14 +1 -1 +1 15 1 -1 -1 -1

15 11 -1 +1 -1 16 14 +1 -1 +1

16 10 +1 -1 -1 17 11 -1 +1 -1

18 10 +1 -1 -1

19 0 0 0

Blocking

To eliminate the SPEED FEED DEPTH X1*X2* BLOC

influence of extraneous X1 X2 X3 X3 K

factors when running an

experiment.

When one has to -1 -1 -1 -1 I

change to a new batch +1 -1 -1 +1 II

of raw materials halfway -1 +1 -1 +1 II

through the experiment.

+1 +1 -1 -1 I

In this case, we need

to divide our experiment -1 -1 +1 +1 II

into two halves (2 +1 -1 +1 -1 I

blocks), one with the

-1 +1 +1 -1 I

first raw material batch

and the other with the +1 +1 +1 +1 II

new batch.

Genichi Taguchi (January 1, 1924 – June 2, 2012)

The Taguchi approach is more effective method than

traditional design of experiment methods such as

factorial design, which is resource and time consuming.

For example, a process with 8 variables, each with 3 states,

would require 38=6561 experiments to test all variables (full

factorial design).

However using Taguchi's orthogonal arrays, only L8

experiments are necessary, or less than 0.3% of the original

number of experiments.

Limitations of the Taguchi method.

Most critical drawback of the Taguchi method is that it does

not account higher order interactions between design

parameters.

Only main effects and two factor interactions are considered.

Parameter diagram (P-diagram) of a process/system

The Taguchi Approach to DOE

Traditional Design of Experiments (Fisher’s DOE)

focused on how different design factors affect the

average result level

Taguchi’s DOE (robust design)

Variation is more interesting to study than the

average

Run experiments where controllable design

factors and disturbing signal factors take on 2

or 3 levels.

Signal to Noise Ratio

a loss function is used

to calculate the

deviation between the

experimental value

and the desired value

The loss function is

further transformed

into utility function

The utility function

developed by Taguchi

is called the Signal-to-

Noise (S/N) ratio

Signal-to-Noise (S/N) ratio

Lower the better (for making

the system response as low

as possible)

reducing variability around a Where:

target) n = the number of tests

y ijk = experimental value

of i th performance

Higher the better (for

characteristic in the j th

making the system response

experiment at the k th test

as high as possible)

Flowchart indicating steps involved in Taguchi method

A Simple case of Taguchi Design

"Process parameter optimization for fly ash brick

by Taguchi method." Materials Research 11.2

(2008): 159-164.

Parameter

Design

Methodology

Selection of control

factors and their levels

are made on the basis

of some preliminary

trial experiments

conducted in the

laboratory and also

from literature review

on the subject.

Selection of Taguchi array

Design of

Experiment

An L9 (34) standard

orthogonal array11 as

shown in Table 2 was

employed

suitable to provide the

minimum degrees of

freedom as 9 [= 1 + 4

x (3–1)] required for

the experimental

exploration.

Larger the

Better

Compressive strength

is a 'larger the better'

type of quality

characteristic since

the goal is to

maximize the

strength.

The F–ratio and the percent

contributions of the various

parameters as quantified

under the respective columns

of Table 6 reveal that, both

water/binder ratio and stone

dust have significant effect on

the compressive strength at

the 95% confidence level.

ash and coarse sand are

insignificant since their F–

ratios are lower than the

critical value (3.55).

process parameters were

found to be A3, B3, C1, and D2

corresponding to water/binder

ratio of 0.4, fly ash of 39%,

coarse sand of 24%, and

stone dust of 30%

Thus a 95% confidence interval (CI) for the predicted mean of optimum QC on a confirmation

test is estimated using the following two equations:

where, F (α, 1, fe) is the F–ratio required for 100 (1– α) percent confidence

interval,

fe is DOF for error,Ve is the error variance,

R is number of replications for confirmation experiment (= 3), and

Neff is effective number of replications.

N is total number of experiments [= 27 (9 x 3)]

and TDOF is the total degrees of freedom [= 8 (2 x 4)

From ANOVA Table, the values are: Ve = 40.963, f.e= 18, and from standard

Statistical Table, the required F–ratio for α = 0.05 is: F (0.05,1,18) = 4.41.

interval is: CI = ± 10.97.

strength is obtained as:

(166.22 ± 10.97) kg.cm–2 i.e. 155.25 < Smp (kg.cm–2) < 177.19

Experimental investigation of surface quality in ultrasonic

machining of WC-Co composites through Taguchi method

http://dx.doi.org/10.3934/matersci.2016.3.1222

Factors Level 1 Level 2 Level 3

A-cobalt content in % 6 24

mm

80A steel s steel.

E-grit size in mesh 200 320 500

F-power rating in % 40 60 80

S.No A B C D E F SR1 SR2 Mean S/N

value ratio

1 1 1 1 1 1 1 0.782 0.751 0.7665 2.308

2 1 1 1 1 2 2 0.787 0.851 0.819 1.7277

3 1 1 1 1 3 3 0.291 0.213 0.252 11.869

2

8

34 2 2 1 3 1 3 1.171 1.187 1.179 -

1.4305

35 2 2 1 3 2 1 0.629 0.641 0.635 3.9441

36 2

A-cobalt 2

content 1 3 3 2 0.515 0.524 0.5195 5.688

C-tool geometry

D-tool material

E-grit size

F-power rating.

Source DOF SR (raw SR (S/N ratio)

data)

F P F P

A 1 4.48 0.039* 0.43 0.52

B 1 13.23 0.001* 5.75 0.026*

C 1 12.47 0.001* 5.83 0.025*

D 2 2.87 0.065 1.44 0.261

E 2 118.12 0.000* 55.96 0.000*

F 2 3.13 0.051 1.63 0.22

AXD 2 7.89 0.001* 2.95 0.076

BXD 2 0.61 0.549 0.34 0.717

CXD 2 1.23 0.299 0.74 0.491

Error 56

Total 71

D-tool

B-thickness of work C-tool material E-grit F-power

A-cobalt content piece geometry size rating.

*Significant at 95%

F- Fisher’s ratio P- Probability value confidence level.

Figure 7. Mean effect plot for surface roughness.

Experimental investigation of surface quality in ultrasonic machining of WC-Co composites through Taguchi method

http://dx.doi.org/10.3934/matersci.2016.3.1222

Figure 8. Interaction plots—(A) raw data, (B) S/N ratio.

Experimental investigation of surface quality in ultrasonic machining of WC-Co composites through Taguchi method

http://dx.doi.org/10.3934/matersci.2016.3.1222

Cause and Effect diagram of various factors that influence

machinability and wear characteristics of AMC

quality in ultrasonic machining of WC-

Co composites through Taguchi

method

Constructions

of Taguchi's

orthogonal

arrays

Kacker, Raghu N., Eric

S. Lagergren, and

James J. Filliben.

"Taguchi’s orthogonal

arrays are classical

designs of

experiments." Journal

of research of the

National Institute of

Standards and

Technology 96.5

(1991): 577.

Number of Interaction Limitations When used

experiments between design

needed variables

method (L9) nonlinear effect and of a project,

interactions initial design

between variables (mostly)

design 27 Coms impractical

for large number of

variables (>5)

factorial (9 for fraction less information

design 1/3 fraction)

composite estimated

design

Selected topics in Statistics

7 Step Process of Statistical Hypothesis Testing

Step 1: State the Null Hypothesis

H0: no statistically significant

difference between sample &

population (or between samples)

Step 2: State the Alternative

Hypothesis

H1: statistically significant difference

between sample & population (or

between samples)

Step 3: Set α

Often, the significance level is set to 0.05

(5%), implying that it is acceptable to have a

5% probability of incorrectly rejecting the In Reality

null hypothesis Decision H0 is TRUE H0 is FALSE

Step 4: Collect Data Accept

Type II Error

OK β = probability of

Step 5: Calculate a test statistic H0

Type II Error

Step 6: Construct Acceptance / Type I Error

α =

Rejection regions Reject H0

probability of

OK

Step 7: Based on steps 5 and 6, Type I Error

draw a conclusion about H 0

Statistical Hypothesis Testing -ANOVA

Step 1: State the Null

Hypothesis

The null hypothesis in ANOVA is always

that there is no difference in means.

Step 2: State the Alternative

Hypothesis

The research or alternative hypothesis

is always that the means are not all

equal

Step 3: Set α

Step 4: Collect Data

Step 5: Calculate a test

statistic

Step 6: Construct Acceptance

/ Rejection regions

Step 7: Based on steps 5 and

6, draw a conclusion about H0

Starting Point

Central aim of

statistical tests:

Determining the likelihood

of a value in a sample,

given that the Null

hypothesis is true: P(value|

H 0)

H0: no statistically significant

difference between sample &

population (or between

samples)

H1: statistically significant

difference between sample &

population (or between

samples)

H0) < 0.05

Types of Error

Population

H0 H1

b-error

H0 1-a

(Type II error)

Sample

H1 a-error

1-b

(Type I error)

Sir Ronald Aylmer Fisher FRS (17 February 1890 – 29 July 1962)

Introduction

tests

Experiments are used widely in the

engineering world

Process characterization & optimization

Evaluation of material properties

Product design & development

Component & system tolerance

determination

“All experiments are designed

experiments, some are poorly designed,

45 Ref: DOX Montgomery

Plan and Conduct of Experiments

Trial-and-Error Approach

Performing a series of experiments each of which

gives some understanding. This requires making

measurements after every experiment so that

analysis of observed data will allow him to decide

what to do next

Design of experiments

A well planned set of experiments, in which all

parameters of interest are varied over a specified

range, is a much better approach to obtain

systematic data

Four Eras in the History of DOX

The agricultural origins, 1918 – 1940s

R. A. Fisher & his co-workers

Profound impact on agricultural science

Factorial designs, ANOVA

The first industrial era, 1951 – late 1970s

Box & Wilson, response surfaces

Applications in the chemical & process

industries

The second industrial era, late 1970s – 1990

Quality improvement initiatives in many

companies

Taguchi and robust parameter design, process

robustness

The modern era, beginning circa 1990

The Basic Principles of DOX

Randomization

Running the trials in an experiment in random

order

Notion of balancing out effects of “lurking”

variables

Replication

Sample size (improving precision of effect

estimation, estimation of error or background

noise)

Replication versus repeat measurements? (see

page 13)

Blocking

48 Dealing withDOX

nuisance factors

6E Montgomery

Strategy of Experimentation

“Best-guess” experiments

Used a lot

More successful than you might suspect, but

there are disadvantages…

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

49 DOX 6E Montgomery

Determining the design space

Ref :http://www.gmpua.com/World/Manu/07/i.htm

Methods of statistical design of experiments

Ref :http://www.gmpua.com/World/Manu/07/i.htm

Formula to find Taguchi’s Loss Fn

Taguchi uses Quadratic Equation to determine

loss Curve

L (x) = k (x-N)²

Where L (x) = Loss Function,

k = C/d² = Constant of proportionality,

where C – Loss associated with sp limit

d - Deviation of specification from

target value

x = Quality Features of selected product,

N = Nominal Value of the product and

(x-N) = Tolerance

Main Effect Plot

Interactions Plot

P Value