Design of Experiments (DOE)

A statistical methodology to systematically investigate a systems input – output

relationship.
In other words, it is used to find cause-and-effect relationships. This information is needed
to manage process inputs in order to optimize the output.
 Design of Experiments (DOE)

1. Cause and Effect diagram

Man Machine

Surface Finish

Material Env.

Robust Design
 Design of Experiments (DOE)

2. Select important factors for study ( Process parameters)

a. Speed ( H L)
b. Feed ( H L)
c. Type of tool (Carbide(C) HSS (S) )

Factors Levels
 Design of Experiments (DOE)

2 Level Designs 21 = 2 runs

22 = 4 runs

23 = 8 Runs
 22 Factorial Design

Ex. No Speed Feed S X F

1 1 1 2

2 2 1 1

3 1 2 1

4 2 2 2
 Design of Experiments (DOE)

3. Plan for experiments:

Ex.No Speed Feed Tool 1 2 3

1 H H C

2 L H C

3 H L C

4 L L C

5 H H S

6 L H S

7 H L S
8 L L S
 Design of Experiments (DOE)

4. Full factorial design:

(Include interactions)

Ex. No Speed Feed Tool SXF FXT SXT SXTXF

1 1 1 1 2 1
2 2 1 1 1 2
3 1 2 1 1 2
4 2 2 1 2 1
5 1 1 2 2 2
6 2 1 2 1 1
7 1 2 2 1 1
8 2 2 2 2 2

Balanced
Factors MF 2 IF 3 IF 4 IF

2 2 1

3 3 3 1

4 4 6 4 1
 Contrast and Effect of factors

Treatment Speed Feed 1 2 3

1 250 0.1 10 11 9
2 300 0.1 13 12 12
3 250 0.2 14 15 17
4 300 0.2 8 4 5
250 (-) 0.1(-)
Speed (A) Feed (B) r = 3 ( replicates)
300 (+) 0.2 (+)

Treatment A B AB 1 2 3 Total
(1) - - + 10 11 9 30
a + - - 13 12 12 37
b - + - 14 15 17 46
ab + + + 8 4 5 17
Contrast = ( Sum of responses at High) – (Sum of responses at Low)

Effect= (Contrast ) / (r * 2 k-1 ) for 2k factorial design

EFFECT

+VE --VE

LEVEL RESPONSE LEVEL RESPONSE

HIGH HIGH HIGH LOW

LOW LOW LOW HIGH

 Treatment 1 2 3 Total
(1) 10 11 9 30
a 13 12 12 37
b 14 15 17 46
ab 8 4 5 17

Main Effect of Factor A = [ a + ab – ((1) + b)] / r x 2k-1 = (37+17-30-46) /6 = -3.66

Main Effect of Factor B = [ b + ab – ((1) + a)] / r x 2k-1 = (46+17-30-37) /6 = -0.66

Effect of Interaction AB = [ (1) + ab – (a + b)] / r x 2k-1 = (30+17-37-46) /6 = -6

A+ & B+ or A- & B - A+ B+
 ANOVA
H0 for A = Response is not influenced by A
H0 for B = Response is not influenced by B
H0 = Response is not influenced by AXB interaction
Treatment 1 2 3 Total
(1) 10 11 9 30
a 13 12 12 37
b 14 15 17 46
ab 8 4 5 17
130

Sum of Squares: SS
(Contrast)2
=
r x 2k
Contrast = ( Sum of responses at High) – (Sum of responses at Low)
Effect= (Contrast ) / (r * 2 k-1 ) for 2k factorial design ( k : no. of
factors)

Sum of Squares: SSA/B/AB

=
(Contrast)2 SST = {∑ ∑ X }
ij
2 – (T2/N)
r x 2k
SSE= SST – (SSA + SSB + SSAB)
Mean Square: MS
SS
=
DOF F0= MS/MSE
Sum of Squares:

•SSA = (37+17-30-46)2/(3 x 22) = 40.33

•SSB= (46+17-30-37)2/(3 x 22) = 1.33
•SSAB= (30+17-37-46)2/(3 x 22) = 108
•SST = [102+112+92+132+.......+52] – (1302/N)
= 1574 – 1408.33 = 165.67
•SSE= SST – (SSA + SSB + SSAB) = 165.67- (40.33+1.33+108)
= 16.01
Source of Variation SS DOF MS= SS/DOF F0= MS/MSE

A 40.33 1 40.33 20.165

B 1.33 1 1.33 0.665
AB 108 1 108 54
Error 16.01 8 2
Total 165.67 11 15.1

For p value or α value of 0.05

Df1 of 1 ( Numerator)
Df2 of 8 ( Denominator)
We get F*α = F*0.05,1,8 = 5.32
 Interpreting F statistic

F0 of A and F0 of AB interaction are

greater than F* and hence Factor A and
Decision Rule : Interaction of AB have significant effect
Reject H0 if F0 > Fα or are influential in deciding the
response
(1) 20
a 25
b 30
ab 35
(1) 20
a 35
b 30
ab 25
 Example (15 marks)

Replicates

Trial No A (hours) B( 0C) 1 2 3 4

1 12 25 34 35 32 40

2 18 25 48 43 40 47

3 12 45 30 28 32 31

4 18 45 70 75 65 77

1. Estimate the factor effects. Which effects appear to be large?

2. Use analysis of variance to confirm your conclusions in part 1 taking p value of 0.05
3. Based on the analysis what levels of reaction time and temperature will you suggest for
higher the better response.
Treatment 1 2 3 4 Total
(1) 34 35 32 40 141
a 48 43 40 47 178
b 30 28 32 31 121
ab 70 75 65 77 287
727

Source of Variation Effect

A 25.375
B 11.125
AB 16.125
Contrast = ( Sum of responses at High) – (Sum of responses at Low)
Effect= (Contrast ) / (r * 2 k-1 ) for 2k factorial design ( k : no. of
factors)

Sum of Squares: SSA/B/AB

=
(Contrast)2 SST = {∑ ∑ X }
ij
2 – (T2/N)
r x 2k
SSE= SST – (SSA + SSB + SSAB)
Mean Square: MS
SS
=
DOF F0= MS/MSE
 ANOVA

MS= F0=
Source of Variation SS DOF SS/DOF MS/MSE

A 2575.6 1 2575.6 180.49

B 495 1 495 34.7
AB 1040 1 1040 72.89
Error 171.25 12 14.27
Total 4281.9 15 285.46
 2^3 factorial design

A B C AB AC BC ABC Trtmnt.
1 - - - + + + - (1)
2 + - - - - + + a
3 - + - - + - + b
4 + + - + - - - ab
5 - - + + - - + c
6 + - + - + - - ac
7 - + + - - + - bc
8 + + + + + + + abc
Treatment 1 2 3 Total
(1) 6 8 9 23
a 10 16 15 41
b 18 12 15 45
ab 12 9 10 31
c 20 26 29 75
ac 34 28 32 94
bc 36 44 46 126
abc 25 22 24 71
Source of Variation Effect

A -2.667
B 3.333
C 18.833
AB -8.833
AC -3.333
BC 1.333
ABC -3.500
Source of Variation SS DOF MS= SS/DOF F0= MS/MSE

A 42.66 1 42.66 4.023

B 66.66 1 66.66 6.299
C 2128.167 1 2128.167 201.093
AB 468.167 1 468.167 44.238
AC 66.667 1 66.667 6.299
BC 10.667 1 10.667 1.008
ABC 73.500 1 73.500 6.945
Error 169.33 16 10.583
Total 3025.833 23
 C-

AB+

A+B+ A-B-

A+ B+ C-