Mrwtaguchismaller

Download as xls, pdf, or txt
Download as xls, pdf, or txt
You are on page 1of 13

Example of Taguchi calculations for

a smaller-the-better characteristic

Statement: The goal of an experiment is to improve the production of quality by minimize the nu
Three factors with two levels are studied
The factors of customer requirements are selected
The results are presented in the following table:

A Material used
B Durability
C Rigidity and Toughness

Run A B C R1 R2
1 1 1 1 55 63
2 1 2 1 52 51
3 1 1 2 66 65
4 1 2 2 52 63
5 2 1 1 54 71
6 2 2 1 60 62
7 2 1 2 70 77
8 2 2 2 52 69

The first run represents the current conditions.


Find the optimal condition and the gain it will provide if adopted.

STEP 1: Compute S/N ratios

In this case, the quality characteristic is the amount of wear and has to be minimize
n=2 k=8
The formula used for the S/N ratios is so:

where:
Run Number of defects _j2 _j A B C
1 55 63 3497.0 -35.4 1 1 1
2 52 51 2652.5 -34.2 1 2 1
3 66 65 4290.5 -36.3 1 1 2
4 52 63 3336.5 -35.2 1 2 2
5 54 71 3978.5 -36.0 2 1 1
6 60 62 3722.0 -35.7 2 2 1
7 70 77 5414.5 -37.3 2 1 2
8 52 69 3732.5 -35.7 2 2 2

STEP 2: Compute factor effects

Each factor has two level, two effects by factor have to be calculated.

Level A B C
1 -141.2 -145.1 -141.4
2 -144.8 -140.9 -144.6

STEP 3: Complete the ANOVA

a/ Variations of factors
We use the following formulas to compute the variation of each
factor and the total variation ST.
Every factor has two levels:

k: number of runs k=8


A B C T
Variations S 1.56 2.20 1.31 5.07

b/ Degree of freedom

Reminder: fT=k-1 so fT= 7


f factor P= number of level-1

Results:
A B C T
f 1 1 1 3

c/ Variances

Reminder: Total variance: Variance of factor

Results:
A B C T
Variance V 1.56 2.20 1.31 1.69

According to these results, we can identify factors making up the larger


variances terms: B
So factors considered as source error is A and C
We can now estimate variation and variance for the error source.

d/ Error variation Se and error variance Ve

Reminders:

fe= _ (f factor M)
Results: Se= SA+SC Se=
So fe=
fe= fC+ Ve=

e/ Pure variations

Reminders: For the factors considered as significant (A,C,E,I and J):

For the error source:

Results:
A B C error
Pure variation 0.25 0.89 0.00 5.23

f/ Degree of contribution

Reminders: For the factors considered as significant (B and C):

For the error source:

Results:
A B C ERROR
_(%) 4.9 17.6 0.0 103.3

We can now draw the ANOVA table with all the results:

Source f S V S' _ (%)


A 1 1.56 1.56 0.25 3.9
B 1 2.20 2.20 0.89 14.0
C 1 1.31 1.31 0.00 0.0
e 1 1.31 1.31 5.23 82.1
Total 3 6.38 100

STEP 4: Investigate the optimum condition

The table with effects of every factor allows to select the best level for
each factor to maximize the Signal-to-Noise ratio:

Level A B C
1 -141.23 -145.09 -141.38
2 -144.76 -140.90 -144.61

According to these values, the optimum condition is:

Factor A B C
Level 1 2 1

Evaluate quality improvement

a/ Process average under optimum conditions

Reminders:

where

Results:
-35.75

STEP 5:
A1 = -70.62

C1 = -70.69

Finally: _= -105.56 decibel

b/ Process average under current conditions

The current combination of level is the following (first run):

Factor A B C
Level 1 1 1

So the current process average is:

? = T + (A2-T)+(C2-T)

A2 = -72.38

C2 = -72.31

_0= -108.94 decibel _-_0= 3.38

By adopting the optimal condition, the process average increases by about 10.4 dec

So _2improvement= 4.59E-01 2.18

The Loss function under the optimal condition will be:


where L0 is the Loss function under current conditions

The quality improvement will be:


ity by minimize the number of defects of the products.

d has to be minimized.
iation of each
Variance of factor P:
1.31
1
1.31

A,C,E,I and J):

B and C):
ect the best level for
decibel

ses by about 10.4 decibel.

You might also like