Quality Evaluations
• According to the traditional concept of loss function that as long as the
product’s quality characteristics is within the specification limits or
tolerances, there are no losses incurred. The lose takes the form of a
step function with a constant value A outside of the specification limits or
tolerances.
• The concept is simple but effective. Taguchi defines quality as “the total
loss imparted to society from the time a product is shipped to the
customer.”
41
Quality Evaluations
• Loss function L(y) makes no distinction between products
– Whose quality characteristic is exactly on target at m, or
– Whose quality characteristic is just above lower specification limit (LSL), or
– Whose quality characteristics is just below the upper specification limit (USL)
• The conventional method of computing the cost of quality is based on the number of parts rejected
and reworked in a production environment. This method of quality evaluation is incapable of
distinguishing between two samples, both within the specification limits but with different distributions
of targeted properties.
42
Quality Evaluations
• It is also doubtful that the loss function L(y) remains constant at a value of
A beyond the specification limits USL and LSL.
• Since the Taguchi loss function increases quadratically with increasing
deviation from the target value, the Taguchi loss function can be applied
to three different situations or cases namely
– Target is best (or) Nominal is best
– Smaller is better
– Larger is better
Target is best (or) Nominal is best
Larger is better
Smaller is better
Ranjit K . Roy, A Primer on the Taguchi Method, Second
43
Quality Evaluations (S/N Ratio)
• Taguchi experiments often use a 2-step optimization process.
• In step 1 use the signal-to-noise ratio to identify those control factors that
reduce variability.
• In step 2, identify control factors that move the mean to target and have a
small or no effect on the signal-to-noise ratio.
• Higher values of the signal-to-noise ratio (S/N) identify control factor
settings that minimize the effects of the noise factors.
• The signal-to-noise ratio measures how the response varies relative to
the nominal or target value under different noise conditions. You can
choose from different signal-to-noise ratios, depending on the goal of your
experiment.
• The important contribution of Taguchi is proposing the signal-to-noise (S/N)
ratio. It was developed as a proactive equivalent to the reactive loss
function.
44
Quality Evaluations (S/N Ratio)
Mean
Standard Deviation
or Variation
[Link]
Method-05Sep08
45
Quality Evaluations
Target is best or The Nominal-the-Best Case
46
Quality Evaluations
Target is best or The Nominal-the-Best Case
47
Quality Evaluations
Target is best or The Nominal-the-Best Case
48
8-Steps in Taguchi Methodology
1. Choose control factors and their levels
2. Identify uncontrollable (noise) factors and decide on how they will be simulated.
3. Select the response variable(s) and determine the performance measures (mean, standard
deviation, SNR, etc.)
4. Setup the experimental layout (choose appropriate design array (s))
5. Conduct the experiments and collect data
6. Analyze the data (effects, ANOVA, regression)
7. Choose optimal control factor levels and predict the performance measure at these levels
8. Confirm the optimal levels by experimentation.
49
Taguchi Analysis (Example)
• Instead of 8 combinations, we tried only four combinations of setting
and from these four we were able to predict the GR achievable for
those combinations of controls which we have not tried.
• For 3 — two level factors, the combinations possible are only = 8.
Here we tried 4 experiments.
• But suppose there are seven factors like speed, infeed, dressing
traverse etc. each with three setting levels, then the possibility is =
2187 combinations.
• The effect of all these 2187 experiments can be predicted with just 18
experiments. (for 3 level factors of 7 numbers, we use L18 array).
50
Taguchi Analysis (Orthogonal Array)
L8 (2^7) ORTHOGONAL ARRAY
Expt. Columns
No 1 2 3 4 5 6 7
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 1 2 1 2 L9 (3^4) ORTHOGONAL ARRAY
6 2 1 2 2 1 2 1
7 2 2 1 1 2 2 1
8 2 2 1 2 1 1 2 [Link] Columns
1 2 3 4
1 1 1 1 1
2 1 2 2 2
3 1 3 3 3
4 2 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3 1 3 2
8 3 2 1 3
9 3 3 2 1
51
Taguchi Analysis (Orthogonal Array)
52
Taguchi Analysis (Example Problem)
• The objective of the experiment is to obtain minimum surface roughness of
the parts machined by wire cut EDM machine using Taguchi’s DOE
technique of process parameters.
• Surface roughness(Ra) values were obtained using Taylor-Hobson
Surtronic instruments and the sequence are 2.5275, 2.3520, 2.2540,
2.4650, 2.7000, 2.8125, 2.350, 2.2875.
• Find out the Main effect and draw the factor effect diagram and identify the
optimum WEDM process parameters.
53
Taguchi Analysis (Example Problem)
• Since, the number of factors given are 7 with 2 levels. Hence, we can
choose L8 orthogonal array.
Expt. N Columns
o 1 2 3 4 5 6 7
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 1 2 1 2
6 2 1 2 2 1 2 1
7 2 2 1 1 2 2 1
8 2 2 1 2 1 1 2
54
Taguchi Analysis (Example Problem)
• L8 orthogonal array.
• Smaller-the-Better: η = -10 log10( 𝟐)
• Hence solving for η
Expt. No Columns
Surface Rou
Pull on Time Pulse off time ( Peak Current( Wire Tension ( Fluid Press η
Wire Feed (E) Fluid Rate (F) ghness (y)
t (A) B) C) D) ure (G)
1 125 (1) 35 (1) 11 (1) 1000 (1) 7 (1) 8 (1) 13 (1) 2.5275 η1
2 125 (1) 35 (1) 11 (1) 1200 (2) 8 (2) 9 (2) 15 (2) 2.3520 η2
3 125 (1) 40 (2) 12 (2) 1000 (1) 7 (1) 9 (2) 15 (2) 2.2540 η3
4 125 (1) 40 (2) 12 (2) 1200 (2) 8 (2) 8 (1) 13 (1) 2.4650 η4
5 130 (2) 35 (1) 12 (2) 1000 (1) 8 (2) 8 (1) 15 (2) 2.7000 η5
6 130 (2) 35 (1) 12 (2) 1200 (2) 7 (1) 9 (2) 13 (1) 2.8125 η6
7 130 (2) 40 (2) 11 (1) 1000 (1) 8 (2) 9 (2) 13 (1) 2.350 η7
8 130 (2) 40 (2) 11 (1) 1200 (2) 7 (1) 8 (1) 15 (2) 2.2875 η8
µ
55
Taguchi Analysis (Example Problem)
• L8 orthogonal array.
• Smaller-the-Better: η = -10 log10( 𝟐)
• Hence solving for η
Expt. No Columns
Surface Rou
Peak current Pulse on time ( Peak Current( Wire Tension ( Fluid Press η
Wire Feed (E) Fluid Rate (F) ghness (y)
(A) B) C) D) ure (G)
1 125 (1) 35 (1) 11 (1) 1000 (1) 7 (1) 8 (1) 13 (1) 2.5275 -8.053823285 η1
2 125 (1) 35 (1) 11 (1) 1200 (2) 8 (2) 9 (2) 15 (2) 2.3520 -7.428746348 η2
3 125 (1) 40 (2) 12 (2) 1000 (1) 7 (1) 9 (2) 15 (2) 2.2540 -7.059078234 η3
4 125 (1) 40 (2) 12 (2) 1200 (2) 8 (2) 8 (1) 13 (1) 2.4650 -7.836338472 η4
5 130 (2) 35 (1) 12 (2) 1000 (1) 8 (2) 8 (1) 15 (2) 2.7000 -8.627275283 η5
6 130 (2) 35 (1) 12 (2) 1200 (2) 7 (1) 9 (2) 13 (1) 2.8125 -8.981850622 η6
7 130 (2) 40 (2) 11 (1) 1000 (1) 8 (2) 9 (2) 13 (1) 2.350 -7.421357245 η7
8 130 (2) 40 (2) 11 (1) 1200 (2) 7 (1) 8 (1) 15 (2) 2.2875 -7.187222055 η8
µ -7.82446144
56
Taguchi Analysis (Example Problem)
• Determine the factor effects
mA1 =1/4 (η1 + η2 + η3 + η4) = 0.25 * (-8.053-7.428-7.059-7.836) = -7.5945
mA2 =1/4 (η5 + η6 + η7 + η8) = 0.25 * (-8.627-8.981-7.42-7.187) = -8.053
mB1 =1/4 (η1 + η2 + η5 + η6) =
mB2 =1/4 (η3 + η4 + η7 + η8) =
mC1 =
mC2 =
mD1 =
mD2 =
mE1 =
mE2 =
mF1 =
mF2 =
mG1 =
mG2 =
57
Taguchi Analysis (Example Problem)
• Determine the factor effects
mA1 =1/4 (η1 + η2 + η3 + η4) = 0.25 * (-8.053-7.428-7.059-7.836) = -7.5945
mA2 =1/4 (η5 + η6 + η7 + η8) = 0.25 * (-8.627-8.981-7.42-7.187) = -8.053
mB1 =1/4 (η1 + η2 + η5 + η6) = -8.272
mB2 =1/4 (η3 + η4 + η7 + η8) = -7.37
mC1 =1/4 (η1 + η2 + η7 + η8) = -7.522
mC2 =1/4 (η3 + η4 + η5 + η6) = -8.125
mD1 =1/4 (η1 + η3 + η5 + η7) = -7.789
mD2 =1/4 (η2 + η4 + η6 + η8) = -7.858
mE1 =1/4 (η1 + η3 + η6 + η8) = -7.82
mE2 =1/4 (η2 + η4 + η5 + η7) = -7.827
mF1 =1/4 (η1 + η4 + η5 + η8) = -7.925
mF2 =1/4 (η2 + η3 + η6 + η7) = -7.722
mG1 =1/4 (η1 + η4 + η6 + η7) = -8.072
mG2 =1/4 (η2 + η3 + η5 + η8) = -7.575
58
Taguchi Analysis (Example Problem)
• Determine the factor effects (graph)
59
Taguchi Analysis (Example Problem)
• Determine the factor effects (graph)
60
Taguchi Analysis (Example Problem)
• A1, B2, C1, D1, E2, F2, G2 giving the minimum effect in reducing the
surface roughness.
η = µ + (mA1 - µ) + (mB2 - µ) + (mC1 - µ) + (mD1 - µ) + (mE2 - µ) + (mF2 - µ) + (mG2 - µ)
= -7.823 + (-7.594+7.823) + (-7.37+7.823) + (-7.597+7.823) + (-7.789+7.823) + (-7.827+7.823) + (-
7.722+7.823) + (-7.573+7.823)
= -6.533
η = -6.533
i.e., η = -10 log10( 𝒚𝟐 ) = -6.533
log10( 𝒚𝟐 ) = 0.6533
𝒚𝟐 = antilog (0.6533)
𝒚𝟐 = 4.5009
𝒚 = 2.12
Hence, Surface Roughness = 2.12 with the above combination.
61