Quality Control

Dr. Everette S. Gardner, Jr.

 x Correlation:
Strong positive
x x Positive
x Negative
x x
*Strong negative

Acoustic trans.,
Im Engineering Competitiv
po

Check force
rta characteristic e

resistance
resistance
close door
needed to

open door
needed to
nc

Door seal
e ts x= Us
evaluation

on level
A= Comp. A

window
Energy

ground
oc

Energy

Water
us B= Comp. B
Customer to (5 is best)
m
requiremen er
1 2 3 4 5
ts to close
Easy 7 x AB
Stays open on a hill 5 x AB
Easy to open 3 xAB
Doesn’t leak in rain 3 A xB
No road noise 2 x A B
Importance weighting 10 6 6 9 2 3 Relationships:
Reduce energy

Strong = 9

energy to 7.5
Medium = 3
level to 7.5

Target values
force to 9 Small = 1

Maintain
Maintain

Maintain
Reduce

current
Reduce

current
current

level
level
level

ft/lb
ft/lb

lb.

5 Source: Based on John R.

B BA BA
Hauser
Technical 4 x B B B xA x and Don Clausing, “The
House of 3 A x Quality,” Harvard Business
A
Review, evaluation (5 is 2 x A May-June 1988.
best) x
1
Quality 2
Taguchi analysis
Loss function
L(x) = k(x-T)2
where
x = any individual value of the quality characteristic
T = target quality value
k = constant = L(x) / (x-T)2

Average or expected loss, variance known

E[L(x)] = k(σ2 + D2)
where
σ2 = Variance of quality characteristic
D2 = ( x – T)2

Note: x is the mean quality characteristic. D2 is zero if the

mean equals the target.
Quality 3
Taguchi analysis (cont.)
Average or expected loss, variance unkown
E[L(x)] = k[Σ ( x – T)2 / n]

When smaller is better (e.g., percent of impurities)

L(x) = kx2

When larger is better (e.g., product life)

L(x) = k (1/x2)

Quality 4
Introduction to quality control
charts
Definitions
• Variables Measurements on a continuous scale, such as length or
weight
• Attributes Integer counts of quality characteristics, such as nbr.
good or bad
• Defect A single non-conforming quality characteristic, such as a
blemish
• Defective A physical unit that contains one or more defects

Types of control charts

Data monitored Chart name Sample

size
• Mean, range of sample variables MR-CHART 2 to 5 units
• Individual variables I-CHART 1 unit
• % of defective units in a sample P-CHART at least 100
units
• Number of defects per unit C/U-CHART 1 or more units
Quality 5
Sample
mean
value
0.13% Upper control limit

Normal
99.74% Process mean tolerance
of
process
Lower control limit

0.13%

0 1 2 3 4 5 6 7 8
Sample number

Quality 6
Reference guide to control
factors
n A A2 D3 D4 d2 d3
2 2.121 1.880 0 3.267 1.128 0.853
3 1.732 1.023 0 2.574 1.693 0.888
4 1.500 0.729 0 2.282 2.059 0.880
5 1.342 0.577 0 2.114 2.316 0.864

• Control factors are used to convert the mean of sample

ranges
( R ) to:
(1) standard deviation estimates for individual
observations, and
(2) standard error estimates for means and ranges of
samples

For example, an estimate of the population standard

deviation of individual observations (σx) is:
σx = R / d 2
Quality 7
 Reference guide to control
factors (cont.)
• Note that control factors depend on the sample size n.

• Relationships amongst control factors:

A2 = 3 / (d2 x n1/2)
D4 = 1 + 3 x d3/d2
D3 = 1 – 3 x d3/d2, unless the result is negative, then D3 =
0

A = 3 / n1/2
D2 = d2 + 3d3
D1 = d2 – 3d3, unless the result is negative, then D1 = 0

Quality 8
Process capability analysis
1. Compute the mean of sample means ( X ).

2. Compute the mean of sample ranges ( R ).

3. Estimate the population standard deviation (σx):

σx = R / d 2

4. Estimate the natural tolerance of the process:

Natural tolerance = 6σx

5. Determine the specification limits:

USL = Upper specification limit
LSL = Lower specification limit

Quality 9
Process capability analysis
(cont.)
6. Compute capability indices:
Process capability potential
Cp = (USL – LSL) / 6σx

Upper capability index

CpU = (USL – X ) / 3σx

Lower capability index

CpL = ( X – LSL) / 3σx

Process capability index

Cpk = Minimum (CpU, CpL)

Quality 10
Mean-Range control chart
MR-CHART
1. Compute the mean of sample means ( X ).

2. Compute the mean of sample ranges ( R ).

3. Set 3-std.-dev. control limits for the sample means:

UCL = X + A2R
LCL = X – A2R

4. Set 3-std.-dev. control limits for the sample ranges:

UCL = D4R
LCL = D3R

Quality 11
Control chart for percentage
defective in a sample — P-CHART
1. Compute the mean percentage defective ( P ) for all samples:
P = Total nbr. of units defective / Total nbr. of units
sampled

2. Compute an individual standard error (S P ) for each sample:

SP = [( P (1-P ))/n]1/2

Note: n is the sample size, not the total units sampled.

If n is constant, each sample has the same standard
error.

3. Set 3-std.-dev. control limits:

UCL = P + 3SP
LCL = P – 3SP Quality 12
Control chart for individual
observations — I-CHART
1. Compute the mean observation value ( X )
X = Sum of observation values / N
where N is the number of observations

2. Compute moving range absolute values, starting at obs.

nbr. 2:
Moving range for obs. 2 = obs. 2 – obs. 1
Moving range for obs. 3 = obs. 3 – obs. 2
…
Moving range for obs. N = obs. N – obs. N – 1

3. Compute the mean of the moving ranges ( R ):

R = Sum of the moving ranges / N – 1
Quality 13
Control chart for individual
observations — I-CHART (cont.)
4. Estimate the population standard deviation (σX):
σX = R / d 2
Note: Sample size is always 2, so d2 = 1.128.

5. Set 3-std.-dev. control limits:

UCL = X + 3σX
LCL = X – 3σX

Quality 14
Control chart for number of
defects per unit — C/U-CHART
1. Compute the mean nbr. of defects per unit ( C ) for all samples:
C = Total nbr. of defects observed / Total nbr. of units
sampled

2. Compute an individual standard error for each sample:

SC = ( C / n)1/2

Note: n is the sample size, not the total units sampled.

If n is constant, each sample has the same standard error.

3. Set 3-std.-dev. control limits:

UCL = C + 3SC
LCL = C – 3SC

Notes:
● If the sample size is constant, the chart is a C-CHART.
● If the sample size varies, the chart is a U-CHART.
● Computations are the same in either case.
Quality 15
 Quick reference to quality
formulas
• Control factors
n A A2 D3 D4 d2 d3
2 2.121 1.880 0 3.267 1.128 0.853
3 1.732 1.023 0 2.574 1.693 0.888
4 1.500 0.729 0 2.282 2.059 0.880
5 1.342 0.577 0 2.114 2.316 0.864

• Process capability analysis

σx = R / d 2
Cp = (USL – LSL) / 6σx CpU = (USL – X ) /
3σx
CpL = ( X – LSL) / 3σx Cpk = Minimum (CpU,
CpL)
Quality 16
 Quick reference to quality
formulas (cont.)
• Means and ranges
UCL = X + A2R UCL = D4R
LCL = X – A2R LCL = D3R

• Percentage defective in a sample

SP = [( P (1-P ))/n]1/2 UCL = P + 3SP
LCL = P – 3SP

• Individual quality observations

σx = R / d2 UCL = X + 3σX
LCL = X – 3σX

• Number of defects per unit

SC = ( C / n)1/2 UCL = C + 3SC
LCL = C – 3SC
Quality 17
Multiplicative seasonality
The seasonal index is the expected ratio of
actual data to the average for the year.

Actual data / Index = Seasonally adjusted data

Seasonally adjusted data x Index = Actual data

Quality 18
Multiplicative seasonal
adjustment
1. Compute moving average based on length of
seasonality (4 quarters or 12 months).

2. Divide actual data by corresponding moving average.

3. Average ratios to eliminate randomness.

4. Compute normalization factor to adjust mean ratios so

they sum to 4 (quarterly data) or 12 (monthly data).

5. Multiply mean ratios by normalization factor to get

final seasonal indexes.

6. Deseasonalize data by dividing by the seasonal index.

7. Forecast deseasonalized data.

8. Seasonalize forecasts from step 7 to get final

forecasts.
Quality 19
Additive seasonality
The seasonal index is the expected difference
between actual data and the average for the
year.

Actual data - Index = Seasonally adjusted

data

Seasonally adjusted data + Index = Actual

data

Quality 20
Additive seasonal adjustment
1. Compute moving average based on length of
seasonality (4 quarters or 12 months).

2. Compute differences: Actual data - moving average.

3. Average differences to eliminate randomness.

4. Compute normalization factor to adjust mean

differences so they sum to zero.
5. Compute final indexes: Mean difference –
normalization factor.

6. Deseasonalize data: Actual data – seasonal index.

7. Forecast deseasonalized data.

8. Seasonalize forecasts from step 7 to get final

forecasts.
Quality 21
How to start up a control chart
system
1. Identify quality characteristics.

2. Choose a quality indicator.

3. Choose the type of chart.

4. Decide when to sample.

5. Choose a sample size.

6. Collect representative data.

7. If data are seasonal, perform seasonal adjustment.

8. Graph the data and adjust for outliers.

Quality 22
How to start up a control chart
system (cont.)
9. Compute control limits

10. Investigate and adjust special-cause variation.

11. Divide data into two samples and test stability of limits.

12. If data are variables, perform a process capability study:

a. Estimate the population standard deviation.
b. Estimate natural tolerance.
c. Compute process capability indices.
d. Check individual observations against
specifications.

13. Return to step 1.

Quality 23