Professional Documents
Culture Documents
Statistical Quality Control: Unit V
Statistical Quality Control: Unit V
STATISTICAL QUALITY
CONTROL
B. Thilaka
Applied Mathematics
Statistical Quality Control
• Quality Control – Process control &
acceptance sampling.
• Tolerance Limits
• Acceptance Sampling
Statistical Quality Control
• Quality- measurable / countable property
of a product
• Potency of a drug, breaking strength of a
yarn, outside diameter of a ball bearing,
number of imperfections in a piece of
cloth,….
Quality Control
The concept of quality control in manufacturing was
first advanced by Walter Shewhart
of the Bell Telephone Laboratories. He issued a
memorandum on May 16, 1924 that featured a
sketch of a modern control chart.
1931 - published a book on statistical quality
control, "Economic Control of Quality of
Manufactured Product"
set the tone for subsequent applications of
statistical methods to process control.
Quality Control
Two other Bell Labs statisticians, H.F.
Dodge and H.G. Romig spearheaded
efforts in applying statistical theory to
sampling inspection.
The work of these three pioneers
constitutes most of the theory of statistical
quality and control
Quality Control - Statistical
Process Control
Measurements(variables) Attributes
(Mean and Range) (p, np and c)
Control charts
Control charts for measurements:
Observations are measurements
Variables are the quality characteristics of a
product that are measurable.
WHERE? HOW?
EBOLA
Control charts
In many industries, processes will be
divided into stages and these principles of
SQC are applied to the individual stages
separately.
Unless a product is found to meet the
specified norms, it will not be permitted to
move to the next stage.
This stage-wise SQC scheme ensures that
the desired quality level is maintained at
all levels.
Control charts
Examples???
Food Processing
Wine-making
n 1 2
S ` S j
2
Further j is a better estimator than
n
S 2j . The average of the above values is
not an unbiased estimator of σ2.
However, computation of the mean
standard deviation is avoided by using the
range data instead of calculating the
adjusted standard deviation.
Control charts for measurements
The x chart:
1 k
We compute R Rj.
k j 1
Central line: X
Control Limits : X A2 R
Control charts for measurements
The R- chart :
While the x chart monitors the central
value of the process, the R-chart monitors
the variation of the process.
Central line = d 2
UCL = D2
LCL = D1
Control charts for measurements
The R- chart :
If σ is unknown, then it may be estimated
from past data as in the case of the x
chart. Further
Central line = R
UCL = D4 R
LCL = D3 R
Control charts for measurements
Thus, there is a close connection between the use
of control charts and the testing of hypothesis. A
point on the x chart that is out of control
corresponds to a sample for which the null
hypothesis is rejected.
1
Also, R Ri
N
1
5 6 5 ... 6 = 5.8
10
Control charts for measurements
From the table of control chart constants,
for sample size n=5, we have
A2 = 0.577, D3 = 0 and D4 = 2.115.
Sample Number : 1 2 3 4 5
Weights of boxes(X): 10.0 10.3 11.5 11.0 11.3
10.2 10.9 10.7 11.3 11.6
11.3 10.7 11.4 10.7 11.0
12.4 11.7 12.4 11.4 12.1
Control charts for measurements
Sample Number : 6 7 8 9 10
Weights of boxes(X): 10.7 11.3 12.3 11.0 11.3
11.4 11.4 12.1 13.1 12.1
10.7 11.1 12.7 13.1 10.7
11.0 10.3 10.7 12.4 11.5
Sample Number : 11 12 13 14 15
Weights of boxes(X): 12.5 11.9 12.1 11.9 10.6
11.9 12.1 11.1 12.1 11.9
11.8 11.6 12.1 13.1 11.7
11.3 11.4 11.7 12.0 12.1
Control charts for measurements
Solution:
Since the given data is raw data, we first
compute the means and ranges of the given
samples.
Sample 1 2 3 4 5 6 7
No. (i)
1
Now, X Xi
N
= 1.33
Sample No. 1 2 3 4 5 6
7 8 9 10 11 12 13 14 15
13.5 11.5 13.5 13.0 14.5 9.5 12.0 10.5 11.5
6.2 4.3 3.4 4.1 3.9 5.1 4.7 3.3 3.3
Solution:
HW
Control charts for measurements
4. The following data gives the coded measurements
of 10 samples each of size 5, drawn from a
production process at intervals of 1 hour. Draw the
appropriate control charts and comment on the state
of control of the process.
Sample 1 2 3 4 5 6
Number(i)
9 10 10 8 7 12
Coded 15 11 13 13 9 15
Measurements 14 13 8 11 10 7
(X) 9 6 12 10 4 16
13 10 7 13 5 10
Control charts for measurements
(i) 7 8 9 10
9 15 10 16
9 15 13 14
(X) 9 10 14 12
13 13 7 14
5 17 11 14
Solution: HW
Control charts for attributes
The control charts seen earlier plots
quality characteristics that can be
measured and expressed numerically. We
measure weight, height, position,
thickness, etc.
If a particular quality characteristic cannot
be represented numerically, or if it is
impractical to do so, then using a quality
characteristic to sort or classify an item
that is inspected into one of two "buckets“
is a good alternate option.
Control charts for attributes
BINOMIAL
In other words,
Central line : p
p(1 p)
Control limits: p3
n
Control charts for attributes (p)
If no standard is given(practical situations),
p will have to be estimated from past data.
d1 d 2 ... d k
p
n1 n2 ... nk
Control charts for attributes (p)
Hence the three control-chart values are
Central Line : p
p (1 p )
UCL : p3
n
p (1 p )
LCL : p 3
n
Control charts for attributes (p)
Note :
Example:
number of defects per hundred yards in
carpeting industry
number of defects per roll in newsprint
industry.
Control charts for attributes (c)
Control Limits for no-of-defects chart :
POISSON - WHY???
Control charts for attributes (c)
k
i 1
i
UCL : c c
LCL : c c
Control charts for attributes
Note:
when p ≥0.05 or np ≥4
(i) 5 6 7 8 9 10
(np) 19 15 17 11 13 20
`(i) 11 12 13 14 15
(np) 10 8 9 5 8
Control charts for attributes
Solution:
Consider ∑np=12+15+10+..+5+8=180
1
np np
N
180
= 12
15
Also,
12
p (since each sample
200
contains 200 items).
The control limits are thus given by
Control charts for attributes
Central line :
p 0.06
LCL : p 3 p(1 p)
n
0.06 x0.094
0.06 3
200
= 0.01
Control charts for attributes
UCL : p (1 p )
p3
n
0.06 x0.094
0.06 3
200
=0.11
The fraction defectives (values of p) for the
given samples are:
0.06, 0.075, 0.05, 0.04, 0.095, 0.075,
0.085, 0.055, 0.065, 0.1, 0.05, 0.04, 0.045,
0.025, 0.04.
Control charts for attributes
Control charts for attributes
Sample No. : 1 2 3 4
No. inspected : 90 65 85 70
No. of defectives : 9 7 3 2
5 6 7 8 9 10
80 60 70 95 90 75
9 5 3 9 6 7
Control charts for attributes
Solution:
Since the size of the sample varies from
sample to sample, we do not construct the
np-chart.
We use the p-chart only if 0.75n ni 1.25n
for all i.
1
Now, n ni = (90+65+…+90+75)/10
N = 800/10 = 80
Control charts for attributes
The values of ni are all between 60 and
100 .
Hence, we now draw the p-chart.
CL : p 0.075
p(1 p) 0.075 x0.025
LCL : p 3 0.075 3
n 80
= -0.013
Since LCL cannot be negative, it is assumed
to be zero.
p (1 p) 0.075 x0.025
p 3 0.075 3 0.163
UCL : n 80
Control charts for attributes
(i) 11 12 13 14 15
(np) 4 7 9 3 4
Solution:
CL : np =6
Control charts for attributes
Solution : HW
1
Now, c ci
N
2 4 3 ... 2 1
1
15
= 45/15=3
Control charts for attributes
Aside: Observe that even though c <4,
we use only the c-chart at this point of
time.
We now compute the three control limits :
CL : c = 3
LCL : c 3 c = 3-3√3 = -2.20
Since LCL cannot be negative, LCL=0
UCL : c 3 c = 3+3√3 = 8.20
Control charts for attributes
Control charts for attributes
Since all the points lie within the control
limits, the process is under control.
(i) 6 7 8 9 10 11 12 13
(c) 22 7 13 18 13 16 14 8
14 15 16 17 18 19 20
7 6 4 5 6 8 9
Solution : HW
Tolerance Limits
In quality control, the limiting values between
which measurements must lie if a product is to
be accepted is loosely defined as tolerance
limits.
(i): 8 9 10 11
X : 15.0 16.5 14.9 15.3
R : 3.8 5.0 2.9 13.8
(i): 12 13 14 15
X : 17.8 15.9 14.6 15.2
R : 14.2 4.8 5.0 2.2
Acceptance Sampling
Compute the control limits for X and R-
charts using the above data for all the
samples. Hence examine if the process is
in control. If not, remove the doubtful
samples and recompute the values of X
and R . After testing the state of control,
estimate the tolerance limits and find if the
process will meet the required
specifications.
Acceptance Sampling
Solution:
We first compute the following values for
the given data of 15 samples.
1
X X
N
16.1 15.2 .... 15.2
1
15
230.9
15
= 15.39
Acceptance Sampling
1
R R
N
CL : R = 4.93
LCL : D3 R = 0
UCL : D4 R = 11.25
Acceptance Sampling
From the values calculated above, we see
that though the process is under control
with respect to the average ( X - chart), it is
not under control with respect to the
variability (R- chart), since the R values of
sample numbers 11 and 12 (13.8 and 14.2
respectively) exceed the UCL = 11.25
R
1
3.0 2.1 ... 2.9 5.8 4.0 2.2
13
1
x 45.9 = 3.53
13
Acceptance Sampling
We now compute the revised control limits
for both the charts.
CL : = 3.53
LCL : D3 R = 0
UCL : D4 R = 11.25
(i) 6 7 8 9 10
49 74 67 66 62
(X) 98 63 70 65 68
65 62 68 58 66
64 57 56 52 68
66 62 61 58 73
64 64 66 50 68
Acceptance Sampling
Solution : HW