16-56

a) With a target = 100 and a shift to 102 results in a shift of

102 100
= 0.5 standard deviations.
4

From Table 16-9, ARL = 38. The hours of production are 2(38) = 76.
b) The ARL = 38. However, the time to obtain 38 samples is now 0.5(38) = 19.
c) From Table 16-9, the ARL when there is no shift is 465. Consequently, the time between false alarms is
0.5(465) = 232.5 hours. Under the old interval, false alarms occurred every 930 hours.

d) If the process shifts to 102, the shift is 0 = 102 100 = 1 standard deviation. From Table 16-9, the

/ n

4/ 4

ARL for this shift is 10.4. Therefore, the time to detect the shift is 2(10.4) = 20.8 hours. Although this time
is slightly longer than the result in part (b), the time between false alarms is 2(465) = 930 hours, which is
better than the result in part (c).
Supplementary Exercises
a)
X-bar and Range - Initial Study
-------------------------------------------------------------------------------X-bar
| Range
---| ----UCL: +
3.0 sigma = 64.0181
| UCL: +
3.0 sigma = 0.0453972
Centerline
= 64
| Centerline
= 0.01764
LCL: 3.0 sigma = 63.982
| LCL: 3.0 sigma = 0
|
out of limits = 0
| out of limits = 0
-------------------------------------------------------------------------------Chart: Both
Normalize: No
Estimated
process mean = 64
process sigma = 0.0104194
mean Range
= 0.01764

Xbar/R Chart for C1-C3

Sample Mean

64.02

UCL=64.02

64.01
Mean=64.00

64.00
63.99

LCL=63.98

63.98
Subgroup

10

0.05

Sample Range

16-57

25

UCL=0.04541

0.03
0.02

R=0.01764

0.01
LCL=0

The process is in control.

c)

20

0.04

0.00

R 0.01764
=
= 0.0104
d2
1.693
USL LSL 64.02 63.98
PCR =
=
= 0.641
6
6(0.0104)

b)  = x = 64

15

16-37

The process does not meet the minimum capability level of PCR 1.33.
d)

USL x x LSL
PCRk = min
,
3
3
64.02 64 64 63.98
= min
,

3(0.0104) 3(0.0104)
= min[0.641,0.641]
= 0.641

e) In order to make this process a six-sigma process, the variance 2 would have to be decreased such that
x LSL
= 2.0 for :
PCRk = 2.0. The value of the variance is found by solving PCRk =
3

64 63.98
= 2.0
3
6 = 64.0 63.98
64.0 63.98
=
6
= 0.0033
Therefore, the process variance would have to be decreased to 2 = (0.0033)2 = 0.000011.
f)  x = 0.0104

63.98 64.01 X 64.02 64.01

P (63.98 < X < 64.02) = P

<
<
x
0.0104
0.0104
= P(2.88 < Z < 0.96) = P( Z < 0.96) P( Z < 2.88)
= 0.8315 0.0020 = 0.8295
The probability that this shift will be detected on the next sample is p = 10.8295 = 0.1705

ARL =

1
1
=
= 5.87
p 0.1705

16-38

a)

Xbar/S Chart for C1-C3

Sample Mean

64.02

UCL=64.02

64.01
Mean=64.00

64.00
63.99

LCL=63.98

63.98
Subgroup

10

15

20

25

UCL=0.02382

Sample StDev

16-58

0.02

0.01

S=0.009274

0.00

LCL=0

b)  = x = 64

c) Same as 16-57

s
0.009274
=
= 0.0104
c4
0.8862
USL LSL 64.02 63.98
PCR =
=
= 0.641
6
6(0.0104)

The process does not meet the minimum capability level of PCR 1.33.
d) Same as 16-57

USL x x LSL
PCRk = min
,
3
3
64.02 64 64 63.98
= min
,

3(0.0104) 3(0.0104)
= min[0.641,0.641]
= 0.641

e) Same as 16-57 e). In order to make this process a six-sigma process, the variance 2 would have to be
x LSL
decreased such that PCRk = 2.0. The value of the variance is found by solving PCRk =
= 2.0 for :
3

64 63.98
= 2.0
3
6 = 64.0 63.98
64.0 63.98
=
6
= 0.0033

Therefore, the process variance would have to be decreased to 2 = (0.0033)2 = 0.000011.

16-39

f) Same as 16-57  x = 0.0104

63.98 64.01 X 64.02 64.01

P (63.98 < X < 64.02) = P

<
<
x
0.0104
0.0104
= P(2.88 < Z < 0.96) = P( Z < 0.96) P( Z < 2.88)
= 0.8315 0.0020 = 0.8295
The probability that this shift will be detected on the next sample is p = 10.8295 = 0.1705

ARL =
a)

P Chart for def

UCL=0.2039

Proportion

0.2

P=0.11

0.1

LCL=0.01613
0.0
0

10

20

Sample Number

There are no points beyond the control limits. The process is in control.
b)

P Chart for def2

0.19

1
UCL=0.1764

Proportion

16-59

1
1
=
= 5.87
p 0.1705

0.14
P=0.11
0.09

LCL=0.04363

0.04
0

10

20

Sample Number

There is one point beyond the upper control limit. The process is out of control. The revised limits are:

16-40

P Chart for def2

0.19
UCL=0.1717

Proportion

0.14

P=0.1063
0.09

LCL=0.04093

0.04
0

10

20

Sample Number

There are no further points beyond the control limits.

c) A larger sample size with the same percentage of defective items will result in more narrow control limits.
The control limits corresponding to the larger sample size are more sensitive to process shifts.
a)

U Chart for Defects

2

1
1
UCL=1.503

Sample Count

16-60

U=0.528

LCL=0
0

10

15

20

25

Sample Number

Points14 and 23 are beyond the control limits. The process is out of control.

16-41

b) After removing points 14 and 23, the limits are narrowed.

U Chart for Defects

1.5

Sample Count

UCL=1.302
1.0

0.5

U=0.4261

LCL=0

0.0
0

10

20

Sample Number

c) The control limits are narrower for a sample size of 10

U C hart for defec ts n=10

1.0

1
1

Sample Count

U C L= 0.7514

0.5

U = 0.264

LC L= 0

0.0
0

10

15

20

25

S am ple N um ber

U Chart for defects n=10

0.7
UCL=0.6509

Sample Count

0.6
0.5
0.4
0.3
U=0.2130

0.2
0.1

LCL=0

0.0
0

10

20

Sample Number

16-42

a) Using I-MR chart.

I-MR Chart of C1

Individual V alue

60.3275

U C L=60.327362

60.3270
_
X=60.32641

60.3265
60.3260
60.3255

LC L=60.325458
2

10
12
O bser vation

14

16

18

20

U C L=0.001169
M oving Range

0.00100
0.00075
0.00050

__
M R=0.000358

0.00025
LC L=0

0.00000
2

10
12
O bser vation

14

16

18

20

b) The chart is identical to the chart in part (a) except for the scale of the individuals chart.

I-MR Chart of C1

Individual V alue

0.0015

U C L=0.001362

0.0010
_
X=0.00041

0.0005
0.0000
-0.0005

LC L=-0.000542
2

10
12
O bser vation

14

16

18

20

U C L=0.001169
0.00100
M oving Range

16-61

0.00075
0.00050

__
M R=0.000358

0.00025
LC L=0

0.00000
2

10
12
O bser vation

14

16

18

c)The estimated mean is 60.3264. The estimated standard deviation is 0.0003173.

USL LSL
0.002
=
= 1.0505
6
6(0.0003173)
0.0009 0.0011
,
PCRk = min
= 0.9455
3
3

PCR =

16-43

20

16-62

a)

I-MR Chart of enery

1

Individual V alue

7000

U C L=6669
_
X=5832

6000

LC L=4995

5000
1

4000
3

12

15
18
O bser vation

21

24

27

30

U C L=1028

M oving Range

1000
750
500

__
M R=315

250
0

LC L=0
3

12

15
18
O bser vation

21

24

27

30

b) The data does not appear to be generated from an in-control process. The average tends to drift
to larger values and then drop back off over the last 5 values.
a)
Trial control limits :
S chart
UCL= 170.2482
CL = 86.4208
LCL = 2.59342

600
400

xbar

800

X bar chart
UCL= 670.0045
CL = 558.766
LCL = 447.5275

10

15

20

25

15

20

25

50

150

Index

16-63

10
Index

16-44

b) An estimate of is given by

S / c4 = 86.4208 / 0.9515 = 90.8259

830 558.77 558.77 330
,
=0.8396
3(90.8259) 3(90.8259)

PCR=500/(6*90.8259)=0.9175 and PCRk= min

Based on the capability ratios above (both <1), the process is operating off-center and will result
in a large number of non-conforming units.
c) To determine the new variance, solve

PCRk = 2 for . Since PCRk=

558.77 330
in this exercise, we find =38.128 or 2=1453.77.
3

d) The probability that

X falls within the control limits is

670
.
0045
600
447
.
5275

600

=
P (447.5275 < X < 670.0045) = P
<Z<
90.8259
90.8259

6
6

P( 4.11 < Z < 1.89) = 0.9706

Thus, p=0.0294 and ARL=1/p=34.01. The probability that the shift will be detected in the next
sample is 0.0294.
For part b) through d), if we remove the out-of-control samples in part a) and recalculate the control limits, we
will get the following control limits for S chart and Xbar chart.
S chart
UCL= 158.9313
CL = 80.67615
LCL = 2.421028
X bar chart
UCL= 655.7918
CL = 551.9477
LCL = 448.1035

b) =84.78839

USL x x LSL
,
PCRK = min
3
3
830 551.95 551.95 330
,
= min
3(84.79)
3(84.79)
= 0.8725
c)

USL x x LSL
PCRK = min
,
3
3
551.95 330
=
3( )
=2
So = 36.9917

16-45

d) In-control distribution

~ N(551.9477, 34.6 )
2

Out-of-control distribution X ~ N(600, 34.6 )

X 655.8, when =600]

448.1 600
655.8 600
= P[
Z
]
34.6
34.6
= P[ 4.39 Z 1.6124 ]

P[448.1

=0.9463
Out-of-control ARL=
16-64

1
= 18.6 19.
1 0.9463

X Control Chart with 2-sigma limits

CL =

UCL = + 2

, LCL = 2
n
n
X

P X > + 2
> 2 = P( Z > 2) = 1 P( Z < 2) = 1 0.97725 = 0.02275
= P
n

/ n

and
X

P X < 2
< 2 = P( Z < 2) = 1 P( Z < 2) = 1 0.97725 = 0.02275
= P
n

/ n

The answer is 0.02275 + 0.02275 = 0.0455. The answer for 3-sigma control limits is 0.0027. The 3-sigma
control limits result in many fewer false alarms.
16-65
a) The following control chart use the average range from 25 subgroups of size 3 to estimate the process
standard deviation. Minitab uses a pooled estimate of variance as the default method for an EWMA control
chart so that the range method was selected from the options. Points are clearly out of control.
EWMA Chart of C3, ..., C5
64.7
64.6
64.5

EWMA

64.4
64.3

UCL=64.2759

64.2

_
_
X=64.1334

64.1
64.0

LCL=63.9910

63.9
2

10

12 14
Sample

16-46

16

18

20

22

24

b) The following control chart use the average range from 25 subgroups of size 3 to estimate the process
standard deviation. There is a big shift in the mean at sample 10 and the process is out of control at

this point.
EWMA Chart of C3, ..., C5
65.75
65.50
65.25

EWMA

65.00
64.75
64.50
UCL=64.380
_
_
X=64.133

64.25
64.00

LCL=63.887
2

10

12 14
Sample

16

18

20

22

24

16-66
a) The data appears to be generated from an out-of-control process.
EWMA Chart of C1
6600
6400
6200

UCL=6111

EWMA

6000

_
_
X=5832

5800
5600

LCL=5553

5400
5200
5000
3

12

15
18
Sample

21

b) The data appears to be generated from an out-of-control process.

16-47

24

27

30

EWMA Chart of C1
7000

6500
UCL=6315

EWMA

6000

_
_
X=5832

5500
LCL=5349
5000

4500
3

12

15
18
Sample

21

24

27

30

a) The process appears to be in control.

EWMA Chart of C1
60.3268
UCL=60.3267273

60.3267

EWMA

60.3266
60.3265
_
_
X=60.32641

60.3264
60.3263
60.3262
60.3261

LCL=60.3260927
2

10
12
Sample

14

16

18

20

b) The process appears to be in control.

EWMA Chart of C1
60.32700

UCL=60.326960

60.32675

EWMA

16-67

60.32650

_
_
X=60.32641

60.32625

60.32600
LCL=60.325860
2

10
12
Sample

16-48

14

16

18

20

16-68

CUSUM Chart of C1
UCL=5.27

5.0

Cumulative Sum

2.5

0.0

-2.5

-5.0

LCL=-5.27
3

12

15
18
Sample

21

24

27

30

Process standard deviation is estimated using the average moving range of size 2 with MR/d2, where d2 =
1.128. The estimate is 1.05. Recommendation for k and h are 0.5 and 4 or 5, respectively for n =1. For this
chart h = 5 was used.
The process is not in control.
K = k = 1, so that k = 0.5
H = h = 10, so that h = 5

CUSUM Chart for hardness

Upper CUSUM

40

Cumulative Sum

16-69

30
20
10

10
0
-10

-10

Lower CUSUM

10

Subgroup Number

16-49

20

16-70

CUSUM Chart for Viscosity

Upper CUSUM

85.8529

Cumulative Sum

80

-80

-85.8529

Lower CUSUM

10

20

Subgroup Number

Process standard deviation is estimated using the average moving range of size 2 with MR/d2, where d2 =
1.128 for a moving range of 2. The estimate is 17.17. Recommendation for k and h are 0.5 and 4 or 5,
respectively, for n = 1.
a)
CUSUM Chart of x
0.020
0.015
Cumulative Sum

16-71

UCL=0.01152

0.010
0.005
0.000

-0.005
-0.010

LCL=-0.01152
2

10

12 14
Sample

16

18

20

22

24

is estimated using the moving range: 0.0026/1.128=0.0023. H and K were computed using
k=0.5 and h=5. The process is not in control.
b) EWMA gives similar results.

16-50

EWMA Chart of x
0.404
UCL=0.403489
0.403

EWMA

0.402
_
_
X=0.401184

0.401

0.400

0.399

LCL=0.398879
2

16-72

10

12 14
Sample

16

18

20

22

24

a) Let p denote the probability that a point plots outside of the control limits when the mean has shifted from
0 to = 0 + 1.5. Then,

3
3

P( LCL < X < UCL) = P 0

< X < 0 +

n
n

1.5

X
1.5
= P
3<
<
+ 3
/ n / n
/ n

= P(6 < Z < 0) = P( Z < 0) P( Z < 6)

= 0.5 [0] = 0.5

Therefore, the probability the shift is undetected for three consecutive samples is (1 p)3 =
(0.5)3 = 0.125.
b) If 2-sigma control limits were used, then

2
2

1 p = P ( LCL < X < UCL ) = P 0

< X < 0 +
n
n

1.5
1.5
X
= P
2<
<
+ 2
/ n / n

/ n
= P (5 < Z < 1) = P ( Z < 1) P ( Z < 5)
= 1 0.84134 [0] = 0.15866
Therefore, the probability the shift is undetected for three consecutive samples is (1 p)3 = (0.15866)3 =
0.004.
c) The 2-sigma limits are narrower than the 3-sigma limits. Because the 2-sigma limits have a smaller
probability of a shift being undetected, the 2-sigma limits would be better than the 3-sigma limits for a mean
shift of 1.5. However, the 2-sigma limits would result in more signals when the process has not shifted
(false alarms).

16-51

16-73.

ARL = 1/p where p is the probability a point falls outside the control limits.
a) = 0 + and n = 1

p = P( X > UCL) + P( X < LCL)

3

0 +
0

=P Z>

/ n

0
0

n
+ P Z <

/ n

= P( Z > 3 n ) + P( Z < 3 n )
= P( Z > 2) + P( Z < 4)
= 1 0.97725 + [0]
= 0.02275

when n = 1

Therefore, ARL = 1/p = 1/0.02275 = 43.9.

b) = 0 + 2

P( X > UCL) + P( X < LCL )

3

0 +
0 2

=P Z>

/ n

0
0 2

n
+ P Z <

/ n

= P ( Z > 3 2 n ) + P ( Z < 3 2 n )
= P( Z > 1) + P ( Z < 5)
when n = 1
= 1 0.84134 + [0]
= 0.15866
Therefore, ARL = 1/p = 1/0.15866 = 6.30.
c) = 0 + 3

P( X > UCL) + P ( X < LCL)

3

0 +
0 3

=P Z>

/ n

0
0 3

n
+ P Z <

/ n

= P( Z > 3 3 n ) + P( Z < 3 3 n )
= P( Z > 0) + P( Z < 6)
when n = 1
= 1 0.50 + [0]
= 0.50
Therefore, ARL = 1/p = 1/0.50 = 2.00.
d) The ARL is decreasing as the magnitude of the shift increases from to 2 to 3. The ARL decrease as
the magnitude of the shift increases since a larger shift is more likely to be detected earlier than a smaller
shift.

16-52

16-74 a) Because ARL = 370, on the average we expect there to be one false alarm every 370 hours. Each 30-day
month contains 30 24 = 720 hours of operation. Consequently, we expect 720/370 = 1.9 false alarms each month

P( X > X + 3 ) + P( X < X 3 ) = P( z > 3) + P( z < 3) = 2(0.00135) = 0.0027

ARL=1/p=1/0.0027=370.37
b) With 2-sigma limits the probability of a point plotting out of control is determined as follows, when
= 0 +
P( X > UCL) + P( X < LCL)
X 0 0 + 2 0
X 0 0 2 0
= P
>
<
+ P

= P( Z > 1) + P( Z < 3)
= 1 P( Z < 1) + [1 P( Z < 3)]
= 1 0.84134 + 1 0.99865
= 0.160
Therefore, ARL=1/p = 1/0.160 = 6.25. The 2-sigma limits reduce the ARL for detecting a shift in the
mean of magnitude . However, the next part of this solution shows that the number of false alarms increases
with 2-sigma limits.

c) 2 limits

P( X > X + 2 ) + P( X < X 2 ) = P( z > 2) + P( z < 2) = 2(0.02275) = 0.0455

AR L= 1/ p= 1/0.0455 = 21.98. This ARL is not satisfactory. There would be too many false alarms. We
would expect 32.76 false alarms per month.
16-75

a)
X-bar and Range - Initial Study
Charting xbar
X-bar
| Range
----| ----UCL: +
3.0 sigma = 140.168
| UCL: +
3.0 sigma = 2.48437
Centerline
= 139.49
| Centerline
= 1.175
LCL: 3.0 sigma = 138.812
| LCL: 3.0 sigma = 0
out of limits = 9
| out of limits = 0
Estimated
process mean = 139.49
process sigma =
0.505159
mean Range
=
1.175
Pro b lem 16-51
141
140.168

140
X-b ar

139.49
139

138.812

138
137
0

12

16

20

2.5

2.48437

2
R an g e

1.5
1.175

1
0.5
0

0
0

8
12
su b g rou p

16

20

There are points beyond the control limits. The process is out of control. The points are 8, 10, 11,
12, 13, 14, 15, 16, and 19.

16-53

b) Revised control limits are given in the table below:

X-bar and Range - Initial Study
Charting Xbar
X-bar
| Range
----| ----UCL: +
3.0 sigma = 140.417
| UCL: +
3.0 sigma = 2.595682
Centerline
= 139.709
| Centerline
= 1.227273
LCL: 3.0 sigma = 139.001
| LCL: 3.0 sigma = 0
out of limits = 0
| out of limits = 0

Xb ar C h art
140.5

U C L= 140.4

140.0

xbar

M ean= 1 39.7
139.5

139.0

LC L= 139.0
0

10

S u b g ro u p N u m b e r

R C h a rt
3 .0
U C L = 2 .5 9 6

1 .5

M e a n= 1 .2 2 7

LC L=0

0
0

10

S u b g ro u p N u m b e r

There are no further points beyond the control limits.

The process standard deviation estimate is given by = R = 1.227273 = 0.5276

d2

2.326

c) PCR = USL LSL = 142 138 = 1.26

6
6(0.528)

142 139.709 139.709 138

USL x x LSL
PCRk = min
,
= min 3(0.528) , 3(0.528)

3
3

= min[1.45,1.08] = 1.08
Because the process capability ratios are less than unity, the process capability appears to be poor. PCR is
slightly larger than PCRk indicating that the process is somewhat off center.

16-54

d) In order to make this process a six-sigma process, the variance 2 would have to be decreased such that
PCRk = 2.0. The value of the variance is found by solving PCRk =

139.709 138
= 2.0
3
6 = 139.709 138
139.709 138
=
6
= 0.2848

x LSL
= 2.0 for :
3

Therefore, the process variance would have to be decreased to 2 = (0.2848)2 = 0.081.

e)  x = 0.528

p = P (139.001 < X < 140.417 | = 139.7)

139.001 139.7 X 140.417 139.7

= P
<
<
x
0.528
0.528

= P (1.32 < Z < 1.35)

= P ( Z < 1.36) P ( Z < 1.32)
= 0.913085 0.093418
= 0.8197

The probability that this shift will be detected on the next sample is 1 p = 10.8197 = 0.1803.

ARL =

1
1
=
= 5.55
1 p 0.1803

16-76
a) The probability of having no signal is

P ( 3 < X < 3) = 0.9973

P(No signal in 3 samples)=(0.9973)3=0.9919

P(No signal in 6 samples)=(0.9973)6= 0.9839
P(No signal in 10 samples)=(0.9973)10= 0.9733
16-77

PCR = 2 but = USL + 3

( 3 )

P( X < USL) = P Z <

= P( Z < 3) = 0.00135

16-78

a) The P(LCL < P < UCL), when p = 0.08, is needed.

LCL = p 3

p (1 p )
0.05(1 0.05)
= 0.05 3
= 0.015 0
n
100

UCL = p + 3

p (1 p )
0.05(1 0.05)
= 0.05 + 3
= 0.115
n
100

Therefore, when p = 0.08

16-55

0.115 0.08
P 0.08

P (0 P 0.115) = P ( P 0.115) = P

0.08(0.92)
0.08(0.92)

100
100

= P ( Z 1.29) = 0.90
Using the normal approximation to the distribution of P . Therefore, the probability of detecting the shift
on the first sample following the shift is 1 0.90 = 0.10.
b) The probability that the control chart detects a shift to 0.08 on the second sample, but not the first, is
(0.90)0.10= 0.09. This uses the fact that the samples are independent to multiply the probabilities.
c) p = 0.10

0.115 0.10
P 0.10

P (0 P 0.115) = P ( P 0.115) = P

0.10(0.90)
0.10(0.90)

100
100

= P ( Z 0.5) = 0.69146
from the normal approximation to the distribution of P . Therefore, the probability of detecting the shift on
the first sample following the shift is 1 0.69146 = 0.30854.
The probability that the control chart detects a shift to 0.10 on the second sample after the shift, but not the
first, is 0.69146(0.30854) = 0.2133.
d) A larger shift is generally easier to detect. Therefore, we should expect a shift to 0.10 to be detected
quicker than a shift to 0.08.

16-79

u=8
a) n = 4

UCL = u + 3

u
8
=8+3
= 12.24
n
4

u
8
=83
= 3.76
n
4

12.24 16
P( U > 12.24 when = 16) = P Z >

16

LCL = u 3

= P(Z > 1.88) = 1 P(Z < 1.88)

= 1 0.03005= 0.96995
P( U

3.76 16
< 3.78) = P Z <

16

= P(Z < 6.12)

=0
So the probability is 0.96995.
b) n = 10

16-56

UCL = u + 3

8
u
= 8+3
= 10.68
10
n

LCL = u 3

8
u
= 83
= 5.32
10
n

10.68 16
= P(Z > 4.22) = 1
P(U > 10.68 when = 16) = P Z >

16

10
So the probability is 1.

16-80

u = 10
a) n = 1

UCL = u + 3

10
u
= 10 + 3
= 19.49
1
n

LCL = u 3

10
u
= 10 3
= 0.51
1
n

19.94 14
P( U > 19. 94 when = 14) = P Z >

14
= P(Z > 1.47)= 1 P(Z < 1.47) = 1 0.9292 = 0.0708
and
P( U < 0.51) =

0.51 14

P Z <
=0
14

b) n = 4

UCL = u + 3

u
10
= 10 + 3
= 14.74
n
4

u
10
= 10 3
= 5.26
n
4

14
.
74

14
P( U > 14.74 when = 14) =
= P(Z >0.40) = 1 0.6554 = 0.3446
P Z>

14

5
.
26

14
= P(Z < 4.67) = 0
P( U < 5.26 when = 14) = P Z <

14

LCL = u 3

16-57