國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
37
CHAPTER 5. DETERMINING THE BEST λ OF THE ECONOMIC
EWMA
X-barCONTROL CHART UNDER DIFFERENT
SHIFT SCALES IN THE MEAN AND VARIANCE
5.1 Data Description and Determining the Optimum Producer Inspection and the Design Parameters of the Economic EWMA
X-barControl Chart
In this section, we use routine “rgamma” to simulate one type of in-control gamma distribution and five types of out-of-control gamma distributions.
We first simulate 25 samples of size 4 data from the in-control gamma distribution with a=25 and b=0.2, with a mean and variance of 5 and 1, respectively.
Table 5-1. In-control Data with Gamma (a=25, b=0.2)
No. simulation data with n=4
X
No. simulation data with n=4X
1 6.283 6.397 4.894 5.691 5.816 14 5.849 6.036 5.879 3.814 5.394 2 4.071 5.513 4.908 5.222 4.929 15 6.359 4.882 4.89 2.939 4.767 3 5.222 4.037 5.591 5.076 4.982 16 6.288 3.448 6.242 2.928 4.727 4 2.54 5.11 3.753 5.155 4.14 17 5.595 5.352 4.419 4.756 5.03 5 6.439 4.634 5.628 4.692 5.348 18 5.124 4.054 5.542 4.358 4.77 6 5.2 5.276 3.601 2.725 4.201 19 5.816 4.549 5.241 5.435 5.26 7 5.265 4.934 5.253 3.719 4.793 20 4.593 4.515 5.584 5.111 4.951 8 3.954 5.004 5.628 6.364 5.237 21 6.918 5.952 5.245 4.799 5.728 9 4.04 3.106 5.334 4.119 4.15 22 4.881 4.955 6.066 4.761 5.166 10 4.102 7.045 4.045 5.118 5.077 23 4.525 5.358 5.983 3.144 4.753 11 4.317 4.943 5.044 4.384 4.672 24 5.442 4.729 5.419 5.011 5.15 12 3 5.595 5.794 4.064 4.613 25 7.023 5.363 5.197 3.921 5.37613 4.992 5.155 7.028 5.743 5.729
X 4 . 99
As a producer, we do not know the parameters of the gamma distribution; thus, we used the MLE method, which maximizes cumulative products of p.d.f for given data to estimate
aˆ
andbˆ of the simulation data.
With 100 data in Table 5-1, we estimate the parameters of in-control data, and obtain
a ˆ
I 24 . 349
,b
ˆI
0.205, Mean a
ˆI b
ˆI
4.99︿
, and
Var a ˆ
I b ˆ
I2 1 . 023
︿
.
‧
subject to 0.5≦h≦8. If the producer decides to inspect, we maximize EAP (Equation 4-7) to determine the optimum h* and ω*, subject to 0.5≦h≦8 and 2≦ω.I. To compare the profit model with different λ, we adopt moderate shifts in the mean and variance of out-of-control gamma data.
Let a=26 and b=0.25, that is, the out-of-control mean and variance are 6.5 and 1.625, respectively, which means δ1=1 and δ2=0.05. We simulate 15 samples of size 4 data from the out-of-control gamma distribution with a=26 and b=0.25, as follows:
Table 5-2. Out-of-control Data with Gamma (a=26, b=0.25)
No. simulation data with n=4
X
With 60 data in Table 5-2, we estimate the parameters of out-of-control data, and obtain
a ˆ
O 25 . 268
,b
ˆO
0.265 ,ˆ 0 . 919
‧
Table 5-3. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.6 0.05 1 0.6 0.05 According to Table 5-3, with and without inspection, h*, EWMAX-bar chart, andARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.05, we increase 54.9% profit per unit time when we have an inspection.
(2) If λ=0.6, we increase 42.5% profit per unit time when we have an inspection.
(3) If λ=1, we increase 44.2% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.6, we have the largest EAP*
and smallest ARL1 for the moderate shifts in the mean and variance. Therefore, we suggest that the producer takes inspection with USL*=8.66, use the economic EWMAX-bar chart with λ=0.6 and take four samples every 0.5 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
40
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-1. The Economic EWMAX-bar Chart (λ=1) with In-control Data
For EWMAX-bar chart with λ=1, Figure 5-1 shows that no points are out of limits for in-control samples.
Figure 5-2. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-2 shows that No. 6, 8, 10, 12, and 15 are out of limits; the first true alarm is on No. 6.
Figure 5-3. The Economic EWMAX-bar Chart (λ=0.6) with In-control Data For EWMAX-bar chart with λ=0.6, Figure 5-3 shows that no points are out of limits for in-control samples.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
41
Figure 5-4. The Economic EWMAX-bar Chart (λ=0.6) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.6, Figure 5-4 shows that No. 3 to No. 15 are out of limits; the first true alarm is on No. 3.
Figure 5-5. The Economic EWMAX-bar Chart (λ=0.05) with In-control Data For EWMAX-bar chart with λ=0.05, Figure 5-5 shows that no points are out of limits for in-control samples.
Figure 5-6. The Economic EWMAX-bar Chart (λ=0.05) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.05, Figure 5-6 shows that No. 4 to No. 15 are out of limits; the first true alarm is on No. 4.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
42
II. To compare the profit model with different λ, we adopt small shifts in the mean and variance of out-of-control gamma data.
Let a=28 and b=0.21, that is, the out-of-control mean and variance are 5.88 and 1.2348, respectively, which means δ1=3 and δ2=0.01. We simulate 15 samples of size 4 data from the out-of-control gamma distribution with a=28 and b=0.21, as follows:
Table 5-4. Out-of-control Data with Gamma (a=28, b=0.21)
No. simulation data with n=4
X
1 6.156 5.224 7.595 6.145 6.28
2 5.026 4.593 8.429 6.001 6.012
3 5.205 5.889 4.731 3.281 4.776
4 4.828 5.989 4.256 4.911 4.996
5 7.144 7.129 5.382 6.513 6.542
6 5.664 5.578 6.608 6.595 6.111
7 7.939 4.867 5.966 6.424 6.299
8 5.524 5.661 7.156 6.023 6.091
9 5.271 5.992 5.396 7.1 5.94
10 7.212 5.952 7.16 5.058 6.345
11 7.683 5.009 6.831 5.237 6.19
12 6.226 4.889 4.416 5.096 5.157
13 5.478 4.99 5.898 6.196 5.641
14 6.961 6.639 4.215 6.964 6.195
15 5.797 7.595 7.614 7.204 7.053
975 .
5 X
With 60 data in Table 5-4, we estimate the parameters of out-of-control data, and obtain
a ˆ
O 30 . 883
,b
ˆO
0.193 ,ˆ 6 . 534
1
,ˆ 0 . 012
2
, Mean︿
5.975 , and156 . 1
Var︿
. Hence, we have a 0.974 mean shift scale and a 1.063 s.d. shift scale.‧
same as the X-bar probability chart. We compare the optimum results of profit model with λ=1, 0.5, and 0.05, as follows:Table 5-5. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.5 0.05 1 0.5 0.05 According to Table 5-5, with and without inspection, h*, EWMAX-bar chart, andARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.05, we increase 1.66% profit per unit time when we have an inspection.
(2) If λ=0.5, we increase 1.6% profit per unit time when we have an inspection.
(3) If λ=1, we increase 1.85% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.5, we have the largest EAP*
and smallest ARL1 for the small shifts in the mean and variance. Therefore, we suggest that the producer takes inspection with USL*=8.66, use the economic EWMAX-bar chart with λ=0.5 and take four samples every 0.5 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
44
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-7. The Economic EWMAX-bar Chart (λ=1) with In-control Data
For EWMAX-bar chart with λ=1, Figure 5-7 shows that no points are out of limits for in-control samples.
Figure 5-8. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-8 shows that No. 15 is out of limits; the first true alarm is on No. 15.
Figure 5-9. The Economic EWMAX-bar Chart (λ=0.5) with In-control Data For EWMAX-bar chart with λ=0.5, Figure 5-9 shows that no points are out of limits for in-control samples.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
45
Figure 5-10. The Economic EWMAX-bar Chart (λ=0.5) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.5, Figure 5-10 shows that No. 6 to No. 11, 14, and 15 are out of limits; the first true alarm is on No. 6.
Figure 5-11. The Economic EWMAX-bar Chart (λ=0.05) with In-control Data For EWMAX-bar chart with λ=0.05, Figure 5-11 shows that no points are out of limits for in-control samples.
Figure 5-12. The Economic EWMAX-bar Chart (λ=0.05) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.05, Figure 5-12 shows that No. 7 to No. 15 are out of limits; the first true alarm is on No. 7.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
46
III. To compare the profit model with different λ, we adopt only moderate shifts in the mean of out-of-control gamma data.
Let a=42.45 and b=0.154, that is, the out-of-control mean and variance are 6.5 and 1, respectively, which meansδ1=17.25 and δ2=-0.046. We simulate 15 samples of size 4 data from the out-of-control gamma distribution with a=42.45 and b=0.154, as follows:
Table 5-6. Out-of-control Data with Gamma (a=42.45, b=0.154)
No. simulation data with n=4
X
1 5.211 6.206 8.77 6.833 6.755
2 6.232 6.273 7.556 6.781 6.71
3 7.759 6.392 4.876 6.144 6.293
4 8.373 7.25 6.74 5.986 7.087
5 5.014 7.998 5.377 5.547 5.984
6 7.65 5.385 5.822 6.622 6.37
7 7.911 6.213 10.067 6.721 7.728
8 7.416 5.252 6.553 6.646 6.467
9 6.259 7.502 5.828 7.118 6.677
10 4.747 5.193 7.432 6.134 5.876
11 5.911 7.173 6.807 6.404 6.574
12 7.335 6.388 6.383 6.643 6.687
13 6.699 8.889 5.93 5.889 6.852
14 6.41 4.786 5.894 7.576 6.167
15 6.871 6.969 7.435 5.815 6.773
6 .
6 X
With 60 data in Table 5-6, we estimate the parameters of out-of-control data, and obtain
a ˆ
O 41 . 331
,b
ˆO
0.16 ,ˆ 16 . 983
1
,ˆ 0 . 045
2
, Mean︿
6.6 , and05 . 1
Var︿
. Hence, we have a 1.603 mean shift scale and a 1.017 s.d. shift scale.‧
the X-bar probability chart. We compare the optimum results of profit model with λ=1, 0.6, and 0.05, as follows:Table 5-7. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.6 0.05 1 0.6 0.05first true alarm on which sample According to Table 5-7, with and without inspection, h*, EWMAX-bar chart, and
ARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.05, we increase 8.6% profit per unit time when we have an inspection.
(2) If λ=0.6, we increase 7.4% profit per unit time when we have an inspection.
(3) If λ=1, we increase 7.7% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.6, we have the largest EAP*
and smallest ARL1 for the only moderate shifts in the mean. Therefore, we suggest that the producer takes inspection with USL*=8.66, use the economic EWMAX-bar
chart with λ=0.6 and take four samples every 0.5 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
48
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-13. The Economic EWMAX-bar Chart (λ=1) with In-control Data For EWMAX-bar chart with λ=1, Figure 5-13 shows that no points are out of limits for in-control samples.
Figure 5-14. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-14 shows that No. 1, 2, 4, 7, 9, 12, 13, and 15 are out of limits; the first true alarm is on No. 1.
Figure 5-15. The Economic EWMAX-bar Chart (λ=0.6) with In-control Data For EWMAX-bar chart with λ=0.6, Figure 5-15 shows that no points are out of limits for in-control samples.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
49
Figure 5-16. The Economic EWMAX-bar Chart (λ=0.6) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.6, Figure 5-16 shows that No. 2 to No. 15 are out of limits; the first true alarm is on No. 2.
Figure 5-17. The Economic EWMAX-bar Chart (λ=0.05) with In-control Data For EWMAX-bar chart with λ=0.05, Figure 5-17 shows that no points are out of limits for in-control samples.
Figure 5-18. The Economic EWMAX-bar Chart (λ=0.05) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.05, Figure 5-18 shows that No. 3 to No. 15 are out of limits; the first true alarm is on No. 3.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
50
IV. To compare the profit model with different λ, we adopt only moderate shifts in the variance of out-of-control gamma data.
Let a=15.385 and b=0.325, that is, the out-of-control mean and variance are 5 and 1.625, respectively, which means δ1=-9.615 and δ2=0.125. We simulate 15 samples of size 4 data from the out-of-control gamma distribution with a=15.385 and b=0.325, as follows:
Table 5-8. Out-of-control Data with Gamma (a=15.385, b=0.325)
No. simulation data with n=4
X
1 5.783 4.26 7.003 4.471 5.379
2 5.485 4.479 7.158 3.814 5.234
3 4.143 7.228 7.417 4.916 5.926
4 6.487 5.547 2.772 4.822 4.907
5 5.967 4.88 5.348 3.423 4.904
6 4.063 3.395 5.573 2.929 3.99
7 5.252 4.555 6.222 4.037 5.016
8 3.769 4.488 2.513 4.25 3.755
9 4.227 7.095 4.826 5.706 5.463
10 4.821 4.497 6.032 4.128 4.87
11 5.136 4.276 7.218 5.864 5.624
12 4.27 4.185 7.084 5.688 5.307
13 6.958 3.505 6.282 4.767 5.378
14 4.593 3.912 4.769 7.98 5.313
15 5.699 7.59 4.129 5.879 5.824
126 .
5 X
With 60 data in Table 5-8, we estimate the parameters of out-of-control data, and obtain
a ˆ
O 15 . 608
,b
ˆO
0.328,ˆ 8 . 741
1
,ˆ 0 . 123
2
, Mean︿
5.126 , and684 . 1
Var︿
. Hence, we have a 0.126 mean shift scale and a 1.281 s.d. shift scale.‧
the X-bar probability chart. We compare the optimum results of profit model with λ=1, 0.3, and 0.05,, as follows:Table 5-9. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.3 0.05 1 0.3 0.05first true alarm on which sample
According to Table 5-9, with and without inspection, h*, EWMAX-bar chart, and
ARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.05, we increase 1.5% profit per unit time when we have an inspection.
(2) If λ=0.3, we increase 1.4% profit per unit time when we have an inspection.
(3) If λ=1, we increase 1.4% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.3, we have the largest EAP*
and smallest ARL1 for the only moderate shifts in the variance. Therefore, we suggest that the producer takes inspection with USL*=8.66, use the economic EWMAX-bar chart with λ=0.3 and take four samples every 0.5 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
52
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-19. The Economic EWMAX-bar Chart (λ=1) with In-control Data For EWMAX-bar chart with λ=1, Figure 5-19 shows that no points are out of limits for in-control samples.
Figure 5-20. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-20 shows that no points are out of limits; it has no true alarm.
Figure 5-21. The Economic EWMAX-bar Chart (λ=0.3) with In-control Data For EWMAX-bar chart with λ=0.3, Figure 5-21 shows that no points are out of limits for in-control samples.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
53
Figure 5-22. The Economic EWMAX-bar Chart (λ=0.3) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.3, Figure 5-22 shows that no points are out of limits; it has no true alarm.
Figure 5-23. The Economic EWMAX-bar Chart (λ=0.05) with In-control Data For EWMAX-bar chart with λ=0.05, Figure 5-23 shows that no points are out of limits for in-control samples.
Figure 5-24. The Economic EWMAX-bar Chart (λ=0.05) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.05, Figure 5-24 shows that no points are out of limits; it has no true alarm.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
54
V. To compare the profit model with different λ, we adopt only small shifts in the variance of out-of-control gamma data.
Let a=15.385 and b=0.325, that is, the out-of-control mean and variance are 5 and 1.625, respectively, which means δ1=-9.615 and δ2=0.125. We simulate 15 samples of size 4 data from the out-of-control gamma distribution with a=15.385 and b=0.325, as follows:
Table 5-10. Out-of-control Data with Gamma (a=15.385, b=0.325)
No. simulation data with n=4
X
1 3.461 5.507 6.388 3.679 4.759
2 3.152 3.912 3.297 4.57 3.733
3 4.175 3.619 3.699 5.961 4.363
4 5.62 5.381 6.398 3.62 5.255
5 4.68 3.814 5.188 5.141 4.706
6 3.949 5.531 6.154 5.891 5.381
7 5.054 4.684 6.065 5.079 5.22
8 6.155 3.538 3.748 4.434 4.469
9 7.259 5.163 5.59 5.38 5.848
10 6.283 6.908 5.931 6.023 6.287
11 4.941 4.361 6.621 3.688 4.903
12 4.891 5.098 5.064 4.271 4.831
13 4.771 5.773 7.21 4.154 5.477
14 4.146 6.678 4.142 4.289 4.814
15 5.892 5.977 5.689 4.045 5.401
03 .
5 X
With 60 data in Table 5-10, we estimate the parameters of out-of-control data, and obtain
a ˆ
O 22 . 087
,b
ˆO
0.228,ˆ 2 . 261
1
,ˆ 0 . 023
2
, Mean︿
5.03, and145 . 1
Var︿
. Hence, we have a 0.039 mean shift scale and a 1.058 s.d. shift scale.‧
same as the X-bar probability chart. We compare the optimum results of profit model with λ=1, 0.2, and 0.05, as follows:Table 5-11. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.2 0.05 1 0.2 0.05first true alarm on which sample
According to Table 5-11, with and without inspection, h*, EWMAX-bar chart, and
ARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.05, we increase 0.5% profit per unit time when we have an inspection.
(2) If λ=0.2, we increase 0.5% profit per unit time when we have an inspection.
(3) If λ=1, we increase 0.5% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.2, we have the largest EAP*
and smallest ARL1 for the only small shifts in the variance. Therefore, we suggest that the producer takes inspection with USL*=8.66, use the economic EWMAX-bar chart with λ=0.2 and take four samples every 0.5 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-25. The Economic EWMAX-bar Chart (λ=1) with In-control Data For EWMAX-bar chart with λ=1, Figure 5-25 shows that no points are out of limits for in-control samples.
Figure 5-26. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-26 shows that no points are out of limits; it has no true alarm.
Figure 5-27. The Economic EWMAX-bar Chart (λ=0.2) with In-control Data For EWMAX-bar chart with λ=0.2, Figure 5-27 shows that no points are out of limits for in-control samples.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
57
Figure 5-28. The Economic EWMAX-bar Chart (λ=0.2) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.2, Figure 5-28 shows that no points are out of limits; it has no true alarm.
Figure 5-29. The Economic EWMAX-bar Chart (λ=0.05) with In-control Data For EWMAX-bar chart with λ=0.05, Figure 5-29 shows that no points are out of limits for in-control samples.
Figure 5-30. The Economic EWMAX-bar Chart (λ=0.05) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=0.05, Figure 5-30 shows that no points are out of limits; it has no true alarm.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
58
VI. Comparing the profit model with different λ using an example of service time data.
We consider a quality variable with an exponential distribution, and take the service time data from Yang et al. (2012). The service time is an important quality characteristic for a bank branch in Taiwan. To measure the efficiency in the service system of a bank branch, the sampling service times (in minutes) are measured from 10 counters every 2days for 30days; that is, 15 samples of size n=10 are taken from an in-control service system. These data have been analyzed and have a right-skewed distribution, as shown in Table 5-12.
Table 5-12. In-control Service Time Data.
No. In-control service data with n=10
X
1 0.88 0.78 5.06 5.45 2.93 6.11 11.59 1.2 0.89 3.21 3.81 2 3.82 13.4 5.16 3.2 32.27 3.68 3.14 1.58 2.72 7.71 7.67 3 1.4 3.89 10.88 30.85 0.54 8.4 5.1 2.63 9.17 3.94 7.68 4 16.8 8.77 8.36 3.55 7.76 1.81 1.11 5.91 8.26 7.19 6.95 5 0.24 9.57 0.66 1.15 2.34 0.57 8.94 5.54 11.69 6.58 4.73 6 4.21 8.73 11.44 2.89 19.49 1.2 8.01 6.19 7.48 0.07 6.97 7 15.08 7.43 4.31 6.14 10.37 2.33 1.97 1.08 4.27 14.08 6.71 8 13.89 0.3 3.21 11.32 9.9 4.39 10.5 1.7 10.74 1.46 6.74 9 0.03 12.76 2.41 7.41 1.67 3.7 4.31 2.45 3.57 3.33 4.16 10 12.89 17.96 2.78 3.21 1.12 12.61 4.23 6.18 2.33 6.92 7.02 11 7.71 1.05 1.11 0.22 3.53 0.81 0.41 3.73 0.08 2.55 2.12 12 5.81 6.29 3.46 2.66 4.02 10.95 1.59 5.58 0.55 4.1 4.50 13 2.89 1.61 1.3 2.58 18.65 10.77 18.23 3.13 3.38 6.34 6.89 14 1.36 1.92 0.12 11.08 8.85 3.99 4.32 1.71 1.77 1.94 3.71 15 21.52 0.63 8.54 3.37 6.94 3.44 3.37 6.37 1.28 12.83 6.83
766 .
5 X
Since the 150 in-control data follows exponential distribution, we estimate the parameters of the exponential distribution, and obtain
b
ˆI
5.766, Mean︿
5.766, and Var︿
33.244. We use the routine “ks.test” to test in-control data with Kolmogorov-Smirnov test method and have a p-value = 0.6714; therefore, we do not reject the data drawn from the exponential distribution with bI=5.766.‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
59
The new data set of service times from a new automatic service system of the bank branch, 10 new samples of size 10, were collected and listed in Table 5-13.
The out-of-control service time data are as follows:
Table 5-13. Out-of -control Service Time Data.
No. Out-of-control service data with n=10
X
1 3.54 0.01 1.33 7.27 5.52 0.09 1.84 1.04 2.91 0.63 2.42 2 0.86 1.61 1.15 0.96 0.54 3.05 4.11 0.63 2.37 0.05 1.53 3 1.45 0.19 4.18 0.18 0.02 0.7 0.8 0.97 3.6 2.94 1.50 4 1.37 0.14 1.54 1.58 0.45 6.01 4.59 1.74 3.92 4.82 2.62 5 3 2.46 0.06 1.8 3.25 2.13 2.22 1.37 2.13 0.25 1.87 6 1.59 3.88 0.39 0.54 1.58 1.7 0.68 1.25 6.83 0.31 1.88 7 5.01 1.85 3.1 1 0.09 1.16 2.69 2.79 1.84 2.62 2.22 8 4.96 0.55 1.43 4.12 4.06 1.42 1.43 0.86 0.67 0.13 1.96 9 1.08 0.65 0.91 0.88 2.02 2.88 1.76 2.87 1.97 0.62 1.56 10 4.56 0.44 5.61 2.79 1.73 2.46 0.53 1.73 7.02 2.13 2.90045 .
2 X
With 100 data in Table 5-13, we estimate the parameters of out-of-control data, and obtain
b
ˆO
2.045,ˆ 3 . 72
2
, Mean︿
2.045, and Var.︿
4.184. Hence, we have a -0.645 small mean shift scale and a 0.355 small s.d. shift scale. During calculation, the out-of-control mean and variance are smaller than the in-control mean and variance.We use the routine “ks.test” to test out-of-control data and we have a
p-value=0.4182; therefore, we do not reject the data drawn from the exponential
distribution with bO=2.045.We let aI =1, n=10, θ=0.01, e=0.05, D=20, T=250, s0=5, s1=0.1, W=500, kc=10,
A=600, IC=0.1, P
C=300, PU=150, and R=200. If the producer decides not to inspect,we maximize EAP (Equation 3-12) to determine the optimum h*, subject to 0.5≦h≦8. If the producer decides to inspect, we maximize EAP (Equation 4-7) to
determine the optimum h* and ω*, subject to 0.5≦h≦8 and 2≦ω.
‧
Table 5-14. The Optimum Results Comparison of Profit Model with Different λ
Inspection Without With
λ
1 0.4 0.1 1 0.4 0.1EAP*
-42444.27 -43125.13 -36886.39 -5577.36 -5973.12 -2346.79P
W 292.532 292.532 292.532 - - -ARL
1 2.692.55
3.9 2.692.55
3.9UCL
12.786 8.972 6.98 12.786 8.972 6.98LCL
1.778 3.531 4.718 1.778 3.531 4.718first true alarm on which sample According to Table 5-14, with and without inspection, h*, EWMAX-bar chart, and
ARL
1 are the same at each λ. However, with inspection, we increased the profit per unit time as follows:(1) If λ=0.1, we increase 93.6% profit per unit time when we have an inspection.
(2) If λ=0.4, we increase 86.1% profit per unit time when we have an inspection.
(3) If λ=1, we increase 86.9% profit per unit time when we have an inspection.
If we use the economic EWMAX-bar chart with λ=0.1, we have the largest EAP*, but largest ARL1. If we use the economic EWMAX-bar chart with λ=0.4, we have the smallest EAP*, but smallest ARL1. To maximize EAP* for small shifts in the mean and variance we suggest that the producer takes inspection with USL*=17.297, use the economic EWMAX-bar chart with λ=0.1 and take 10 samples every 8 unit time.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
61
To find the detection ability for the three types of EWMAX-bar chart, we plot the in-control and out-of-control statistics on them.
Figure 5-31. The Economic EWMAX-bar Chart (λ=1) with In-control Data For EWMAX-bar chart with λ=1, Figure 5-31 shows that no points are out of limits for in-control samples.
Figure 5-32. The Economic EWMAX-bar Chart (λ=1) with Out-of-control Data Plots the out-of-control statistics on the EWMAX-bar Chart with λ=1, Figure 5-32 shows that No. 2, 3, and 9 are out of limits; the first true alarm is on No. 2.
Figure 5-33. The Economic EWMAX-bar Chart (λ=0.4) with In-control Data For EWMAX-bar chart with λ=0.4, Figure 5-33 shows that no points are out of
Figure 5-33. The Economic EWMAX-bar Chart (λ=0.4) with In-control Data For EWMAX-bar chart with λ=0.4, Figure 5-33 shows that no points are out of