23_045191 ch15.

qxd 8/16/06 11:02 PM Page 245

Chapter 15: Constructing Control Plans and Charts 245

7. Shipments of washers for a pump assembly 8. At the end of a cellphone manufacturing
process come in deliveries of 10,000 items. process, each phone is put through a final
When each shipment is received, a sample functional test. The unit either “passes” or
of 100 washers is measured for thickness. “fails.” The number of total units produced
What type of control chart should you and the number passing the final test are
use to track the ongoing thickness of the recorded each hour. Select the right type
washers? of control chart for tracking each hour’s
production performance.
Solve It
Solve It

Interpreting Your Control Charts

Selecting and creating the right control chart is only a start. After creating the chart, the real
point is to correctly interpret it. What you’re looking for is the evidence of special causes,
which are the out-of-the-ordinary events that throw performance off its normal level.

To interpret a control chart, the plotted sequence of points on the chart is compared to the
calculated control limits (more on calculating control limits later in this chapter). When the
position or sequence of the points, in relation to each other or to the control limits, follows a
non-normal pattern, you know that a special cause has occurred.

Table 15-1 provides a summary of the non-normal patterns to look for in control charts.

Table 15-1 Tests for Special Causes for Rules 1 through 6

Chart Description Example 1 Example 2 Interpretation
Stable and Chart points don’t 20
The process is stable,
20
predictable form a particular 15
not changing. Only
15
pattern and they 10
common-cause
10
lie within the variation is affecting
upper and lower the process.
control limits.
(continued)
23_045191 ch15.qxd 8/16/06 11:02 PM Page 246

246 Part V: Improving and Controlling

Table 15-1 (continued)

Chart Description Example 1 Example 2 Interpretation
Beyond One or more chart 20
Alerts you that a
20
control points lie beyond 15
special cause has
15
limits the upper and 10
affected the process.
10
lower control Investigate to
limits. determine the source
of the special cause.
Run Chart points are Suggests that the
20 20
on one side of process has
15 15
the center line. undergone a
10 10
The number of permanent change.
consecutive May require you to
points on one side compute new control
is the “length” of limits for the shifted
the run. process.
Trend A continued rise 7 1
Indicates a special
20 20 2
or fall in a series 2 3
4
5
6 3
4
5
6
cause with a gradual,
15 15
of chart points. 1 7
cumulative effect.
10 10
(Seven or more Investigate possible
consecutive special cause
points in the same sources.
direction.)
Cycle Chart points show Indicates a special
20 20
the same pattern cause with a cyclical,
15 15
changes (for repetitive effect.
10 10
example, rise or Investigate possible
fall) over equal special cause
periods of time. sources.
Hugging Chart points are Suggests a possible
20 20
close to the 10
error in data sub-
15 15
center line or to a 10
grouping or selection.
10
control limit line. 10
Verify validity of
sampling plan and/or
investigate possible
special cause sources.

To be more specific and clinical in finding evidence of special cause sources, you also divide
the distance between the central line of the control chart and each calculated control limit
into thirds, as shown in Figure 15-4.

UCL
Zone A: +3 sigma
Χρησιμοποιήστε τους κανόνες του
Sample Value

Zone B: +2 sigma
Figure 15-4:
Zone C: +1 sigma
παραπάνω πίνακα για να
Dividing the
control Zone C: -1 sigma αξιολογήσετε ένα διάγραμμα I-MR
region of the Zone B: -2 sigma και να αποφασίσετε κατά πόσον η
chart into
LCL
Zone A: -3 sigma
διεργασία είναι εκτός ελέγχου.
zones.
Sample Group or Time Sequence
23_045191 ch15.qxd 8/16/06 11:02 PM Page 247

Chapter 15: Constructing Control Plans and Charts 247

The distance above and the distance below the central limit, the third closest to the
central line, is called Zone C. The middle third is called Zone B and the third farthest
from the central line is called Zone A. Here are eight specific rules, based on the three
zones, that indicate special causes:

1. Any point beyond either control limit

2. Nine points in a row on the same side of the center line
3. Six points in a row, all increasing or all decreasing
4. Fourteen points in a row alternating up and down
5. Two out of any three consecutive points in Zone A and all three on the same side
of the central average
6. Four out of any five consecutive points in Zone B or A, and all five on the same
side of the central average
7. Fifteen points in a row in Zone C, on either side of the central average line
8. Eight points in a row without a point in Zone C, on either side of the center line.

Checking all of these rules manually for each point on the control chart is tedious
work. We will do it here in theses examples to illustrate the concepts, but in practice,
usually your computer software will compute the control limits for you from the
sample data and check for these rules automatically.

Q. For the control chart shown in the following figure, identify any evidence of special causes. State
which rule from the eight-point list applies.

15

14 UCL = 13.814
13

12

11
Value

10 X = 10.193

7
LCL = 6.572
6
5 10 15 20 25 30 35 40 45 50
Observation

A. The obvious evidence of the presence of a special cause is the two points at observation num-
bers 14 and 15 that are beyond the upper control limit, which is based on the first rule in the list
of eight. Going through all of the other seven rules is a bit tedious, but you’ll find one other rule
providing evidence of a special cause — number 5. Observations 14, 15, and 16 are all in Zone A
or beyond, which fits the bill for rule number 5. Clearly something out of the ordinary happened
in the process at this time.
23_045191 ch15.qxd 8/16/06 11:02 PM Page 249

Chapter 15: Constructing Control Plans and Charts 249

Constructing an Individuals and
Moving Range (I-MR) Chart
The individuals and moving range chart is probably the most widely used control chart
because it requires only a sequence of individual numerical measurements. A daily recording
of minimum temperature, a sequential accounting of sales figures, or an ordered history of
length measurements — all of these measurements can be readily charted using an I-MR chart.

As with all of the control charts for continuous data types, an I-MR chart is a dual chart,
which means you create a chart for the individuals and another chart with the same horizon-
tal time axis for the moving range of the data. Having a second chart to track the range
between the sequential points allows you even greater resolution at detecting special causes:
If any of the test rules is broken for either the individuals or for the moving range chart, you
have evidence of a special cause.

In a pinch, you can quickly create just the individuals part of the I-MR chart. However, you
should create both parts whenever you can.

You create the moving range part (MR) of the I-MR chart by plotting the moving range (MRi)
between each of the sequential observations in your data. With the plotted moving ranges
you draw a calculated central average line ( MR). You also calculate and draw an upper con-
trol limit (UCLMR) and a lower control limit (LCLMR) around the moving range data. You calcu-
late these values from the following formulas:

MR i = x i - x i - 1

MR = 1
k-1
_ MR 2 + MR 3 + f + MR k i where k is the number of points you are charting
UCL MR = 3.267MR
LCL MR = 0

You create the individuals part (I) of the I-MR chart by plotting the individual sequential
measurements (xi). You include a calculated center average line designated as X . You also
calculate and draw an upper control limit (UCLx) and a lower control limit (LCLx).

X = 1 _ x 1 + x 2 + f + x k i where k is the number of points you are charting

k
UCL x = X + 2.659MR
LCL x = X - 2.659MR

Q. Create an I-MR chart for the following monthly sales data collected over the last two years. Note
any signs of special causes:
Month Number Sales ($) Month Number Sales ($) Month Number Sales ($)
1 $1,030,821 9 $1,053,196 17 $930,029
2 $991,989 10 $994,710 18 $990,836
3 $1,001,260 11 $1,027,961 19 $994,533
4 $955,303 12 $988,566 20 $996,449
5 $1,009,530 13 $1,003,554 21 $1,025,286
6 $978,663 14 $968,384 22 $1,151,151
7 $963,531 15 $937,275 23 $981,045
8 $1,015,794 16 $1,097,691 24 $992,442
23_045191 ch15.qxd 8/16/06 11:03 PM Page 250

250 Part V: Improving and Controlling

A. A calculation table will help solve this problem. Create a table with a row for each month’s sales
dollars. In a new column, calculate the moving range between each of the sequential data points.
Your table will look like this:
Month Number Sales ($) MR Month Number Sales ($) MR
1 $1,030,821 XXX 13 $1,003,554 $14,988
2 $991,989 $38,832 14 $968,384 $35,171 0
3 $1,001,260 $9,272 1 15 $937,275 $31,109
Στον πίνακα 4 $955,303 $45,958 7 16 $1,097,691 $160,416
αυτό
5 $1,009,530 $54,227 17 $930,029 $167,662
επισημαίνονται
διάφορα 6 $978,663 $30,866 7 18 $990,836 $60,807
μικρολαθάκια 7 $963,531 $15,132 19 $994,533 $3,697
στο τελευταίο 8 $1,015,794 $52,262 3 20 $996,449 $1,917 6
δεκαδικό ψηφίο 9 $1,053,196 $37,403 2 21 $1,025,286 $28,836 7
της στήλης MR.
10 $994,710 $58,487 6 22 $1,181,430 $156,144
11 $1,027,961 $33,252 1 23 $981,045 $200,385
12 $988,566 $39,395 24 $992,442 $11,397
Now you simply plug the table values into the formulas to generate all the data for the I-MR chart.
Here are the calculations:

MR = 1
k-1
_ MR 2 + MR 3 + f + MR k i = 1 _ $38, 832 + $9, 272 + f + $11, 397i = $55, 983
24 - 1
UCL MR = 3.267MR = 3.267 $ $55, 983 = $182, 897
LCL MR = 0

X = 1 _ x 1 + x 2 + f + x k i = 1 _ $1, 030, 821 + $991, 989 + f + $992, 442i = $1, 004, 595
k 24
UCL x = X + 2.659MR = $1, 004, 595 + 2.659 $ $55, 983 = $1,153, 454
LCL x = X - 2.659MR = $1, 004, 595 - 2.659 $ $55, 983 = $855, 376
With these values calculated, you can draw your I-MR chart. Plot the points and add the center
and control limit lines for each part of the dual chart (next figure).

120,000
Στο διάγραμμα αυτό, η UCL = 1,153,454
Individual Value

110,000
κλίμακα του άξονα Υ
είναι λάθος: έπρεπε να 100,000 X = 1,004,595

είναι σε εκατομμύρια. 90,000

LCL = 855,736
2 4 6 8 10 12 14 16 18 20 22 24
Observation

200,000
UCL = 182,897
Moving Range

150,000

100,000

50,000 MR = 55,983

0 LCL = 0
2 4 6 8 10 12 14 16 18 20 22 24
Observation

When reviewing the chart for evidence of special causes, you’ll find that only the 22nd month, with
its individual value and the corresponding moving range (MR23), sends up a flag for violation of the
rules. So, you know that something out of the ordinary must have occurred in the 22nd month.
23_045191 ch15.qxd 8/16/06 11:03 PM Page 251

Chapter 15: Constructing Control Plans and Charts 251

11. Construct an I-MR chart for the following 12. An airport administrator found some his-
production line yield data. Review your fin- torical data that gives the number of air-
ished chart for evidence of special causes, craft landings per week at his airport.
and get some scrap paper handy. During the period of this found data, the
administrator’s peers believe that a drop in
Production Date Order Yield aircraft traffic occurred. Create an I-MR
1 0.80 chart of the data and determine whether
there is indeed any evidence of a signifi-
2 0.81 cant drop in traffic.
3 0.76
Week Landings
4 0.79
1 158
5 0.80
2 151
6 0.78
3 150
7 0.82
4 169
8 0.79
5 149
9 0.78
6 150
10 0.83
7 143
11 0.78
8 149
12 0.79
9 143
13 0.83
10 158
14 0.77
11 145
15 0.75
12 149
16 0.79
13 132
17 0.82
14 140
18 0.80
15 171
19 0.80
16 129
20 0.83
17 126
21 0.75
18 136
22 0.85
19 136
23 0.78
20 140
24 0.77
21 128
25 0.76
22 136
26 0.76
23 124
27 0.81
24 128
28 0.79
25 126
29 0.84
26 123
30 0.77
31 0.79 Solve It
32 0.80

Solve It
23_045191 ch15.qxd 8/16/06 11:03 PM Page 267

Chapter 15: Constructing Control Plans and Charts 267

j Identify any evidence of special causes in the control chart shown in the figure from the origi-
nal question earlier in this chapter.
Going carefully through all the numbered rules for each of the points in the chart, you find that
observation 12 concludes a trend of six points in a row increasing, which is a sign of a special
cause. Also, at observation 24, the charted point is beyond the control limits. This is another
indicator of special cause variation. These indicators are highlighted in the version of the con-
trol chart in the next figure.

0.35 1

0.30 UCL = 0.3015

0.25

0.20
Value

3
0.15 X = 0.1499

0.10

0.05

0.00 LCL = 0.0016

9 6 9 12 15 18 21 24 27 30
Observation

k Construct an I-MR chart for the production line yield data shown in the original question earlier
in this chapter. Review your finished chart for evidence of special causes.
A calculation table helps solve this problem. In a new column, calculate the moving range
between each of the sequential data points. Your table should look like this:
Production Date Order Yield MR Production Date Order Yield MR
1 0.80 17 0.82 0.04 3
2 0.81 0.01 18 0.80 0.02
3 0.76 0.05 19 0.80 0.00
4 0.79 0.03 20 0.83 0.03
5 0.80 0.00 1 21 0.75 0.08
Και στον
6 0.78 0.02 22 0.85 0.10
πίνακα αυτό
επισημαίνονται 7 0.82 0.04 23 0.78 0.07
μικρολαθάκια 8 0.79 0.03 24 0.77 0.01
στο τελευταίο 9 0.78 0.01 25 0.76 0.01
δεκαδικό ψηφίο 10 0.83 0.05 26 0.76 0.00
της στήλης MR.
11 0.78 0.04 5 27 0.81 0.05
12 0.79 0.00 1 28 0.79 0.02
13 0.83 0.04 29 0.84 0.05
14 0.77 0.06 30 0.77 0.07
15 0.75 0.02 31 0.79 0.02
16 0.79 0.03
4 32 0.80 0.01
Now you simply plug the table values into the formulas to generate all the data for the I-MR
chart. Here are the calculations:
23_045191 ch15.qxd 8/16/06 11:03 PM Page 268

268 Part V: Improving and Controlling

MR = 1
k-1
_ MR 2 + MR 3 + f + MR k i = 1 ^ 0.01 + 0.05 + f + 0.01h = 0.03
32 - 1
UCL MR = 3.267MR = 3.267 $ 0.03 = 0.11
LCL MR = 0

X = 1 _ x 1 + x 2 + f + x k i = 1 ^ 0.80 + 0.81 + f + 0.80h = 0.79

k 32
UCL x = X + 2.659MR = 0.79 + 2.659 $ 0.03 = 0.88
LCL x = X - 2.659MR = 0.79 - 2.659 $ 0.03 = 0.71
With these values calculated, you can draw your I-MR chart. Plot the points and add the center
and control limit lines for each part of the dual chart (see the next figure).

0.90
UCL = 0.8827
Individual Value

0.85

0.80 X = 0.7934

0.75

0.70 LCL = 0.7042

3 6 9 12 15 18 21 24 27 30
Observation

UCL = 0.1096
0.100
Moving Range

0.075

0.050
MR = 0.0335
0.025

0.000 LCL = 0
3 6 9 12 15 18 21 24 27 30
Observation

When reviewing the chart for evidence of special causes, note that everything looks good: You
have no runs, trends, or points beyond the control limits.

l An airport administrator found some historical data that gives the number of aircraft landings
per week at his airport. During the period of this found data, the administrator’s peers believe
that a drop in aircraft traffic occurred. Create an I-MR chart of the data (flip to the original ques-
tion shown earlier in the chapter to see the data) and determine whether there is indeed any
evidence of a significant drop in traffic.
A calculation table helps solve this problem. In a new column, calculate the moving range
between each of the sequential data points. Here’s what your table should look like:
Week Landings MR Week Landings MR
1 158 14 140 8
2 151 7 15 171 31
3 150 1 16 129 42
4 169 19 17 126 3
5 149 20 18 136 10
6 150 1 19 136 0
7 143 7 20 140 4
8 149 6 21 128 12
9 143 6 22 136 8
10 158 15 23 124 12
11 145 13 24 128 4
12 149 4 25 126 2
13 132 17 26 123 3
23_045191 ch15.qxd 8/16/06 11:03 PM Page 269

Chapter 15: Constructing Control Plans and Charts 269

Now you simply plug the table values into the formulas to generate all the data for the I-MR
chart. Here are the calculations:

MR = 1
k-1
_ MR 2 + MR 3 + f + MR k i = 1 ^7 + 1 + f + 3 h = 10.2
26 - 1
UCL MR = 3.267MR = 3.267 $ 10.2 = 33.3
LCL MR = 0

X = 1 _ x 1 + x 2 + f + x k i = 1 ^158 + 151 + f + 123h = 141.9

k 26
UCL x = X + 2.659MR = 141.9 + 2.659 $ 10.2 = 169.0
LCL x = X - 2.659MR = 141.9 - 2.659 $ 10.2 = 114.8
With these values calculated, you can draw your I-MR chart. Plot the points and add the center
and control limit lines for each part of the dual chart (see the following figure).

1
168 UCL = 169.1
2
Individual Value

156 2
2 2
144 X = 141.88
132
2 2
120 2
LCL = 114.76
3 6 9 12 15 18 21 24
Observation
1
40
UCL = 33.33
Moving Range

30

20

10 MR = 10.2

0 LCL = 0
3 6 9 12 15 18 21 24
Observation

You can see by studying your I-MR chart that an out-of-the-ordinary change in airport traffic defi-
nitely occurred during this period. Observations 1 through 12 are all above the center line. That
fits rule number two: Nine or more points in a row are along one side of the center line. The same
rule is detected again starting with observation 16. But this time, the run of points are all beneath
the center line. It looks like airport traffic was at a higher level — a spike in traffic occurred coin-
cident with observation 15 and then a new, lower level of traffic emerged afterward.

m As an in-process check, a fruit juice bottler takes a three-item sample of caps every ten minutes
and measures their height. (If the height of the caps varies too much, the cap feed machines
have to be adjusted.) The original question shown earlier in this chapter contains a sequence
of 30 three-item samples of cap height measurements.
Create an X -R control chart for the cap height data and investigate whether any special causes
are present.
For each hourly sample of n = 3 points, calculate the sample average X and the range R. Your
table will look like this:
Sample X1 X2 X3 X R
1 19.86 19.74 20.07 19.89 0.33
2 19.98 20.07 19.94 20.00 0.13
3 20.02 19.99 20.06 20.02 0.07
4 19.95 19.80 20.00 19.92 0.20