Scribd
Upload
7 views

Advanced Control Charts

The document discusses advanced control charts used in industry, including stabilized control charts, exponentially weighted moving average (EWMA) charts, and cumulative sum control charts (CuSum). It highlights the adaptation of classical statistical process control methods for short production runs and explains how each type of chart functions, including their formulas and applications. The document emphasizes the importance of these charts in detecting process shifts and maintaining quality control in manufacturing.

Uploaded by

Ismail Saiel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
7 views

Advanced Control Charts

The document discusses advanced control charts used in industry, including stabilized control charts, exponentially weighted moving average (EWMA) charts, and cumulative sum control charts (CuSum). It highlights the adaptation of classical statistical process control methods for short production runs and explains how each type of chart functions, including their formulas and applications. The document emphasizes the importance of these charts in detecting process shifts and maintaining quality control in manufacturing.

Uploaded by

Ismail Saiel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

ADVANCED CONTROL CHARTS

In addition to the control charts already introduced, several other types of control
charts are used in industry. This section briefly reviews three of them: stabilized con-
trol charts, exponentially weighted moving average (EWMA) charts, and cumulative
sum control charts. The reader is encouraged to study more advanced books on sta-
tistical quality control for further details on these and other types of control charts.

Short Production Runs and Stabilized Control Charts


Control charts were developed for high-volume manufacturing situations in which
production runs lasted for weeks or months. In some industries, particularly as the
pressures for increased manufacturing flexibility increase, short production runs,
perhaps for only a few hours, are common. In such situations it may be impossible to
collect enough samples to compute control limits. Even if it is possible, by the time
the data are collected and the chart is constructed, the production run might be over,
thus defeating the purpose of the chart.
Fortunately, classical SPC methods can often be modified to apply to short pro-
duction runs. Three approaches can be used for variables data. First, tables of special
control chart constants for control limits compensate for the fact that a limited
number of samples are available. As more data become available, this approach
updates control limits until no further updates are needed and standard control
chart factors can be used. A second approach is to “code” the data by subtracting the
nominal value from the actual measurements. For example, consider a drill press
in which each run requires varying depths of cut. Instead of measuring values of
the actual depth of cut for a particular part, one can measure the deviation of depth
from the target. In this way, differences between products and production runs are
removed. In effect, this approach monitors process characteristics rather than product
characteristics.
Finally, the data can be transformed so that they are independent of the unit of
measure; such charts are called stabilized control charts. The idea is similar to the
familiar statistical concept of transforming normally distributed random variables
to a normal random variable with mean 0 and variance 1 by subtracting the mean
– –
and dividing by the standard deviation. Common transformations are (x– – x )/R and
– – –
R/R. For the x–-chart, control limits on the transformation (x– – x )/R are UCL = A2 and
–
LCL = –A2. Control limits for the R-chart based on the transformation R/R are UCL =
D4 and LCL = D3.
Stabilized charts can be developed for any type of attribute chart by using the
transformation

z = (sample statistic – process average)/standard deviation

For example, to develop a stabilized p-chart, use

z = (p – p–)/s p

Because z is measured in standard deviations, the upper and lower control limits are
+3 and –3, respectively. Readers are encouraged to consult Pyzdek’s book cited in the
bibliography for further information.

1
EWMA Charts
The exponentially weighted moving average (EWMA) chart was introduced for ap-
plications in chemical and process industries in which only one observation per time
period may be available. These measurement applications are the same situations in
which charts for individuals are used, except that the EWMA chart incorporates in-
formation on all the past data, not simply the last observation. The term exponentially
weighted refers to the fact that the data are weighted, with more weight being given to
the most recent data. (You may have studied exponential smoothing as a forecasting
technique; the same principle applies.)
The statistic that is plotted on the chart is

zt = αx–t + (1 – α)zt – 1

Here, zt is the exponentially weighted moving average after observation t is taken; x–t
is the value of observation t; zt – 1 is the previous exponentially weighted moving av-
erage; and α is a weighting factor between 0 and 1. This formula can be written in an
alternate fashion:

zt = zt – 1 + α(x–t – zt – 1)

which states that the current value of the statistic is equal to the previous value plus
some fraction of the difference between the current observation and its last estimate.
Note that when α = 1, the formula reduces to the ordinary x–-chart.
The standard error of the exponentially weighted moving average is

α σx α
σz = σx– =
t 2 – α 
n 2 – α

Therefore the control limits are given by

– σ α
UCLz = x + 3 x
 n 2 – α

– σx α
LCLz = x – 3

n 2 – α

The EWMA chart is more sensitive to small process level shifts than x–- or indi-
vidual charts. The smaller the value of α, the more easily are smaller shifts detectable.
This chart is useful when the acceptable process limits are narrow. However, this sen-
sitivity can lead to an excessive number of unnecessary adjustments to the process
and, consequently, unnecessary costs.

Cumulative Sum Control Charts


The cumulative sum control chart (CuSum chart) was designed to identify small but
sustained shifts in a process level much more quickly than ordinary x–-charts. Because
it gives an early indication of process changes, it is consistent with the management
philosophy of doing it right the first time and not allowing the production of non-
conforming products.

2
The CuSum chart incorporates all past data by plotting cumulative sums of the
deviations of sample values from a target value; that is,

St = (x–i – x–0)

where x–i is the average of the ith subgroup, x–0 is the standard or reference value, and
St is the cumulative sum when the ith observation is taken. Note that when n = 1, x–i
is the value of the ith observation.
The CuSum chart looks different from ordinary x–- and R-charts. In place of a cen-
ter line and horizontal control limits, a “mask” is constructed that consists of a loca-
tion pointer and two angled control limits as illustrated in Figure ACC.1. The mask is
located on the chart so that the point p lies on the last point plotted. The distance d
and the angle θ are the design parameters of the mask. (This text does not discuss
how these are computed. Readers are referred to Chapter 10 of Grant and Leaven-
worth or Chapter 7 of Montgomery [both cited in the bibliography at the end of the
chapter] for details.)
If no previous points lie outside the control limits, the process is assumed to be in
control. If, for example, a shift in the process mean raises it above the reference value,
each new value added to the cumulative sum will cause St to increase and result in an
upward trend in the chart. Eventually a point will fall outside the upper control limit,
indicating that the process has fallen out of control (illustrated in Figure ACC.2). The
opposite will occur if the mean shifts downward.

Figure ACC.1 CuSum Chart for Sample Averages

St

15

10
Lower control limit
5

0
pθ
–5
d
–10

–15
Upper control limit

1 2 3 4 5 10 15 20
Sample Number

3
Figure ACC.2 CuSum Chart Illustration of Lack of Control

St

20
15
10
Points outside upper
5 control limit
0
–5
–10
–15
–20

Sample Number

You might also like