PDHonline Course P209 (3 PDH)

Statistical Methods for Process

Improvement - Part 3: Using Process
Control Charts

Instructor: Davis M. Woodruff, PE, CMC

2012

PDH Online | PDH Center

5272 Meadow Estates Drive
Fairfax, VA 22030-6658
Phone & Fax: 703-988-0088
www.PDHonline.org
www.PDHcenter.com

An Approved Continuing Education Provider

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Statistical Methods for Process Improvement

Part 3: Using Process Control Charts

Davis M. Woodruff, PE, CMC

“Industrial processes are much like small children, they behave

better within well defined limits of behavior.” – M. J. Moroney

Introduction to Process Control Charts

The control chart is primarily a tool used to analyze data which is generated over a period of time.
Developed by Dr. Walter A. Shewhart in 1924, the control chart remains one of the most important
tools in the SPC arsenal. Although simple to use, they are very powerful for process analysis.

In Part 3 of Statistical Methods for Process Improvement we will learn about using several types of
control charts and how to interpret what a process is ‘telling’ us. However, before considering any
specific chart, let's examine control charts in general.

Process Control Charts

What are they? First of all, while the term "control chart" has been universally accepted and used, it
must be understood that the chart does not control anything. It only provides information which
serves as a basis for action. Control charts are effective only if those responsible for making
decisions act on the information the chart reveals. In other words, control charts are tools which can
be used, misused, or not used at all. Their effectiveness is directly related to the understanding of
and proper application by the people using them, you and your colleagues.

To adequately control any repetitive operation requires continuous monitoring. The control chart is
a graphical representation of this monitoring process.

How do they work? Control charts are all basically like the conceptual chart shown below. Along
the vertical axis a scale is established consistent with the data being collected. The horizontal axis
reflects the data sequence or time interval. As data is obtained, it can be plotted relative to these
two axes.

©2010 Davis M. Woodruff Page 2 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

The Conceptual Control Chart

UCL

Average or
Central Line

Measurement Scale

LCL

Time sequence

Since we normally obtain data by sampling, let's assume we take a certain sample size from a
source at regular intervals. Sample statistics (mean, median, range, fraction defective, etc.) are
calculated. We know from the Central Limit Theorem that with proper sampling, these will vary in a
specific way; namely, a normal distribution. By calculating the mean and standard deviation for
each statistic and applying our knowledge of the normal distribution, we can estimate, or predict the
group behavior of each sample statistic. This is done by calculating the grand mean and plotting it
on the chart as a solid central line. We then measure up 3 standard deviations to plot the upper
control limit and down 3 standard deviations to plot the lower control limit. These are shown on the
chart as dashed lines.

Actually, what we have constructed is the normal distribution curve of the sample statistic and its
±3σ limits. We know that for a normal distribution, the ±3σ limits represent 99.73% of the
distribution. We would expect that future samples would result in a sample statistic that would fall
within the upper and lower control limits 99.73% of the time. This is a realistic expectation only if
nothing has changed in the original distribution or source of the data.
What do they tell us? A control chart provides guides (control limits) which will indicate whether a
population has shifted. If the limits are representative of the process we are monitoring, they also
indicate the extent that common cause variation is influencing the distribution. Variation between
the upper and lower limits can be expected and considered normal. Operator efforts to reduce this
variation are ineffective; and in fact, such actions generally lead to more rather than less variation, a
condition known as over control. A chart operating within the limits is saying, common cause
variation at work--do not adjust. Since only common cause variation is present, special causes are
absent; and we conclude that the process is in statistical control.
©2010 Davis M. Woodruff Page 3 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Data points which fall outside the control limits are very unlikely. If this situation occurs, it is an
indication that the distribution has shifted. Such shifts are the result of special cause variation.
Because special cause variation is present, we conclude that the process is not in control.
Investigations should begin immediately to determine the source of the special cause variation so
that corrective action can be taken to bring the process back under control.

Control charts simply tell us if the process is in control or out of control, that is, whether a search for
special causes is warranted or the process should be left alone. More on the interpretation of
control charts will be presented later in this course.

What do they not tell us? Too often people want to judge the data plotted on a control chart by the
blueprint tolerance or specifications. Usually this is an attempt to verify that the process is
producing acceptable quality results. People who do this suffer from a serious misunderstanding or
misinterpretation of control charts.
_
A typical control chart, say an X & R chart, is based on subgroup averages. Blueprint specifications
are for individual measurements. Simply because a subgroup average falls within the specification
limits does not mean that the individual measurements are within those limits. Conversely, if an
individual measurement is outside the control limits, it is erroneous to conclude that the process is
out of control.

Remember, the primary function of a control chart is to indicate whether the process is in control or
out of control, not to judge the acceptability of an individual data point or part. It is possible for a
process to be in control yet produce out-of-spec material. A process also may be out of control and
produce entirely acceptable material.

Types of Control Charts

The two major divisions in control charts result from the fact that there are two types of data -
variable data and attribute data.

Whenever a record is made of an actual measured quality characteristic, such as a dimension

expressed in thousandths of an inch, the quality is said to be expressed by variables. Variables are
©2010 Davis M. Woodruff Page 4 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

measurable characteristics such as a dimension, weight, purity, temperature, yield, tensile strength,
flow rate, angularity, etc.

If a record shows only the number of articles conforming and failing to conform to specified
requirements, the quality is said to be a record by attributes. Attributes are countable characteris-
tics. For example, most visual examinations concern attributes. A shaft is either cracked or not; a
bearing surface has an acceptable finish or it doesn't; a pump seal fails or it does not; scratches are
present or not.

Types of Control Charts

Variables Charts

_
X-R Average and Range

~- R
X Median and Range

_
X-S Average and Standard Deviation

X - Rm Individual and Moving Range

Attributes Charts

p Fraction Nonconforming

np Number of Nonconforming Units

c Number of Nonconformities

u Number of Nonconformities per unit

Requirements for Using Control Charts

Before applying control chart techniques to any process or operation certain essential requirements
must be satisfied. First, the people in decision-making positions should understand and support
efforts to continually improve processes and quality through the use of statistical methods such as
control charts. When this has been achieved, the following criteria must be satisfied for a business
to really experience process improvement using control charts.

©2010 Davis M. Woodruff Page 5 of 36

www.PDHcenter.com PDH Course P209 www.PDHonline.org

1. Understand the process being studied. Understanding is necessary in order to make

decisions regarding chart selection, sample size, sampling frequency, characteristics
to be monitored and actions which can be taken to reduce variability.

2. Determine what characteristics are to be monitored. Usually these would be the ones
most critical for satisfactory functioning of the process or product or perhaps the ones
with the tightest specifications. They could be the characteristics most often found to
be nonconforming or the ones most expensive to repair. A technique such as Pareto
analysis is often used to select a single characteristic to be charted.

3. Define and understand the measurement system. The data on the chart is going to be
no better than the measuring process or device. The required accuracy must be
established so that proper measuring equipment can be selected. Measuring
equipment should be reviewed for accuracy and repeatability, and calibrated regu-
larly.

4. Select a chart type to best accomplish the purpose. Often this is dictated by the cost
of measurement and the loss incurred if important changes go undetected when and if
they should occur. Also, the type of information needed may dictate which chart to
use.

In general use an attribute chart if

a. Measurements are not possible, as in visual inspection.

b. Measurements are not practical due to cost of measuring equipment or ex-

cessive testing time.

c. The part has many characteristics to evaluate.

d. Plans are to use 100% inspection.

Use a variables chart if

a. A critical characteristic or a variable characteristic is involved.

b. More precise control is desired than is possible with attribute charts.

©2010 Davis M. Woodruff Page 6 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

5. Determine the sample size and frequency. For variables charts, the most common
sample size is five. The sample subgroup should be selected to allow minimum op-
portunity for variation within the group. However, the larger the sample size, the more
normal the distribution of subgroup averages even if the population is non-normal.
Economic factors also affect the selection of sample sizes. If subgroups cannot be
used, an individual and moving range chart may be used. In part 4 of this 5 part series
of courses, the impact of sample size will be discussed further.

Attribute charts usually have a much larger sample size - from 30 to several hundred
or more. Since attribute data may be available from preexisting quality reports or
studies, the time and expense have already been incurred.

It is difficult to offer any general rule for frequency of sampling. Each case is different
and must be decided on its own merits after considering the costs, risks and potential
benefits. Usually, when beginning a chart, frequent samples are taken to allow con-
clusions to be reached quickly. Later, sampling frequency possibly may be reduced
based on what is learned about the process.

Control Chart Selection

In manufacturing plants some type of process trouble is, unfortunately, a rather common state of
affairs. Control charts can be invaluable trouble-shooting aids for the engineer as well as for the
operators. Control charts provide information in at least four areas of primary concern:

1. Basic variability of the characteristic

2. Consistency of performance
3. Average level of the characteristic
4. Type of variability present

The following chart depicts a general flow chart selection guide for the different types of basic
control charts. Next is a table that presents a summary of the general control chart formulas for the
more commonly used Shewhart control charts. These are discussed in the sections on variables
and attribute charts.

©2010 Davis M. Woodruff Page 7 of 36

www.PDHcenter.com PDH Course P209 www.PDHonline.org

Control Chart Selection Guide

Data
Collection 

Attribute Data Variable Data

Sample Size (n) Chart Type Sample Size (n) Chart Type
>20, not constant p chart 1 Individual and Moving Range

>20, constant np chart <10 Average & Range

Median & Range

>1; <20, not constant u chart 10 to 25 Average and Std. Deviation

2
1 c chart >25 Average and Variance (σ )

p chart = fraction non conforming

np chart = number nonconforming

u chart = nonconforming/inspection unit

c chart = nonconforminginspection/unit

©2010 Davis M. Woodruff Page 8 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Control Chart Formulas

Tables of Constants for Control Charts

Sample Average & Range Median & Range Individual & Moving Average & Standard Deviation
Size All Charts Charts Range Charts Charts
(n) d2 A2 D3 D4 ~A2 D3 D 4 E2 D3 D4 A3 B3 B4 C4
2 1.128 1.880 0 3.267 –– 0 3.3 2.659 0 3.267 2.659 0 3.267 0.7979
3 1.693 1.023 0 2.574 1.187 0 2.6 1.772 0 2.574 1.954 0 2.568 0.8862
4 2.059 0.729 0 2.282 –– 0 2.3 1.457 0 2.282 1.628 0 2.266 0.9213
5 2.326 0.577 0 2.114 0.691 0 2.1 1.290 0 2.114 1.427 0 2.089 0.9400
6 2.534 0.483 0 2.004 –– 0 2 1.184 0 2.004 1.287 0.030 1.97 0.9515
7 2.704 0.419 0.076 1.924 0.509 0.076 1.9 1.109 0.076 1.924 1.182 0.118 1.882 0.9594
8 2.847 0.373 0.136 1.864 –– 0.136 1.9 1.054 0.136 1.864 1.099 0.185 1.815 0.9650
9 2.97 0.337 0.184 1.816 0.412 0.184 1.8 1.010 0.184 1.816 1.032 0.239 1.761 0.9693
10 3.078 0.308 0.223 1.777 –– 0.223 1.8 0.975 0.223 1.777 0.975 0.284 1.716 0.9727

Summary

©2010 Davis M. Woodruff Page 9 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

• Process control charts are tools which help us analyze data. Specifically, they distinguish
between variability as a result of common causes or special causes.

• Processes are in control if only common causes are present.

• Processes are out of control if special causes are present.

• The two general categories of control charts are Variables Control Charts and Attribute
Control Charts.

• Tables of constants have been developed for use with control charts to utilize the relationship
between the samples and the population based on Shewhart’s research. These are the only
accepted methods of determining control limits.

• When the population is used to determine the control limits for an average and range chart,
false indications of the state of control are evident because the control limits are to be based
on the sampling distribution of averages and not the population distribution of individual
values. The same holds true for the other variables charts.

©2010 Davis M. Woodruff Page 10 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Variables Control Charts

Introduction

By using variables control charts, the process performance can be recorded and monitored with
respect to a selected variable. Variables are those quality characteristics which can be measured.

In this section, the types of variables charts to be discussed are as follows:

_
• Average and Range charts or "X and R" charts.

• ~ and R" charts.

Median and Range charts or "X

_
• Average and Standard Deviation charts or "X and S" charts.

• Chart of Individuals or "X and Rm" charts.

The steps involved in developing all of these variables charts are basically the same. The
differences are associated with the particular parameter to be plotted. These basic steps are
presented below.

Steps for Developing and Plotting

A Variables Control Chart

1. Enter the heading information on the chart. This may include such items
as the name of the part, the part number, department, machine number,
plant identification, operation being performed, etc.

2. Record the information about the sample. These are usually the date
the sample was taken, the time, the initials of the checker, and shift on
which the samples were taken.

3. Entered the measured data. In the appropriate places, enter the

measurements that have been taken.

©2010 Davis M. Woodruff Page 11 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

4. Calculate the statistic being monitored. This will be the average, the
median, the range, the standard deviation, or other statistics.

5. Establish the scales for the graphs. Scales should be used which make
it easy to do the plotting. As a general rule, the range of values for the
central tendency measure should be either the product specification or
twice the difference in the highest and lowest subgroup statistics. The
chart for monitoring variability should have scales beginning at zero and
extending to a value of about 1.5 to 2 times the maximum calculated
measure of variability.

6. Plot the data.

7. Calculate and plot the central lines. There will be a central line for both
portions of the chart - central tendency and variability. These should be
drawn in as solid lines and labeled.

8. Calculate and plot the control limits. These are calculated using the
specific formulas for the statistics being monitored. The formulas are
shown on a previous page in this course. The necessary constants are
shown as well. The control limits should be drawn in as dashed lines on
the chart and labeled.

9. Interpret the chart. This will be discussed in detail later in this course.

Control Limit Calculations

As you remember from the section "Introduction to Control Charts", the upper and lower control
limits are, in reality, simply the ±3σ limits for the sample statistic. Let's look at the formulas which
_
have been developed for X and R charts.

©2010 Davis M. Woodruff Page 12 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

The upper control limit (UCL) and lower control limit (LCL) are calculated as follows:

 
_ _
These equations have the basic form of X ± 3σx-. Where A2R is a means of determining 3 standard
_ _
deviations of the of the sample statistic, in this case, X. By calculating R and using the constant A2
we can determine the 99.73% limits for the statistic we are monitoring. Although A2 is a constant, it
does depend on the sample size. For the Range chart, the distribution is not normal as it is bounded
at Zero and Infinity, thus the calculations take that into consideration. For subgroup sizes ≤6 the
lower control limit for the Range will be zero (see the table of constants).

A table of the constants used in variable control charting is shown in the table below.
Formulas have been derived for specific charts and constants established. But in every case, the
control limit calculations allow you to calculate the ±3 standard deviation limits.

Sample Average & Range Median & Range Individual & Moving Average & Standard Deviation
Size All Charts Charts Range Charts Charts
(n) d2 A2 D3 D4 ~A2 D3 D 4 E2 D3 D4 A3 B3 B4 C4
2 1.128 1.880 0 3.267 –– 0 3.3 2.659 0 3.267 2.659 0 3.267 0.7979
3 1.693 1.023 0 2.574 1.187 0 2.6 1.772 0 2.574 1.954 0 2.568 0.8862
4 2.059 0.729 0 2.282 –– 0 2.3 1.457 0 2.282 1.628 0 2.266 0.9213
5 2.326 0.577 0 2.114 0.691 0 2.1 1.290 0 2.114 1.427 0 2.089 0.9400
6 2.534 0.483 0 2.004 –– 0 2 1.184 0 2.004 1.287 0.030 1.97 0.9515
7 2.704 0.419 0.076 1.924 0.509 0.076 1.9 1.109 0.076 1.924 1.182 0.118 1.882 0.9594
8 2.847 0.373 0.136 1.864 –– 0.136 1.9 1.054 0.136 1.864 1.099 0.185 1.815 0.9650
9 2.97 0.337 0.184 1.816 0.412 0.184 1.8 1.010 0.184 1.816 1.032 0.239 1.761 0.9693
10 3.078 0.308 0.223 1.777 –– 0.223 1.8 0.975 0.223 1.777 0.975 0.284 1.716 0.9727

Constants for Variables Control Charts

©2010 Davis M. Woodruff Page 13 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

_
Average and Range (X - R) Charts

The keys to plotting an Average and Range chart are the central line calculations and the formulas
for the control limits. These are shown below. Constants are from the table above.
Central Lines

=( 1+ 2+ 3 +……. n) n

Or stated in words, it is the sum of the subgroup (sample) averages

divided by the number of subgroups or samples

= ,

Or stated in words it is the sum of the Ranges

divided by the number of subgroups or samples 

Control Limits

38
UCL, 37.355
37
Average Thickness (mm), X-bar

36

35 CL, 34.840
34

33
LCL, 32.325
32

31

30

29
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Sam ple #

10
UCL, 9.219
8
6
Range

4 CL, 4.360

2
0 LCL, 0.000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Sample #

Average and Range Control Chart

©2010 Davis M. Woodruff Page 14 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Note: By doing an internet search, one can find a variety of templates that allow you to simply enter
the data and let excel® or another program do all the work for you. However, it is important to
understand the statistics involved to enable proper interpretation of the results.

Just to be sure the concepts are clear, another example that shows the data by subgroups,
calculations and control chart is shown below:

Data Collection for Average and Range Chart/Process Capability

Subgroup X1 X2 X3 X4 X5 Ave Range

1 41 41 46 44 43 43.0 5
2 42 41 42 41 45 42.2 4
3 41 40 44 43 48 43.2 8
4 43 46 47 49 42 45.4 6
5 47 43 49 44 44 45.4 6
6 44 49 45 46 43 45.4 6
7 46 44 42 40 48 44.0 6
8 45 41 44 48 46 44.8 7
9 49 49 47 45 46 47.2 4
10 47 41 43 42 45 43.6 6
11 46 45 42 48 45 45.2 6
12 41 47 42 46 44 44.0 6
13 49 45 46 43 49 46.4 6
14 46 41 45 45 45 44.4 5
15 47 48 41 43 47 45.2 7
16 46 42 44 49 50 46.2 8
17 42 44 42 41 46 43.0 5
18 41 46 49 47 45 45.6 8
19 45 49 41 43 48 45.2 8
20 42 49 44 47 48 46.0 5
21 49 46 41 45 45 45.2 8
22 42 49 46 42 41 44.0 8
23 41 42 46 43 49 44.2 8
24 41 44 45 44 44 43.6 4
25 48 43 44 50 43 45.6 7
Average 44.7
Ave. Range 6.3

Note: Range = hi-lo in each subgroup

Central Line = Average of Subgroup Averages = 44.7 and the Average Range = 6.3

UCL = 44.7 + 3(0.577 x 6.3) = 44.7 + 10.9 = 55.6

LCL = 44.7 - 3(0.577 x 6.3) = 44.7 - 10.9 = 33.8

UCLR = Ave R x D4 = 6.3 x 2.114 = 13.3

LCLR = Ave R x D3 = 6.3 x 0 = 0

The Average Chart:

©2010 Davis M. Woodruff Page 15 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

60.0

55.0

50.0

45.0
Series2

40.0

35.0

30.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

The Range Chart:

14

12

10

8
Series1
6

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

~ and R) charts
Median and Range (X

The calculations for establishing a median and range chart are like average and range charts,
except the median is used for each subgroup and the overall median is used as the central line. The
©2010 Davis M. Woodruff Page 16 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

R chart is constructed exactly like the R chart for an average and range chart. Median charts are
seldom used, but are available if needed.

_
Average and Standard Deviation (X and S) Charts

The average and standard deviation chart has not been used much in industry because of the
required calculation of the standard deviation. However, with excel® or other statistical packages,
this chart could easily become a manufacturing fixture. The formulas for developing such a chart
are: Again, it is mentioned herein for completeness and to expose the reader to all of the tools
available. To see the formulas used, go to the table of control chart formulas.

Charts of Individuals (X and Rm)

With some processes, it may not be practical to utilize subgroups. Instead, individual measure-
ments or readings are collected. Since it is not possible to determine a range with only one
observation, a moving range is calculated instead. The variables plotted on the control chart are
then the individual readings and the moving range.

The steps in developing this chart are outlined below:

1. Record the data.

2. Calculate the moving range, Rm. By taking the difference between the
first and second observations, record it; then, the difference in the
second and third, third and fourth, etc. Just as the range is the difference
between largest and smallest values within a subgroup, the moving
range is the difference between each subsequent individual value.

3. Plot the individual values, X, and the moving range on their respective
charts.

4. Calculations of chart characteristics:

Central Lines

©2010 Davis M. Woodruff Page 17 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

The central line = ⁄n

Or, in words simply add the individual values and divide by how many values you have.

The central line for ⁄n-1

Or, in words simply add the moving ranges and divide by the number of
moving ranges you have minus 1.

Control Limits

Rm

UCLRm =

LCLRm =

5. Interpretation of the chart is similar to the other variables charts except

the tests for non-randomness are less significant on the moving range.
An example of a moving range chart is shown below.
Chart of Individuals and Moving Ranges for Machine Uptime
130
UCL

120

110

AVE
100

90

80

LCL
70
MARCH APRIL MAY JUNE JULY AUG SEPT OCT NOV DEC JAN FEB MARCH

Individual Values, X

35.0
UCL
30.0

25.0

20.0

15.0 Series1
Ave
10.0

5.0

0.0
1 2 3 4 5 6 7 8 9 10 11 12

Moving Ranges, n=2 for ranges

Individual and Moving Range Chart

©2010 Davis M. Woodruff Page 18 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Control Chart Interpretation

Once data have been recorded on a process control chart, the task of interpreting the chart begins.
The control chart is used to indicate when special cause variation is occurring. The range of
common cause variation or random, inherent dispersion is represented by the upper and lower
control limits on the chart. Whenever nonrandom or special cause variation occurs it can manifest
itself in several ways.

The first rule of chart interpretation is to look for points that fall outside the control limits.
Remember, evaluate the RANGE chart first on a variables chart. If the range chart is out of
control, it is essentially useless to attempt evaluation of the average (individual, median, etc.) chart.
Whether the results are outside the upper or lower limits, the indication is that some unusual
circumstances have developed which have shifted the distribution of the variable. Points falling
outside the control limits are very strong indicators and the likelihood of a "false" alarm is extremely
remote.

Next, look for trends in the data. Trends are depicted by seven points in a row in an increasing or
decreasing pattern. Trends are examples of nonrandom patterns in the data. Another way the
nonrandom behavior of the data exhibits itself is by a non-normal distribution of the control chart
variable. The Central Limit Theorem states that the distribution of the sample means will form a
normal distribution. If we find that this has not occurred, then something has influenced the data. A
quick check for a normal distribution is to count the number of points in the middle third of the chart.
Since the middle third represents the ±1σ limits, approximately 2/3 of the points should be within this
portion of the chart. (If you recall, from Part 2 of Statistical Methods for Process Improvement, Using
the Normal Distribution, ±1σ encompasses 68.26% of the data of a normal distribution.) With most
of the data falling near the central line of the chart, we would expect to find only a few points near the
control limits. Several simple rules for checking for non-randomness are shown below. These rules
are based on probabilities that are essentially equivalent to finding one point outside the control
limits.

©2010 Davis M. Woodruff Page 19 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Simple Tests for Non-randomness

2 of 3 consecutive points outside 2σ limits

4 of 5 consecutive points outside 1σ limits

8 consecutive points on one side of the mean

15 consecutive points within the ±1σ limits

(Check the chart above and below the

central line independently)

This last test is to determine if stratification is occurring. Stratification is associated with a stable
mixture which exhibits an artificial or false stability. Instead of varying randomly within the control
limits, stratified data will tend to hug the central line. Sometimes the distribution is referred to as
"unnaturally quiet".

Do not make the mistake of concluding that a stratified pattern is showing the process to be in good
statistical control. Just the opposite is true. Stratification results whenever samples are taken in a
consistent manner from very different distributions. Sometimes this occurs by taking one part from
each operator in a group, or one part from each machine, or one part from each cavity in a
multi-cavity molding machine, etc. Usually people sample in this manner in an effort to make the
samples "representative". Instead, the deliberate pattern leads to a non-randomness and
subsequently to a lack of natural fluctuation.

Control Chart Typical Random Pattern

There are several other simple tests, similar to these. However, these few tests will allow adequate
interpretation of the chart. These tests are all based upon the probability of occurrences, and are
indications of special causes of variability. One must remember that chart interpretation rules
©2010 Davis M. Woodruff Page 20 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

should be applied along with a good deal of common sense and judgment, based on your
knowledge of the process and measurement system.

Other Tests

While there are “tests” for the number of runs above and below the central line or the length of runs,
the simple tests listed above will provide a solid basis for chart interpretation. If you desire more
information on control chart interpretation, see Statistical Quality Control Handbook, Western
Electric.

Summary

1. Variables control charts are used for measurable characteristics.

2. There are four general types of variables charts that may be used to monitor a
process:

• Average and Range Charts

• Median and Range Charts
• Average and Standard Deviation Charts
• Individuals and Moving Range Charts
3. Control charts must be interpreted for nonrandom variation which is symptomatic of
special causes.
4. Interpretation of control charts will indicate whether a process is in control, out of
control, or developing a trend.
5. The real work begins when the chart is interpreted and corrective or preventive actions
are indicated.
6. Always look for the “test of reasonableness.” Ask yourself, “Does this make sense
given what I know about this process?”

©2010 Davis M. Woodruff Page 21 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Attribute Control Charts

Introduction

An important distinction in process control is that between variables and attributes. When data is
gathered on an actual measured quality characteristic, such as a dimension expressed in
thousandths of an inch, the quality is said to be expressed in variables. When we look at the number
of articles failing to conform to any specified requirement, it is expressed as an attribute. The
section titled "Introduction to Control Charts" discusses attributes and variables in more detail.

Control Charts for Attributes

There are several different types of control charts for attributes. These are summarized below.

1. The p chart monitors the proportion of items nonconforming to specifications.

2. The np chart monitors the actual number of nonconforming items.

3. The c chart monitors the actual number of nonconformities.

4. The u chart monitors the number of nonconformities per unit.

At this point, it may be good to review the difference between nonconformities and nonconforming
units. An article which does not conform in some way to a given specification can be described as
nonconforming. Every nonconforming article contains one or more nonconformities.

Control Chart for Nonconforming Units

The most versatile and widely used attributes control chart is the p chart. This is the chart for the
fraction rejected as nonconforming to specifications. It is applied to quality characteristics which can
be observed as attributes only or to those that are considered as attributes. As long as the result of
the inspection is the classification of an individual article as accepted or rejected, a single chart may
be applied to one or any number of quality characteristics contained in the article.

To become better acquainted with the use of p charts, consider an experiment that fairly well
describes the use of p charts in an industrial environment. Suppose 10,000 beads of the same size
and density are placed in a container. Of these beads, 9,500 are white and 500 are red. The white
©2010 Davis M. Woodruff Page 22 of 36
 www.PDHcenter.com PDH Course P209 www.PDHonline.org

beads represent articles which conform to specifications, and the red beads represent noncon-
forming articles. Samples of 50 beads are drawn from the 10,000 beads. If the beads are replaced
after each sample has been drawn and all the beads are thoroughly mixed before the next drawing,
the theory of probability enables us to calculate the relative frequency, in the long run, of getting
exactly 0, 1, 2, 3, 4, 5, etc. red beads.

Any one sample of beads drawn from the container is a sample from a very large quantity of beads
with 5% red. As a matter of chance, variations from sample to sample in the number of red beads
are inevitable. In a similar way, we may think of a day's production (or other lot) of any manufac-
tured article or part as a sample from a larger quantity with some unknown fraction which do not
conform to specifications.

p Charts

Shown below are the basic calculations necessary to develop a p chart. Just as with variables
charts, the calculations are for the central line and the control limits. On the following page, an
example of a p chart is shown.

Central Lines

The central line is simply the average proportion nonconforming, or as a formula:

= Total Number Nonconforming/Total Number Inspected,

expressed as a decimal

Control Limits

UCLp =

LCLp =

©2010 Davis M. Woodruff Page 23 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Attribute Control Chart: Proportion Non-Conforming (p chart)

Nonconformities in Purchase Orders

N=150/month examined
Type of n/c Jan Feb March Apr May June July Aug Sept Oct Nov Dec Total
Missing signature 1 2 1 7 4 6 5 9 7 6 1 9 58
Incorrect dates 3 1 0 3 9 4 1 9 1 5 7 5 48
Inadequate info 1 2 4 9 5 2 4 7 3 2 2 6 47
Supplier not approved 1 1 3 4 6 0 8 5 2 8 6 3 47
Exceeded limits 6 5 8 4 3 8 6 6 5 5 4 9 69
Other 4 9 2 0 1 4 7 2 5 9 5 2 50
Total non-conforming 16 20 18 27 28 24 31 38 23 35 25 34 319
Proportion non-conforming (p) 0.11 0.13 0.12 0.18 0.19 0.16 0.21 0.25 0.15 0.23 0.17 0.23 0.18
(319/1800)

Ave. Proportion N/C = 0.18 0.30

(Central Line)

UCL = 0.27
0.25
LCL = 0.09

0.20

0.15
Series1

0.10

0.05

0.00

Example of p Chart for Purchase Order Non-conformities

np Charts

©2010 Davis M. Woodruff Page 24 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

The choice among the control chart techniques for attribute data is based primarily upon conven-
ience in the interpretation of the chart. p charts are used when the number of items inspected are
constant or varying, and the statistic of interest is the fraction rejected.

Whenever the subgroup size is variable, the chart must show the fraction rejected rather than the
actual number of rejects. If the subgroup size is constant, the chart for actual number of rejects
(non-conformances) may be used. Such a chart is called an np chart.

The formula for np charts central line and control limits are shown below:

Central Lines

n = Number Nonconforming

Control Limits

UCLnp = n

LCLnp =

c Charts

We have already discussed that a nonconforming article is an article that in some way fails to
conform to one or more given specifications. Each instance of the article's lack of conformance is a
nonconformity. Where it is appropriate to make a total count of the number of nonconformities in
each article, or group of articles, either a c chart or a u chart is used. Examples of c chart
applications could be:

1. The number of n/c rivets in an aircraft wing.

2. The number of imperfections observed in a galvanized sheet.

3. The number of imperfections found in a bolt of cloth.

The basic formulas for charts are:

Central Lines

©2010 Davis M. Woodruff Page 25 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Simply the average number of nonconformities per unit and the sample size is one. ( )

Control Limits

The control limits are simply the 3 standard deviation limits of the distribution, which in this

distribution is actually the square root of the average, or , thus the control limits are:

UCLC =

LCLC =

u Charts

When working with c charts, the quantity c was the number of nonconformities observed in some
specified inspection of a constant size, thus the distribution is bounded at 0 and 1. This inspection
unit could be a single unit of product, 10 products or any number of products. When plotting on
control charts, the total number of nonconformities per inspection unit were plotted just as if the
subgroup were a unit of product.

Whenever there is an evident change in the area of opportunity for occurrence of a nonconformity
from subgroup to subgroup, a chart showing only the total number of nonconformities is not
applicable. For example, if a number of units constitute a subgroup of size n, where n varies from
subgroup to subgroup, defects per unit (c/n) may be an appropriate control statistic. In such cases,
if total defects observed in each subgroup were plotted, the central line on the chart as well as the
control limits would change from one subgroup to another.

The symbol 'u' is used to represent nonconformities per unit or c/n where c is the nonconformities
found and n is the number of items. The central line on the chart would then be u.

Formulas for finding the central line and the control limits for u charts are shown below:

©2010 Davis M. Woodruff Page 26 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Central Lines

Control Limits

UCLU =

LCLU =

Chart Interpretation

For attribute charts, remember that if data points show up on a chart, something is not right with our
process, nonconformities have occurred. So, there is no range chart as in variables charts. Also,
these are not based on a normal distribution, so some of the typical rules do not apply. Instead, look
for:
• A point outside the control limits
• Trends in the data
• Significant runs (8 or more points above or below the mean)

With attribute charts, it is as important to investigate a process performing “better than expected” as
it is a process that is trending towards an upper control limit. Sometimes, with attribute charts we
discover a process is actually improving and when that is the case we need to apply a problem
solving approach to determine why it is better so we can work to sustain the improvements and/or
transfer the knowledge to other processes. For example, let’s use the p chart for purchase orders
we looked at earlier. When the data are summarized for a 12 month period, a Pareto Chart of the
non conformities quickly shows us the top issue, is purchase orders that exceeded limits. Now, we
can begin taking corrective action based on the process performance and not on an isolated
incident.

©2010 Davis M. Woodruff Page 27 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

80
70
60
50
40
30
20
10
Series1
0

Summary Control Charts

Why Use Summary Charts?

It is easy to become inundated with control charts that have been completed by operators,
electronically or other personnel.

For control charts to be utilized as a management tool, it is important to establish an easy to use
method of summarizing control chart information. Summary Control Charts are an easy way for one
to determine the overall impact of SPC in an operation. Both Attribute Charts and Variables Charts
can become summary charts.

p Charts as Summary Charts

The following procedure outlines the methods used:

1. Tabulate the total number of data points plotted on all control charts.
This may be done daily, weekly or monthly. The control charts may be
from one department, one process, one product or the entire plant.
Record the total number of points on the p Chart. This will become the
summary chart.

2. Tabulate the number of out of control points for the same data used in
#1, above.

©2010 Davis M. Woodruff Page 28 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

3. Calculate p for each set of data. This is the fraction non-conforming to

control limits. Plot on the summary chart.

4. When sufficient data are available, calculate the control limits for the
summary chart. You may use average sample size or the formula for
moving control limits, depending upon the actual data.

5. Evaluate the chart as any other attribute chart.

The p Chart as a summary control chart allows you to evaluate the overall performance of SPC
efforts in a given area by:

1. Observing the total number of data points plotted on charts.

2. Comparing how processes are performing in relation to control limits.

3. Improvements in performance can be observed.

4. More data are available for decision making.

Average and Standard Deviation Charts as Summary Charts

The following procedure outlines the methods used:

1. When a particular variables control chart is filled with data, it is time to

summarize the data for future reference and use.

_ _
2. Calculate X (or X) for the data on the chart.

_
3. Calculate S or σ using R/d2 for the chart. You may calculate S directly
from all the data points if you prefer.

4. Transfer the values from #2 and #3 to the Summary Chart as data for
one sample for this particular variable.

©2010 Davis M. Woodruff Page 29 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

5. After completing at least ten charts for a variable, and ten data points for
the summary chart, calculate control limits for the summary chart.

6. Evaluate the summary chart as any other variables chart.

Summary variables charts enable one to view the data for a particular variable over a longer period
of time, thus allowing evaluation by:

1. Results being monitored over time.

2. Long term trends or shifts are detected in a process or product.

3. Data are available for decision making concerning process changes,

capital expenditures, etc.

4. Process data are made more manageable and useful.

Getting the Most Benefits From Control Charts in Your Business

To gain the most benefit it is important to define and implement a uniform method of reviewing
control charts for continual improvement.

1. Define a routine review plan for control charts.

2. Summarize out of control conditions and special causes on an established frequency.
3. Maintain the summary control charts.
4. Review major special causes and do Pareto and/or other studies as required.
5. Re-calculate control limits as needed.
6. Assign specific responsibilities to specific personnel.
7. Take appropriate corrective actions using your corrective action process

These three general portions of a process control/corrective action plan can be expanded and
tailored to any of your facilities to ensure consistent response to what our processes are "telling us".

©2010 Davis M. Woodruff Page 30 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Glossary of Statistical Terms

Accuracy - The closeness of agreement between and observed value and an accepted reference value.

Alternative Hypothesis - The hypothesis that is accepted if the null hypothesis is disproved.

Attribute Data - Qualitative data that typically shows only the number of articles conforming and the number of articles

failing to conform to a specified criteria. Sometimes referred to as Countable Data.

Average - The sum of the numerical values in a sample divided by the number of observations.

Bar Chart - A chart that uses bars to represent data. This type of chart is usually used to show comparisons of data

from different sources.

Batch - A definite quantity of some product or material produced under conditions that are considered uniform.

Bias - A systematic error which contributes to the difference between a population mean of measurements or test

results and an accepted reference value.

Bimodal Distribution - A distribution with two modes that may indicate mixed data.

Binomial Distribution - A distribution resulting from measured data from independent evaluation, where each

measurement results in either success or failure and where the true probability of success remains constant from

sample to sample.

Cells - The bars on a histogram each representing a subgroup of data.

Common Cause - A factor or event that produces normal variation that is expected in a given process.

Confidence - The probability that an interval about a sample statistic actually includes the population parameter.

Control Chart - A chart that shows plotted values, a central line and one or two control limits and is used to monitor a

process over time.

Control Limits - A line or lines on a control chart used as a basis for judging the significance of variation from subgroup

to subgroup. Variation beyond a control limit shows that special causes may be affecting the process. Control limits are

usually based on the 3 standard deviation limits around an average or central line.

Countable Data - The type of data obtained by counting -attribute data.

Data - Facts, usually expressed in numbers, used in making decisions.

Data Collection - The process of gathering information upon which decisions to improve the process can be based.

Detection - A form of product control, not process control, that is based on inspection that attempts to sort good and bad

output. This is an ineffective and costly method.

©2010 Davis M. Woodruff Page 31 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Distribution - A group of data that is describable by a certain mathematical formula.

Frequency Distribution - A visual means of showing the variation that occurs in a given group of data. When enough

data have been collected, a pattern can usually be observed.

Histogram - A bar chart that represents data in cells of equal width. The height of each cell is determined by the

number of observations that occur in each cell.

k - The symbol that represents the number of subgroups of data. For example, the number of cells in a given histogram.

Lower Control Limit (LCL) - The line below the central line on a control chart.

Mean - The average value of a set of measurements, see Average.

Measurable Data - The type of data obtained by measurement. This is also referred to as variables data. An example

would be diameter measured in millimeters.

Median - The middle value ( or average of the two middle values) of a set of observations when the values have been

ranked according to size.

Mode - The most frequent value in a distribution. The mode is the peak of a distribution.

n - The symbol that represents the number of items in a group or sample.

np - The symbol that represents the central line on an np chart.

Nonconformities - Something that doesn't conform to a drawing or specification; an error or reason for rejection.

Normal Distribution - A symmetrical, bell-shaped frequency distribution for data. This is a distribution that is often

seen in industry.

Null Hypothesis - The hypothesis tested in test of significance that there is no difference (null) between the population

of the sample and the specified population (or between the populations associated with each sample).

Out of Control - The condition describing a process from which all special causes of variation have not been eliminated.

This condition is evident on a control chart by the presence of points outside the control limits or by nonrandom patterns

within the control limits.

p - The symbol on a p chart that represents the proportion of nonconforming units in a sample.

Parameter - A constant or coefficient that describes some characteristics of a population (e.g. standard deviation,

average, regression coefficient).

Pareto Charts - A bar chart that arranges data in order of importance. The bar representing the item that occurs or

costs the most is placed on the left-hand side of the horizontal axis. The remaining items are placed on the axis in

descending (most to least) order.

©2010 Davis M. Woodruff Page 32 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Population - All members, or elements, of a group of items. For example, the population of parts produced by a

machine includes all of the parts the machine has made. Typically in SPC we use samples that are representative of the

population.

Prevention - A process control strategy that improves quality by directing analysis and action towards process

management, consistent with the philosophy of continuous quality improvement.

Process - Any set of conditions or set of causes working together to produce an outcome. For example, how a product

is made.

Product - What is produced; the outcome of a process.

Quality - Conformance to requirements.

Random Sampling - A data collection method used to ensure that each member of a population has an equal chance of

being part of the sample. This method leads to a sample that is representative of the entire population.

Range - The difference between the highest and lowest values in a subgroup.

Repeatability - The variation in measurements obtained when one operator uses the same test for measuring the

identical characteristics of the same samples.

Reproducibility - The variation in the average of measurements made by different operators using the same test when

measuring identical characteristics of the same samples. (In some situations it is the combination of operators,

instruments and locations.)

Run Chart - A line chart that plots data from a process to indicate how it is operating.

Sample - A small portion of a population.

Sampling - A data collection method in which only a portion of everything made is checked on the basis of the sample

being representative of the entire population.

Significance Level () - The risk we are willing to take of rejecting a null hypothesis that is actually true.

Skewed Distribution - A distribution that tapers off in one direction. It indicates that something other than normal,

random factors are effecting the process. For example, TIR is usually a skewed distribution.

Special Cause - Intermittent source of variation that is unpredictable, or unstable; sometimes called an assignable

cause. It is signaled by a point beyond the control limits or a run or other nonrandom pattern of points within the control

limits. The goal of SPC is to control the special cause variation in a process.

Specification - The extent by which values in a distribution differ from one another; the amount of variation in the data.

Standard Deviation (σ) - The measure of dispersion that indicates how data spreads out from the mean. It gives

information about the variation in a process.

©2010 Davis M. Woodruff Page 33 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Statistic - A quantity calculated from a sample of observations, most often to form an estimate of some population

parameter.

Statistical Control - The condition describing a process from which all special causes of variation have been eliminated

and only common causes remain, evidenced by the absence of points beyond the control limits and by the absence of

nonrandom patterns or trends within the control limits.

Statistical Methods - The means of collecting, analyzing, interpreting and presenting data to improve the work process.

Statistical Process Control (SPC) - The use of statistical techniques to analyze data, to determine information, and to

achieve predictability from a process.

Statistics - A branch of mathematics that involves collecting, analyzing, interpreting and presenting masses of

numerical data.

Subgroup - A group of consecutively produced units or parts from a given process.

Tally or Frequency Tally - A display of the number of items of a certain measured value. A frequency tally is the

beginning of data display and is similar to a histogram.

Tolerance - The allowable deviation from standard, i.e., the permitted range of variation about a nominal value.

Tolerance is derived from the specification and is NOT to be confused with a control limit.

Trend - A pattern that changes consistently over time.

Type I Error (α) - The incorrect decision that a process is unacceptable when, in fact, perfect information would reveal

that it is located within the "zone of acceptable processes".

Type II Error(β) - The incorrect decision that a process is acceptable when, in fact, perfect information would reveal that

it is located within the "zone of rejectable processes".

u - The symbol used to represent the number of nonconformities per unit in a sample which may contain more than one

unit.

Upper Control Limit (UCL) - The line above the central line on a control chart.

Variables - A part of a process that can be counted or measured, for example, speed, length, diameter, time,

temperature and pressure.

Variable Data - Data that can be obtained by measuring. See Measurable Data.

Variation - The difference in product or process measurements. A change in the value of a measured characteristic.

The two types of variation are within subgroup and between subgroup. The sources of variation can be grouped into two

major classes: Common causes and Special Causes.

X - The symbol that represents an individual value upon which other subgroup statistics are based.

©2010 Davis M. Woodruff Page 34 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

_
X (x bar) - The average of the values in a subgroup.

__
X(x double bar) - The average of the averages of subgroups.

©2010 Davis M. Woodruff Page 35 of 36

 www.PDHcenter.com PDH Course P209 www.PDHonline.org

Suggested Readings and/or References:

Duncan, A.J., Quality Control and Industrial Statistics, 5th ed. Irwin.

Grant, E.L. and Leavenworth, R.S., Statistical Quality Control, 5th ed., McGraw Hill.

Blank, L.T., Statistical Procedures for Engineering, Management and Science, 1st ed, McGraw Hill.

Wadsworth, H.M, Stephens, K.S. and Godfrey, A.B., Modern Methods for Quality Control and Improvement, 1st ed.,

Wiley.

Juran, J.M., et al, Quality Control Handbook, McGraw Hill

Ott, E.R., Process Quality Control: Troubleshooting and Interpretation of Data, McGraw hill

Shewhart, W.A., Economic Control of Quality of Manufactured Products, Van Nostrand.

Levin, R.I., Statistics for Management, 3rd ed., McGraw Hill.

Feigenbaum, A.V., Total Quality Control, 3rd ed., McGraw Hill

Statistical Quality Control Handbook, Western Electric.

Charbonneau, Harvey C. and Webster, Gordon L., Industrial Quality Control.

©2010 Davis M. Woodruff Page 36 of 36