Control Charts

Control Chart
NORMAL BEHAVIOR

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 Control Chart
UNHEALTHY TREND NEEDING INVESTIGATION

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 Control Chart
UNHEALTHY TREND NEEDING INVESTIGATION

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 TOO MANY PLOTS ABOVE THE CENTER LINE
( BEARS INVESTIGATION )

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 ERRATIC BEHAVIOR BEARS INVESTIGATION

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 OUT-OF-CONTROL PROCESS. STOP, INVESTIGATE, CORRECT!

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 TWO PLOTS NEAR “UCL” - INVESTIGATE

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

 Control Chart
TWO PLOTS NEAR “LCL” - INVESTIGATE

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE AVERAGES PLOTTED OVER TIME

Process Variability – Two Types
Natural Process Variability

An example are the miniscule

changes in product output
caused by alternating electric
current flow to manufacturing
equipment
 Natural Process Variability

UPPER CONTROL LIMIT

PROCESS
AVERAGE

LOWER CONTROL LIMIT

SAMPLE MEANS PLOTTED OVER TIME

Assignable Cause Variability

Not random,
but controllable,
that is, the
cause or causes
can be identified
and eliminated
Classification of Control Charts
 Control Charts
Four Versions

I. X – bar chart

II. R – chart

III. p – chart

IV. c - chart
X – Bar Chart

 Measures the central tendency of a process,

that is, the variability or dispersion around
the normal process average.

 The process average itself can be the desired,

historical, or the originally-designed average
of the process.
If a sample of four items had weights of:
2 ounces
10 ounces
4 ounces
7 ounces
R-Chart
the range would be 8 ounces

 Measures the range ( R ) between the largest and

smallest or the heaviest and lightest items within
each randomly selected sample.

 In other words, it measures the gain or loss in

uniformity within the process.

UNIFORMITY IS AFFECTED BY LOOSE OR WORN PARTS,

OPERATOR SLOPPINESS, ERRATIC FLOW OF LUBRICANTS
TO A MACHINE, AND SO ON

If four samples had ranges of:

6 ounces
8 ounces
9 ounces
2 ounces
The average range would be 6.25 ounces
The Central Limit Theorem

_ = is the mean of
X X the sample means

The distribution of =
sample means x
is called the
sampling Regardless of the particular probability distribution
distribution of a product or service characteristic, the distribution
of its sample averages will follow a normal
distribution even if the number of observations
in each sample is as small as four !
The Process or Target Mean
A MAJOR ELEMENT OF THE CONTROL CHART

The normal distribution

of the sample averages also the mean of all
( sample means ) possible sampling
is called the distributions for a
sampling distribution
= given sample size ( n )

the mean of the

X
Both are often estimated
sampling distribution by averaging sample
is means drawn from
the mean of the means a normal process

=
x is the process mean or average for a “healthy” process !
 The Standard Deviation
A MEASURE OF DISPERSION AROUND THE MEAN

USED TO ESTABLISH THE UPPER AND LOWER CONTROL LIMITS

THE STANDARD DEVIATION POPULATION

OF THE STANDARD
SAMPLING DISTRIBUTION DEVIATION
( THE STANDARD ERROR ) ( KNOWN )

σ X
σ_ =
X
The standard error
is the measure of dispersion
√n SAMPLE
around the mean of the
SIZE
sampling distribution
 X – Bar Chart

THE
NUMBER OF
UPPER CONTROL LIMIT ( UCL ) NORMAL
STANDARD
= DEVIATIONS

The Template X + zσ - THE

x STANDARD
ERROR
LOWER CONTROL LIMIT ( LCL )

THE MEAN
=
X - zσ x-
+/- 1 z
OF THE
SAMPLE +/-2z
MEANS
(process average) +/-3z

LCL = UCL
X
 THE NUMBER OF NORMAL STANDARD DEVIATIONS (Z)

= _
X +/- 3σx
Widest Control Limits !

99.7% of the time, the sample means will fall within the above “z”
values
control limits if the process has only random variations* are used

= _
when the
process
+/- 2σx standard

X deviation
(σ)
is known
95.5% of the time, the sample means will fall within the above
control limits if the process has only random variations*

= _
+/- 1σx Narrowest Control Limits !
X
68% of the time, the sample means will fall within the above
control limits if the process has only random variations*
* NATURAL VARIABILITY
Setting Control Chart Limits
depends Narrow
control
on the limits
nature could
produce
of the multiple
narrowest control limits process “false alarms”
or whereas
wide
industry control
middle control limits norms limits
could
allow
widest control limits serious
process
changes
to go
= undetected
-3 z -2 z -1 z x +1 z +2 z +3 z

THE NUMBER OF NORMAL STANDARD DEVIATIONS ( Z )

The Significance of “Z”

_
X

THE IDEA BEHIND CONTROL CHARTS !

If the mean of a randomly

selected sample were to
3 z = 99.7 % fall outside of the 1z, 2z, or 3z
2 z = 95.5 % control limits, we would be
1 z = 68.3 % 68.3%, 95.5%, or 99.7%
sure, respectively, that the
process had changed !
 X-Bar Chart Example
SOLUTION

The mean of all twelve ( 12 ) sample means is easily

calculated to be exactly 16.0 ounces and the population
standard deviation is calculated to be exactly one ( 1 ) ounce.
We therefore have:
=
X =the 16.0 oz. σx = 1.0 oz. n=9 z=3
process mean population standard deviation sample size wide control limits

THE CONTROL LIMITS ARE:

=
UCL = X + zσ _ = 16 + 3 ( 1/√9 ) = 16 + 3 ( 1/3 ) = 17.0 oz.
X
=
LCL = + zσ _ = 16 - 3 ( 1/√9 ) = 16 - 3 ( 1/3 ) = 15.0 oz.
X X
 X-Bar Chart Based on
Average Range Values

Process standard deviations ( σx ) are either not available

or difficult to compute. Consequently, we usually calculate
_
control limits based on the average range values ( R s )
rather than on standard deviations.

RANGE is defined as the difference between the largest

and smallest items in each sample.

For example, if the heaviest box of Oat Flakes in the 1st

hour weighed 19 ounces and the lightest box weighed
14 ounces, the range for that hour would be 5 ounces.
X-Bar Chart Control Limits
BASED ON AVERAGE RANGE VALUES

= _
UCL = X + A2 R
= _ where the

LCL = X - A2 R
process
standard
deviation
(σ)
is unknown
_
where:
R = average range of all the samples
A2 = value found in the Table
=
X = mean of the sample means
 The Factor Table
Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
2 1.880 3.268 0
3 1.023 2.574 0
4 .729 2.282 0
5 these values
approximate
the values of
.577 2.114 0
6 .483 2.004 0
the unknown
standard errors,
and also reflect
the “t” values

7 for small
samples, since .419 1.924 0.076
the “z” values
cannot be used
 The Factor Table
Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
8 0.373 1.864 0.136
9 0.337 1.816 0.184
10 0.308 1.777 0.223
12 0.266 1.716 0.284
14 0.235 1.671 0.329
16 0.212 1.636 0.364
 The Factor Table
Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
18 0.194 1.608 0.392
20 0.180 1.586 0.414
25 0.153 1.541 0.459
 X-Bar Chart Control Limits
AVERAGE RANGE VALUE EXAMPLE

Super Cola bottles soft drinks labeled “net weight 16 ounces”. An overall
process average of 16.01 ounces has been found by taking several batches
of samples in which each sample contained five ( 5 ) bottles. The average
range of the process is .25 ounces.

Determine the upper and lower control limits for averages in this process.
Solution

Looking in the Table for a sample size of 5 ( n = 5 ) , in the “mean factor

A2” column, we find the number “ .577 “ Thus, the upper and lower control
chart limits are:
 The Factor Table
Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
2 1.880 3.268 0
3 1.023 2.574 0
4 .729 2.282 0
5 .577 2.114 0
6 .483 2.004 0
7 .419 1.924 0.076
X – Bar Chart UCL and LCL

UCL = 16.01 + ( .577 )( .25 )

= 16.01 + .144
SOLUTION = 16.154 ounces

LCL = 16.01 - ( .577 )( .25 )

= 16.01 - .144
= 15.866 ounces
 R - Charts

 We are also interested in the process dispersion or variability.

 Even though the process average is under control, the process

variability ( uniformity ) may not be.

 For example, something may have worked itself loose in a piece

of equipment. As a result, the sample means may remain the
same but the variation within the samples could be entirely too
large.

 The theory behind Range Charts is the same as for the process
average control charts.
 R-Chart Formulas
Limits are established that contain + / - three ( 3 ) standard deviations
of the distribution for the average range “ R “ :

_
UCLR = D4 R
_
LCLR = D3 R
where :
UCLR = upper control chart limit for the range

LCLR = lower control chart limit for the range

D4 and D3 are values from the Table

 The Factor Table
Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
2 1.880 3.268 0
3 1.023 2.574 0
4 .729 2.282 0
5 .577 2.114 0
6 .483 2.004 0
7 .419 1.924 0.076
 Sample Size Mean Factor Upper Range Lower Range
(n) ( A2 ) ( D4 ) ( D3 )
2 1.880 3.268 0
3 1.023 2.574 0
4 .729 2.282 0
for n = 5 5 .577 2.114 0
6 .483 2.004 0
7 .419 1.924 0.076
 R-Chart
The average range of a process is 5.3 pounds. If the sample size
is five ( 5 ) , determine the upper and lower control chart limits.

SOLUTION
Looking in the Table for a sample size of “ 5 “ , we find that
D4 = 2.114 and D3 = 0
_
EXAMPLE UCLR = D4 R
= ( 2.114 )( 5.3 pounds ) = 11.2 pounds
_
LCLR = D3 R
= ( 0 )( 5.3 pounds ) = 0 pounds
Attribute Control Charts

c – Charts
Count the
p – Charts number
Measure the of
“percent” defects
defective in a
in a sample
sample
 p - Charts

 The chief way to control attributes

 The normal distribution can be used to calculate
p - charts when sample sizes are large enough*
 The procedure resembles the x - bar chart
approach which was based on the central limit
theorem. * ATTRIBUTES NORMALLY FOLLOW THE BINOMIAL PROBABILITY DISTRIBUTION
p-Chart UCL and LCL Formulae
_
UCLp = p + z σ p^
_ ^
LCLp = p – z σ p

where:
_
p = mean fraction defective in the sample

z = the number of standard deviations

( z = 2 for 95.5% limits ; z = 3 for 99.7% limits )

σ^
= standard deviation of the sampling distribution
p
 The Standard Deviation
of the
Sampling Distribution

p(1–p)
σ^p =
ESTIMATED
BY THE

n FORMULA
SHOWN
HERE
WHERE n = THE SIZE OF EACH SAMPLE
 p-Chart
Data entry clerks key in thousands of insurance records each day.
100 records entered by each clerk were carefully examined to make
sure they contained no errors. Twenty ( 20 ) clerks were examined.
The number of errors for each of the 20 clerks were computed and
shown below:
Clerk Number Clerk Number
Number Errors Number Errors
1 6 11 6
2 5 12 1
3 0 13 8
4 1 14 7
EXAMPLE
5 4 15 5
Basically, 6 2 16 4
we have
20 samples 7 5 17 11
containing
100 items
8 3 18 3
each 9 3 19 0
10 2 20 4
 p - Chart EXAMPLE

_ total number of errors

p=
total number of records examined
80 NOTE:
“100”
= = .04 IS THE
(100)(20) SIZE
OF
EACH
SAMPLE ( n )

^ (.04)(1 - .04)
σp = = .02
100
 p-Chart
EXAMPLE
_
UCLp = p + z σ^p = .04 + 3 (.02) = .10

_ BECAUSE
WE

LCLp = p – z σp^ = .04 – 3 (.02) = 0

CANNOT
HAVE A
NEGATIVE
PERCENT
DEFECTIVE

Setting the control limits to include 99.7% of the random variation

in the data entry process when it is in control ( z = 3 )
p-Chart Structure & Elements

DATA ENTRY EXAMPLE

.11
.10 UCLp = .10
.09
.08
DEFECTIVE

.07
FRACTION

.06
.05
.04 p = .04
.03
.02
.01
.00 LCLp = .00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SAMPLE NUMBER
 p-Chart
Data entry clerks key in thousands of insurance records each day.
100 records entered by each clerk were carefully examined to make
sure they contained no errors. Twenty ( 20 ) clerks were examined.
The number of errors for each of the 20 clerks were computed and
shown below:
Clerk % Clerk %
Number Errors Number Errors
1 .06 11 .06
We can
2 .05 12 .01
then 3 .00 13 .08
convert
the
4 .01 14 .07
number 5 .04 15 .05
of errors
to the
6 .02 16 .04
percentage 7 .05 17 .11
of errors
8 .03 18 .03
9 .03 19 .00
10 .02 20 .04
p-Chart Structure & Elements
..and plot
them on
the newly DATA ENTRY EXAMPLE
constructed
control chart

.11
.10 UCLp = .10
.09
.08
DEFECTIVE

.07
FRACTION

.06 X
.05
.04 p = .04
.03
.02
.01
.00 LCLp = .00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SAMPLE NUMBER
 c-Charts
 Used to control the number of defects per unit of
output.
 Used to monitor processes where a large number
of potential errors can occur but the actual number
that do occur is relatively small.
 Defects may be bad circuits in a micro-chip, burrs
on cloth or metal, blemishes on furniture, etc.
 c-Chart Variables

_
c = the mean number of defects per unit
as well as the variance.
_
√ c = the standard deviation of defects per
unit.
THE POISSON PROBABILITY
DISTRIBUTION IS THE BASIS
FOR c - CHARTS
c-Chart Control Limits

_ _
UCL = c + 3 √ c
_ _
LCL = c - 3 √ c
_
TO COMPUTE 99.7% CONTROL LIMITS FOR c
 C-Chart

EXAMPLE
Red Top Cab Company receives several complaints per
day about the behavior of its drivers. Over a 9-day period
( where days are the units of measure ) the owner receiv-
ed the following numbers of calls from irate passengers:
3, 0, 8, 9, 6, 7, 4, 9, 8
for a total of 54 complaints.

COMPUTE 99.7% CONTROL LIMITS

 c-Chart
EXAMPLE
_
c = ( 54 / 9 ) = 6 complaints per day
therefore:
UCLc = 6 + 3√ 6 = 6 + 3 (2.45) = 13.35

LCLc = 6 - 3√ 6 = 6 - 3 (2.45) = 0.00

After the owner plotted a control chart
summarizing these data and posted
it prominently in the drivers’ locker
room, the number of calls
received dropped to an
average of three
calls per day.
Can you explain why this occurred?
 c-Chart Example
15 It doubles
14 as a
13 Behavior UCLc = 13.35
Modification
12 Chart !
11
COMPLAINTS

10
NUMBER

9
OF

8
7
6 c = 6.00
5
4
3
2
1
0 LCLc = 0.00
1 2 3 4 5 6 7 8 9
DAY
Deciding Which Chart To Use
X-Bar and R-Charts

I. The observations are usually products

which are measured for size and weight *

II. We collect 20 to 25 samples of n = 4, n = 5,

or more, each from a stable process, and
compute the mean for an x-Bar Chart and
the range for an R-Chart.

III. We then track samples comprised of “n”

observations per sample.
* WEIGHT OF A CAN OF SOUP OR LENGTH OF A WIRE BEING CUT
 p-Chart
I. Observations are attributes that can be
categorized as good or bad, pass or fail,
functional or broken, i.e. two states.
II. We deal with fraction, proportion, or
percent defectives.
III. There are several samples with many
observations in each *
FOR EXAMPLE,
TWENTY SAMPLES
OF
n = 100
OBSERVATIONS
EACH
 c-Chart

I. Observations are attributes whose defects

per unit of output can be counted.

II. We deal with the number counted, which is

a small part of the possible occurrences.
Defects may be the number of blemishes on
a desk, crimes in a year, flaws in a bolt of
cloth, or typos in a newspaper.
Statistical Process Control
Via QM for WINDOWS
 Scroll To “ Quality Control “

Applied Management Science for Decision Making, 2e © 2014 Pearson Learning Solutions
Click “New”
to solve a
new problem
 We want to work with
an ‘ x - bar ‘ control chart
 We already know
the
mean of the
sample means
=
X
and the
standard deviation
σ
 In this application,
the sample size of
nine (9) is
insignificant

The ‘Data Input Table’

provides for the
insertion of the
process mean
We desire to set the =
control limits at X
“ 99.7 % “ and the standard deviation
σ
 The UCL = 17.0
The LCL = 15.0
The CL = 16.0

if
=
X and σ are known
If the process mean equals
16.01 ounces and the
average range equals
.25 ounces, and the
sample size equals ‘5’,
we can find the
UCL and LCL
for the x-Bar chart
 The Data Input Table
makes provisions
for the
process mean ( 16.01 )
and the
average range ( .25 )
and the
sample size ( n = 5 )
and for
3 - sigma control limits
For the X-bar Chart:

UCL = 16.1543 oz.

LCL = 15.8658 oz.

For the Range Chart:

UCL = .5288 oz.

LCL = 0 oz.
CL = .25 oz.
The x- Bar Chart
The Range Chart
Click on “ p-charts”
to build a
‘p’ control chart
Insurance records
of 20 clerks were
Examined
Each sample contains 100 records

The number of errors in each

sample are listed

We want 99.7% control limits

on this p-chart ( 3 sigma )
 The
Control Limits
and
Center Line
( 3 sigma )
 One Sample
Violated The
Upper Control Limit
Click on “c-charts”
to build a
c - Chart
 Complaints were
accumulated daily
over 9 days
We want to build a
c - Chart
with 99.7%
control limits
( 3 sigma )
Statistical Quality Control Using
 Template
and
Sample Data
UCL = 15.85556 + (3)(0.357771) = 16.92887

LCL = 15.85556 - (3)(0.357771) = 14.78224

Template and Sample Data
 For sample size ( n = 5 )
_
UCL = 2.115 ( 2.511111 ) = 5.311
UCL = ( D4 ) ( R )
_ LCL = 0 ( 2.511111 ) = 0
LCL = ( D3 ) ( R )
 Template
and
Sample Data
UCLp = .04 + ( 3 ) ( .02 ) = .10

LCLp = .04 - ( 3 ) ( .02 ) = 0

Applied Management Science for Decision Making, 2e © 2014 Pearson Learning Solutions
 Template
and
Sample Data
UCL = 6 + ( 3 ) ( √6 ) = 13.35

LCL = 6 - ( 3 ) ( √6 ) = 0.00
 Statistical Process Control
Manufacturing & Service Sectors

Applied Management Science for Decision Making, 1e © 2013 Pearson Prentice-Hall, Inc. Philip A. Vaccaro , PhD