Statistical Process Control

Statistical Process Control

Purpose
Introduce the concepts of Statistical Process Control

Objectives
1. Be able to use “Xbar and S Charts” for analysis of
continuous data.
2. Be able to use “p” Control Chart for analysis of
discrete data.
3. Be able to determine control limits for each type of
chart.
4. Be able to interpret charts and determine when a
process is “Out of Control”.
5. Be able to explain the importance of taking action
based on the chart information.

06A.1
Statistical Process Control

What is: Statistical

Process Control (SPC)
Statistical -- Probability based decision rules.

Process -- Any repetitive task or steps.

Control -- Monitoring of process performance.

The analysis is based on the same concept as

the “t test” hypothesis test. It allows decisions
about the process to correct problems before
they affect the output. SPC will signal when the
process is “out-of- control”.
Your mission is to find out
why, then fix it so it will
never happen again

SPC
SPCsignals
signalswhen
whenthe
thesteady-
steady-
state
stateprocess
processvariation
variationhas
has
been
beeninfluenced
influencedby
byoutside
outside
assignable
assignablecauses.
causes.

06A.2
Statistical Process Control

Six Sigma and SPC

Six Sigma Quality focuses on moving
the control upstream in a process to
leverage control of the input
characteristics or vital X’s.

Y
Y
X3
X2
X1 Provide increased

X
X stability to
Dependent variables;
Responses Y1,Y2,..., Ym
Control Charts applied to
Process variables;
Independent variables;
Design variables X1, X2,..., Xk

06A.3
Statistical Process Control

When is SPC used?

• What do I want to know?

– Has the critical process variable
(X or Y) output changed over time?
(i.e., Is the process stable?)
• How do I want to see it?
– Real time data based graphical
display showing process changes.

SPC
SPCisisaarigorous
rigorousprocess
processwhich
whichrequires
requires
active
activeengagement
engagement ofof the
theoperational
operational
team
teamtotocollect
collectand
andanalyze
analyzethe
thedata.
data.

06A.4
Statistical Process Control

Control Chart Components

XXBar
BarChart
Chart

102.00
UCL Upper
UpperControl
ControlLimit
Limit
101.00

Avg
100.00
X LCL-A
99.00 Grand
GrandAverage
Average
Avg-Gd
Central
CentralLine
Line
98.00 LCL UCL-A

97.00
Out
OutofofControl
ControlCondition,
Condition,
96.00 Record
Recordaction
actiontaken
takentotofix
fix. . Lower
LowerControl
ControlLimit
Limit
1 3 5 7 9 11 13 15 17 19 21 23 25

Sample
Sample/ /Subgroup
Subgroup(time
(timeordered)
ordered)

Sigma
SigmaChart
Chart
7.00

6.00 UCLs
5.00 Range
4.00 LCL-R
Average
AverageSigma
3.00
s AVG-R
Central
Sigma
Line
2.00 UCL-R Central Line

1.00

0.00
LCLs
1 3 5 7 9 11 13 15 17 19 21 23 25

06A.5
Statistical Process Control

Control Charts
Developed in the 1920s by Dr. Walter Shewhart of
the Bell Laboratories, Statistical Process Control
charts provide a graphical comparison of a measured
process observation against statistically computed
“control limits”(expectation).

• Plot performance over time.

• A process change could involve mean and/or
variance differences, so we always plot average
and variance control charts in pairs (Xbar and S)
• Control limits for the Mean represent two tail
hypothesis test limits to infer sameness or
difference of the observed sample mean.
• The control limits for Sigma or Range represent
where variance would show difference.

06A.6
 Statistical Process Control

Hypothesis test?
A Control Chart is a graphic display of a continuing two
tailed test with HO and HA defined as:

Ho: i

Ha: i

/2
/2 For
For3 limits,
3limits, ==0.00135
0.00135

UCLx

Note: the approximate

X confidence level is 99.7%.

LCLx

/2
/2

When a subgroup average falls outside of the control

chart limits, it makes a graphical statement that a
difference exists between the mean of that sample
and the historical average.

06A.7
Statistical Process Control

Process Stability
The graphic below shows a variety of unstable
process examples. Control chart interpretation
can assist in identifying when these are
occurring as well as what condition exists.

Condition C1: C 2: C3: C 4: C5:

t1:
Time
t2:

t3:

t4:

t5:

t6:

t7:

Mean: SUSTAINED IRREGULAR TREND CONSTANT IRREGULAR

SHIFT SHIFT

Stdev: CONSTANT CONSTANT CONSTANT DECREASED IRREGULAR

Copyright 1995 Six Sigma Academy, Inc.

An
An unstable
unstable process
process has
has no
no
predictable
predictable behavior
behavior and
and good
good
performance
performance may
may not
not be
be
sustainable.
sustainable.
06A.8
Statistical Process Control

Process Stability
• A process output is considered stable when it
consists of only common-cause variation.
• Subgroup averages and variation measures
are between their control limits and display
no evidence of assignable-source (special-
cause) variation.
• If non-random patterns of data appear on the
control chart, or when a point is beyond the
control limits, then this is a strong signal that
assignable-source (special-cause) variation
is present in your process.
region of assignable-source variation
UCL

region of stable process variation

X
(only common-cause variation)

LCL
region of assignable-source variation

AAstable
stableprocess
processwill
willrarely
rarelyproduce
producean anoutput
output
outside
outsideofofthe
theplus
plusand
andminus
minusthree
threesigma
sigmaregion.
region.

06A.9
Statistical Process Control

Types of Control Charts

Types
Types of
of Control
Control Charts
Charts
There are two types of
Variables
Variables Charts
Charts Attributes
Attributes Charts
Charts
control charts:
for
for monitoring
monitoring
continuous
continuous X's
X's
for
for monitoring
monitoring
discrete
discrete X's
X's
Variables charts - Used
for monitoring X variables
Ave
Ave && Range
Range Fraction Defective that are continuous, such
X
X bar & &R
nn << 10,
10,
p Chart
typically n > 50
as a diameter or consumer
typically
typically 33 -- 55 tracks dpu/dpo satisfaction rating.
Average
Average & & Number Defective
Std
Std Deviation
Deviation np Chart
X
X bar
bar &
& ss n > 50 (constant)
nn > 10
10 tracks # def Attribute charts - Used
for monitoring discrete X
Median & Range Number of
X&R Defects variables, such as
n < 10, c Chart
typically 3 - 5 c>5 good/bad counts, or
inventory levels.
Individual & Number of
Moving Range Dev / Unit
XmR u Chart
n=1 n variable

In
Inorder
ordertotoselect
selectthe
theappropriate
appropriatecontrol
control
chart
chartto
tomonitor
monitoryour
yourprocess,
process,first
first
determine
determineififthe
thekey
keyprocess
processvariables
variables(X’s)
(X’s)
are
arecontinuous
continuousor ordiscrete.
discrete.

06A.10
 Statistical Process Control

The X Bar Sigma

(Xbar-S) Control Chart
Xbar/S Chart for Evaluations

4.3
Used to analyze and
3.0SL=4.232
4.2
control continuous process
Sample Mean

4.1 X=4.096

variables.
4.0
-3.0SL=3.959
3.9
3.8

You can use Xbar-S charts

1 1
3.7
Subgroup 0 5 10 15 20 25

0.6
1
– In Measure, to
0.5

graphically separate
Sample StDev

0.4
0.3

common causes of
3.0SL=0.2409
0.2
S=0.1403
0.1

variation from special

-3.0SL=0.03982
0.0

causes
The
The Xbar-S
Xbar-S chart
chart is
is best
best – In Analyze and Improve
generated
generated using
using Minitab
Minitab oror to check for process
other
other statistical
statistical software
software stability before
package.
package. IfIf software
software isis not
not completing a hypothesis
available,
available, use
use Xbar-R
Xbar-R or or test.
other
other manual
manual control
control – In Control to verify
charts.
charts. process control following
improvement work.

06A.11
Statistical Process Control

Constructing Xbar-S
using Minitab
File: GEAPPS>6Sigma>Minitab>Training>Minitab>
Session 4> control chart.mtw
Select Stat > Control Charts > Xbar-S

06A.12
 Statistical Process Control

Constructing Xbar-S
using Minitab
Select the column with the response data and either
enter a value for the subgroup size or identify the
subgroup subscript column(in this case, “Week”)

Select “Tests”.

Determine the criteria

for “Out of Control
Conditions”. Select
either “Perform all
eight tests” or choose
between the eight
tests provided.
06A.13
 Statistical Process Control

Voila!!… Now What?

Minitab generates the Xbar/S Chart for Evaluations
Xbar-S chart . It 4.3

automatically
3.0SL=4.232
4.2

Sample Mean
4.1 X=4.096
4.0

calculates the control

-3.0SL=3.959
3.9
3.8
1

limits. The out-of-

3.7 1
Subgroup 0 5 10 15 20 25

control points are Sample StDev

0.6
0.5
0.4
1

shown on the graph 0.3

0.2
3.0SL=0.2409
S=0.1403

and summarized in
0.1
-3.0SL=0.03982
0.0

the session window.

The numbers on the chart correspond to the eight tests

used to determine an “Out of Control” condition.

06A.14
Statistical Process Control

Analyzing the Control Charts

Xbar/S Chart for Evaluations

4.3
3.0SL=4.232
4.2
Sample Mean

4.1 X=4.096
4.0
-3.0SL=3.959
3.9
3.8
1 1
3.7
Subgroup 0 5 10 15 20 25

0.6
1
0.5
Sample StDev

0.4
0.3
3.0SL=0.2409
0.2
S=0.1403
0.1
-3.0SL=0.03982
0.0

• The week 7 & 16 evaluation averages fell below the

lower limit of 3.957. . . They are out of control.
• This change was driven by some assignable cause
(either system-related or region-initiated).
• Investigate, identify and fix the assignable source of
the variation. Record this on the chart at the time of
occurrence.
• The variation among the regional centers for week 7 is
larger than expected. This also requires investigation,
correction and notation.

An
An“out-of-control”
“out-of-control”indication
indicationcancancome
come
from
fromeither
eitherchart
chart. .
06A.15
Statistical Process Control

Calculating Average Chart

Control Limits
To determine the value of the control limits for mean, you
must first calculate the Overall Process Average
(Grand Mean):

X
X 1  X 2  ... X k  Overall
OverallProcess
ProcessAverage
Average
k
where k = the # of subgroup averages

The Upper Control Limit:is found by the following formula:

UCL X  X  3 / n
The formula for the Lower Control Limit:

LCL X  X  3 / n

For
Forlarger
largersample
samplesizes,
sizes,the
thelimits
limitsfor
foraagiven
givenprocess
process
will
willbe
benarrower,
narrower,and
andthe
thechart
chartsensitivity greater..
sensitivitygreater

06A.16
Statistical Process Control

Calculating Variation Chart

Control Limits
• To define the control limit for “s”,
first calculate “s” for each
subgroup.

 (x
Where ni=number
si  ij  xi ) 2
(ni  1) of observations in
j
the ith subgroup.

• Next calculate average “S”

s    si  k
Where k= the number of subgroups

• Determine the control limit lines.

– Calculations are based on concepts
similar to Mean chart, but are complex.
Luckily, Minitab calculates these limits.

06A.17
 Statistical Process Control

Larger Subgroup Size

Increases Sensitivity

nn==33 nn==10
10 nn==25
25
UCL
UCL
UCL
UCL
UCL
UCL
LCL
LCL
LCL
LCL
Copyright 1995 Six Sigma Academy, Inc.
LCL
LCL

As sample size increases, the control limits draw closer

together. This increases the sensitivity. That is, it
increases the probability of detecting a change.
Control Chart sensitivity is relative to the ratio of the
square roots of the sample size. i.e., a sample size of
25 is 2.5 times as sensitive as a sample of 4 (5/2).
To take advantage of the Central Limit Theorem,
subgroup size must be greater than 2.

06A.18
B22.14
Statistical Process Control

Why Three Sigma

Control Limits?
• Three sigma limits have withstood the test of time.
• The 3 sigma limit results in an  of approximately .
00135. This gives a low chance of reacting when
the process did not truly change. This is important
since there are a large number of tests being
conducted over time.

22  -- 95%
95% Confidence
Confidence Interval
Interval for
for an
an average
average
33  -- 
=.003
=.003 (Reason:
(Reason: Many
Many sequential
sequential tests;
tests;
reduces
reduces potential
potential for
for error.)
error.)
4.5  -- Long
4.5 Long term
term process
process performance
performance goal
goal for
for
individual
individual measurement
measurement compared
compared toto
customer
customer requirement.
requirement.
6.0  -- Short
6.0 Short Term
Term process
process variation
variation goal
goal for
for
individual
individual measurement
measurement compared
compared toto
customer
customer requirement.
requirement.

3  limits provide good sensitivity to change

with low potential for over-reacting
when the process is stable.

06A.19
Statistical Process Control

Variables control chart Example

A consumer services organization wants to monitor
“consumer satisfaction” for their company. Each week, a
survey from each of the company’s ten regional service
centers is evaluated and the scores are tabulated. The
following is an example of how an Xbar-S control chart could
be used to monitor “consumer satisfaction.” (In this example,
higher is better.)
The vital information for creating an Xbar-S control chart:
Total subgroups = 25
Subgroup size, n = 10
Grand average, X = 4.096
S=.1403

Control Limit Formulas:

UCLX  X  A3  s

LCL X  X  A3  s
UCL  B4  s
LCL  B3  s
Actual Control Limit Calculations for the Data
See
UCL = 4.096 + (.975 x 0. 1403) = 4.232 Seethe
theSPC
SPC
table of
table of
constants
constants
LCL = 4.096 - (0.975 x 0.1403) = 3.959 on
onthe
thenext
next
page
page
UCLR = 1.716 x 0.1403 = 0.2408
LCLR = 0.284 x 0.1403 = 0.0398

06A.20
Statistical Process Control

Control Chart Constants

& Control Limits
The following table contains the various constants
that are used for constructing SPC control charts.
Variables Control Chart Control Limit Constants
n A2 A3 D3 D4 B3 B4 d2 c4
1 2.660 3.760 - - - - - -
2 1.880 2.659 0 3.267 0 3.267 1.128 0.7979
3 1.023 1.954 0 2.575 0 2.568 1.693 0.8862
4 0.729 1.628 0 2.282 0 2.266 2.059 0.9213
5 0.577 1.427 0 2.115 0 2.089 2.326 0.9400
6 0.483 1.287 0 2.004 0.03 1.970 2.534 0.9515
7 0.419 1.182 0.076 1.924 0.118 1.882 2.704 0.9594
8 0.373 1.099 0.136 1.864 0.185 1.815 2.847 0.9650
9 0.337 1.032 0.184 1.816 0.239 1.761 2.970 0.9693
10 0.308 0.975 0.223 1.777 0.284 1.716 3.078 0.9727

• The standard deviation used to calculate the

control chart limits is is based on the kind of chart
being plotted.
 For the Xbar chart, it is the standard deviation
of the subgroup means. This is similar to a
pooled standard deviation.
 For the S chart, it is the standard deviation of
the subgroup standard deviations.
• Both formulas are dependent on the size of the
subgroup.

06A.21
Statistical Process Control

Control Chart Use

Control charts can be used in the Measure
and Analyze phases to track process
changes. Analyze for significant changes and
record.
What
What
Xbar/S Chart for two
caused
caused
54 1
1
1 1
this?
this?
53 1 1111
5 5 2 2 3.0SL=52.63
Sample Mean

52 5
2
51 22 2
50 X=50.00
49
48
-3.0SL=47.37
47
Subgroup 0 50 100

4 3.0SL=3.849
Sample StDev

2 S=1.842
2
1

0 -3.0SL=0.00E+00

Control Charts are used in controlling a

process to sustain the improvement.
– Use the chart to monitor and record an
input variable (X), analyze for changes
and control.

06A.22
Statistical Process Control

Changing Control
Chart Limits
• It is best to use historical stable process limits for
control charts, versus limits which change with each
observation. Historical limits define the “expected”
range of data, or the “null hypothesis (Ho)”. (Use
Historical settings in Minitab)
• Change limits when:
– A process change is made and is determined to be statistically
significant(I.e. Ha.).
– A definable practical process change is completed.

06A.23
Statistical Process Control

Chart Interpretation

Charts are interpreted before establishing

Process Capability and on an ongoing
process control basis.

– Interpret Sigma chart first.

– During the initial capability analysis, if you can

identify the special cause variation which
causes “OOC” conditions, you can remove
these points from consideration when you
calculate the control limits.

06A.24
Statistical Process Control

Normal Process
Variation
“Boring…”

This chart represents a predictable process in which

variation is subject only to random variation.

The points move up and down unpredictably, but tend to

cluster ( however, not too closely) around the centerline
and stay inside the control limits.

This pattern is the target for any control chart. It does not
necessarily indicate the best capability of the process nor
that the process can meet the specification, but it indicates
that the process is stable.

06A.25
 Statistical Process Control

Special Cause Change

“What happened?”

Occasionally a factor will

enter a process and cause a
sudden, short-lived change.

S A cause such as this will be

seen in the xBar chart as a
cluster of points at an out-of-
control level. The S chart will
Some Typical Causes: generally not be affected by
• Introduction of a batch of these shifts.
out-of-spec material into
the process
• A temporary shift in
measurement calibration
• A different inspector
• A different type of tool

06A.26
 Statistical Process Control

Process Outliers
“Ah Ha! Now here’s something
interesting.”
Hiccup
Hiccup
Sometimes a process will
get the hiccups and as a
result these “mavericks”
or “fliers” will occur as
Hiccup occasional values that are
clearly not a part of the
basic process distribution.
After one occurs, the
process behaves in its
normal manner until the
Some typical Causes:
• Mistake in measuring next hiccup.
• Bottom piece (or top piece
in stack)
• End of bar, coil, etc.
• Dirt or foreign material

06A.27
Statistical Process Control

Sudden Process Shift

“What did you do?”

Symptom:
Nine points in a row one
side of center After a change, the
process makes the
parts larger on the
average or the yield
greater or the
hardness higher and
so on.
The basic variability of
the process has not
changed, and the
range chart does not
show a change.
Some typical Causes:
• Mis-adjustment or
improper setting
• Change of material or
lubricant,
• Shift change

06A.28
Statistical Process Control

Process Trend
“Where’s it going?”
Symptom:
A trend is a gradual •7 points in a row trending up.
shift of the process •7 points in a row trending down
level, reflected on the
xBar chart alone.
Sometimes raw
material,
measurement or
human factors can
cause a trend but this
is unlikely. The trouble
is usually in the Some typical Causes: :
equipment itself, in the
power supply or prior • Often associated with “tool wear”
process environment. since it often shows tool wear
(usually of an aging or wear out
nature)
• Example:
• bath depletion in a plating
operation or various chemical
operations
• tube wearout in an electronic
circuit

06A.29
Statistical Process Control

Zone Testing
Process is “Out of Control” if

Test
Test Zone
Zone Test
TestCriteria
Criteria
1 • 1 point above +3 sigma
2 A+ 2 out of 3 in A+ or above
3 B+ 4 out of 5 in B+ or above
4 C+ 7 out of 8 in C+ or above
5 C- 7 out of 8 in C- or below
6 B- 4 out of 5 in B- or below
7 A- 2 out of 3 in A- or below
8 • 1 point below -3 sigma
Copyright 1995 Six Sigma Academy, Inc..

The Zones of Probability

A+
B+
C+
C-
B-
A-

06A.30
 Statistical Process Control

Average and Range

Charts (Xbar - R)
Xbar/R Chart for Evaluations

• Since Sigma 4.3

3.0SL=4.229
4.2

calculations are Sample Mean 4.1

4.0
X=4.096

cumbersome when
-3.0SL=3.963
3.9
3.8

done manually, the Subgroup

3.7
0 5
1
10 15
1

20 25

Xbar - R chart is the 1

1.5
preferred method for
Sample Range

1.0

manual control charts. 0.5

3.0SL=0.7673

R=0.4318

-3.0SL=0.09634
0.0

• Variation is shown by a calculation of the Range

of data within the subgroup (Largest - smallest).
• Use A2R bar as approximation of 3/sqrt(n). Use
D3 and D4 times Rbar to find the lower and upper
control limits for Range variation.
• Analysis is done in similar fashion to Xbar - S.

06A.31
 Statistical Process Control

Individuals and Moving

Range Charts (XmR)
The Individual X - Moving
Range chart is
appropriate where there is
no measurable variation
within a subgroup (such
as process temperature,
pressure or other
homogeneous
measurement), or when
rational sub grouped data
is not available (due to
cost or other limiting
factors).

Caution
• When used inappropriately with a process having “within
subgroup” variation, (example control chart data shown above)
the resulting chart sometimes is difficult to read and use.
• When tracking individual measurements, there is no information
about short term and long term variation differences.

06A.32
Statistical Process Control

Exponentially Weighted
Moving Range (EWMA)
Charts
The EWMA Chart is much 4.15
EWMA Chart for Evaluati

more sensitive than any

3.0SL=4.148

of the Shewhart Control

EWMA
4.10
X=4.096

Charts. Each EWMA

4.05

Chart data point -3.0SL=4.043

0 5 10 15 20 25

incorporates information Sample Number

from all of the previous observations, and the chart

can be configured to detect any size shift in the
process. This programmable sensitivity makes them
an excellent tool for monitoring in-control processes.

Note the appearance of the EWMA chart for the same

data we used for the continuous data charts earlier.
We can note trends up in the average broken by
significant average shifts down.

06A.33
Statistical Process Control

Attribute Control Charts

• np
• p

Principle Attribute Charts

• np - Measures counts of defects. Control
limits are based on the Binomial Distribution.
Because raw counted defects are recorded,
subgroup size needs to be equal.

• p - Records the fraction defective for a

sample. Control limits are based on the
Binomial Distribution. Because the proportion
is a ratio of defects to sample size, sample
size does not have to be equal.

06A.34
Statistical Process Control

Attribute Control Chart Example

A local dental group wanted to know why a lot of their patients fail to
keep their appointments. A problem solving team was assembled,
and decided to use a p Chart to track the percentages of “no
shows”. The dental clinic began logging monthly percentages of “no
shows” for each month. Since a “no show” is a defective
appointment, the average total fraction defective is called p.
Calculate control chart limits based on the first 6 months. Use
sample size (n) of 100 appointments per month.
Year 1996
Month Jul Aug Sep Oct Nov Dec
% Failed 40 36 36 42 42 40
Year 1997
Month Jan Feb Mar Apr May Jun
% Failed 20 26 25 19 20 18
Month Jul Aug Sep Oct
%Failed 16 10 12 12

p Chart Formulas:
number of nonconforming units in sample(d )
p
sample size (n )
total number of defective units in all samples (d 1  d 2  ... d n )
p
n1  n2  ...  nk
   
p  1  p  p  1  p 
UCL = p  3 and LCL  p  3
n n

p = 236/600 = 0.393, where di = 40+36+36+42+42+40 = 236

ni=600, the total number sampled in 6 months

UCL = .393+3((.393*.607)/100)½ = 0.5395

LCL = .393- 3((.393*.607)/100)½ = 0.2465

06A.35
 Statistical Process Control

Measurement Performance
Over Time

• Control limits were established from the 1996 “no show” data.
• The advocacy team examined and prioritized the various
reasons why patients were missing their appointments.
• The team determined that providing flex-time for patients
resulted in fewer missed appointments.
• A new flexible appointment policy was adopted in January 1997.
• The control chart shows a dramatic reduction in the number of
missed appointments after the implementation of the flex-time
policy.
• By adopting the new appointment policy, the team was able to
reduce the average percentage of “no shows” from 40% to
20% (20% is the new average for 1997 data only).

06A.36
 Statistical Process Control

Creating the P Chart

File: GEAPPS>6Sigma>Minitab>Training>Minitab
>Session 4>Patients.mtw
Data Sheet Stat>Quality Tools>P Chart

You’ll need two

columns on the
spreadsheet,
one for the count
data and a
second to
identify the
subgroup

Once you have opened

the dialog box, identify
the count column for
the “Variable”. Fill in
the subgroup size (n),
and the historical value
for p (in this case, the p
for the 1996 year data).

06A.37
Statistical Process Control

Improving the Graphic Output

Select the
“Annotate>Title”
button. Enter the
graphic title using the
available lines and
format. Click OK

Next, select the “Stamp”

button.
In that dialogue box,
identify the subgroup
identification
information; I.e.,
months. Click OK.

Next, select the

“Frame>Tick” button.
In that dialogue box,
identify any special
settings of the axis
ticks to make the
graphic easier to
use.

06A.38
Statistical Process Control

Fine Tuning
If there is important
information you want
footnoted, use the
Annotate>Footnote
button. In this case,
the historical P value
reference.

If a reference line is
desired, use the
Frame>Reference button.
In this case, the line at
December 1996 to show
end of year.

Finally, with the graphic

showing, double click the
graph window to open the
editing palettes. These
were used to insert the
year dates and to color
them

06A.39
Statistical Process Control

Some Practical Concerns

in Applying
Statistical Process Control
• Requires a disciplined approach to process
management and data gathering.
• Works best within an automated or semi-automated
environment. (It is a real-time process monitoring tool.)
• Requires an appropriate response to an “out-of-control”
condition.
• The sensitivity of a control chart to detect non-random
variation can be improved by increasing the subgroup
sample size.
• It may be appropriate to recalculate control limits from
rebaseline or confirmation run results.
• Recalculate new control limits only when the process
variation has truly changed (stabilized).
• While it is acceptable to compute temporary control
limits after 5 to 10 subgroups, permanent limits require
at least 25 subgroups of data points that are “in-control”
for both the average and sigma charts.

06A.40
Statistical Process Control

SPC Chart Goals

Continuous
ContinuousData
Data Discrete Data

Target
Target

0%
0%Rejected
Rejected

SPC
SPCfor
forcontinuous
continuousvariables
variablesis
isused
usedto
tosteer
steeraa
process
processtowards
towardsaatarget.
target. Attribute
Attribute(Discrete)
(Discrete)
SPC
SPCcharts
chartsare
areused
usedto tominimize
minimizedefects.
defects.

06A.41
Statistical Process Control

Key Concepts: Tab 3 - SPC

• Statistical Process Control is an excellent upstream
process control tool. Control charts are ideally
suited for monitoring and controlling your “Vital
Few” X variables.
• Control charts monitor the process variation and
generate a signal when the process variation is
influenced by special cause variation.
• SPC Control charts can be used to monitor:
Continuous Variables
Xbar & S
Xbar & Range
Individuals & Moving Range (XmR)
EWMA Control Charts
Discrete Variables (Attribute)
p Chart
np Chart
• Basic tactics for using control charts:
– Plot data immediately.
– Identify and react to “out-of-control” conditions.
– Points outside the control limits
– Search for root cause for “out-of-control”.
– Put a permanent fix in place.
– Do not make adjustments if the process is not out-
of-control.
06A.42