Overview of Statistical

Process Control (SPC)

March 18, 2009

 SPC Defined
• Basic
– Shows the behavior of a characteristic over time
– Shows the influence of different variables on the
characteristic
– Shows where the process is located and how much
variation is present in the process
– Helps us get a process in a state of control
• Advanced
– Is the basis for establishing process capability
• Process capability defines how well (or not well) our process
can meet the needs of our customers
– Separates common cause variation from special
cause variation
2
Process Limits-Not Customer Limits

If the process remains stable and in control, we expect the

process’ output to run between these limits almost 100% of
the time. 3
The Relationship of Process Limits to
Customer Limits (Process Capability)

The process limits are wider than the customer’s

The process limits are tighter than the customer’s
limits. The process is not capable. You have three
limits. This is a capable process and no significant
options: Get your customer to relax his
action is needed other than make sure the process
requirement, 100% inspect the output of the
is followed.
process, or change the process to meet the
customer’s requirements

The process limits are marginally better than the customer’s The process limits are tighter than the customer’s limits but
limits. The process is centered so variation must be the process is off target (to high side). Adjust process to 4
reduced. Great case for six sigma. target. If the process can’t be adjusted, then reduce variation.
Great case for six sigma
 Common Cause and Special Cause
Variation
Lower Process Limit Upper Process Limit
• Common cause variation is
what we expect to happen
99.97% of the time if the
process is in control.
• Common cause variation
exists between the process
limits
• Special cause variation is not
expected to happen and has
assignable causes.
• Special cause variation occurs
outside the process limits
99.97%

5
 Common Cause Special Cause
Variation Variation
Histogram of Process Week One Histogram of Process Week One
20 14

12
15
10

If only common cause

Frequency

Frequency
8
10

5
6

4
variation is present in
0
1.68 1.76 1.84 1.92 2.00
2
the process, the
0
Peen Height 1.68 1.74 1.80 1.86
Peen Height

Histogram of Process Week Two

1.92 1.98
histogram will look the
Histogram of Process Week Two
12 12
same over time
10 10

8
8

Frequency
Frequency

6
6

If special cause
4
4

2
2

0
0
1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00
variation exists, the
1.68 1.74 1.80 1.86 1.92 1.98 Peen Height
Peen Height
Histogramof Process Week Three histogram will change
18

16
4
Histogram of Process Week Three
over time in location
14

12
3
and/or spread
Frequency

10
Frequency

8 2

4 1

0 0
1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00
Peen Height
Peen Height

6
 Examples of Common
Cause/Special Cause Variation
• Body Temperature
– If our body’s processes are in control, we expect temperature to vary slightly
above and below 98.6 degrees F. This is the common cause (or expected)
variation.
– If a virus (special cause) enters our body, a process will be altered and
temperature will spike significantly high
• Teenage Behavior
– Teenagers do teenage things. Always have and always will. Parents must
decide what is expected behavior and what is not expected behavior. The
former (good or bad) usually warrants a stern lecture while the latter deserves
punishment.
– Special causes often drive teenage behavior. The breakup by a girlfriend. Being
cut from a sports team. Making a bad grade. If they are not acting as expected,
we often must find the special cause before acting in return.
– An example-My eighteen year old often challenges me on my philosophies and
opinions. That is fine. I expect him to do that. I’m glad he does it. Sometimes
he goes too far with his mother and can be disrespectful. That’s outside the
boundaries of normal behavior and incurs my wrath.

7
 Exercise #1
• Run product from machine determine
upper and lower process limits
• Run product to see how the process limits
hold
• Introduce special causes
– Increased standard deviation
– Shift in average

8
 The Sections of a Control Chart:
Process Information
• Process information:
This is needed to keep
production records to go
along with the data.
• What you should record:
– Date data was collected
– Time data was collected
– Who collected the data

9
 The Sections of a Control Chart:
Subgroups
• The data is recorded in subgroups
• The subgroups are set up to be a
certain size. The size of a
subgroup is the number of
readings recorded.
– Typical sizes are three and
five
• A completed control chart is one
with at least twenty completed
subgroups on the page

10
 The Sections of a Control Chart:
Subgroup Statistics
• Once the data is recorded in the
subgroups, we need to perform
calculations for each subgroup
– A measure of where the process
is located. The mean (or average)
X shows us where the process is
located
– A measure of how much variation
R is in the process. The range
shows us how much variation is in
the process.

11
 What is a Mean?
• The mean is the
center of weight for
data. Also called
average.

50% 50%
Weight Weight

Mean

12
 How to Calculate a Mean
• Add up the measurements
and divide by the number
of measurements
– Add up measurements:
o 1.81+1.81+1.82
o Sum=5.44
o Number of measurements: 3
– Divide sum by the number of
measurements
5.44
= 1.813
3
Note: Always record the mean to 1.82
one more decimal place than the
original data point 1.813 13
 What is a Range?
• The range indicates
how similar (or dis-
Largest
similar) the measurement:
measurements are in 1.80

a subgroup Smallest
measurement:
• To calculate the 1.75

range Range:
1.80-1.75=0.05
– Subtract the smallest
measurement from the
largest measurement

14
 Exercise #2
• Collect subgroups of data
• Calculate mean and range

15
 Overall Mean X

1 2 3 4 5 6 7 8 9 10

Number of means:10 Sum:2.73+2.71+2.72+2.72+2.72+2.72+2.70.2.72+2.71+2.71 27.16

= 2.716
Sum=27.16 10
Divide sum by number of means 16
 Overall Range R
• Number of ranges: 10
• Sum of ranges:
o 0.09+0.05+0.05+0.06+0.12+0.08+0.08+0.06+0.07+0.04
o Sum=0.7
o Divide Sum by number of ranges 0.7 = 0.07
10

17
 Sections of a Control Chart: Plot of
Means and Ranges

Plot of
X
Means
Plot of
Ranges R

•The plots show

how the process is
behaving over time
•We expect the
points to fall above
and below the
center line which is 18
the overall mean
 Sections of the Control Chart:
Control Limits
• Control limits are calculated for means
and ranges
• Control limits represent the boundaries
between normal and abnormal
variation or common cause from
special cause variation
• Common cause variation is:
– What we expect to happen the majority
of time.
– Common cause variation is everything
between the limits. You can also call it
50/50 variation. When you flip a coin,
there is a 50% chance of getting a
head and a 50% chance of getting a
tail. Meaning, the only thing driving the
outcome is chance. Same with
production. If only common cause
variation is present, there is a 50%
chance of being above the target and a
50% chance of being below the target.
The majority of points should fall within
the limits.

19
 Exercise #3
• Collect more subgroups and calculate
control limits

20
 Interpreting Charts
• There are different pictures you might see
in the plots of means and ranges.
• Key point: Look for abnormal patterns in
the data. Something is causing the
abnormal pattern. This “something” is
called a assignable cause.

21
 Interpreting Control Charts and
Taking Action
• The averages are
Process In Control with Chance Variation randomly falling above
and below the centerline.
15 • There are no points
outside the upper control
10 limit.
• The variation is common
X 5 cause variation. No
special causes of
0 variation are present

22
 Trends
• The plot of averages was
Trends behaving randomly but
something occurred to
make the process start
drifting upward.
1500 • The process is no longer
behaving randomly.
1000 Special cause variation is
present
• Find the assignable
X 500 cause
• Document your actions
0 on the control chart

23
 Jumps in Process Level
• The process is not
Jumps in Process Level exhibiting random
behavior
1500 • Special cause
variation exists
1000 • Find the assignable
cause
500 • Document your
actions on the control
0 chart

24
 Cyclic Pattern
• There is a repeating
Recurring Cycles cycle to the data
• This is not random
600 behavior
400 • Find the assignable
cause
200 • Document your
0 actions on the control
chart

25
 Point Near the Control Limit

• Point at the upper control

1500 limit but not outside the
upper control limit
• Proper action to take:
1000 – Pull another sample and
plot the average and range.
If the average is still near
500 the upper limit, action may
be needed
– Document your actions on
0 the control chart

26
Point Well Outside the Upper Limit
• This is a strong signal
Process In Control with Chance Variation
that an assignable
cause exists for this
special cause
1500 variation
1000 • Find the assignable
cause
500 • Document your
actions on the control
0 chart

27
 Point Just Above or Just Below
Control Limit
• Don’t take the limit so literally.
Remember, there is a small
Process In Control with Chance Variation probability of a point falling
outside the limit. We can
expect this to happen less than
1% of the time.
1500 • Proper action to take:
– Don’t be so quick to adjust the
1000 machine or process
– Pull another sample and plot
the average and range. If the
500 average is still near the upper
limit, action may be needed
– Document your action on the
0 control chart

28
Taking and Documenting Action
• When special cause
variation is present,
find and eliminate the
assignable cause
• Document the actions
taken on the control
chart. Record the
date and time for the
action

29
 Exercise #4
• Collect more subgroups and evaluate
chart
– Change in process level
– OOC point

30