Process Control Charts

Any process has a certain amount of natural

variability. But how can we tell if the process’s
variability has gone “out of control”?
Example: An automated process whose intent is to
fill a bag with 200 pounds of cement.

Process Control is a technique for inferring that an

unplanned change has taken place in a process
measured by a process variable X.
Example: X is the exact weight of a bag of cement
intended to weigh 200 pounds.
1
 Alternative Meanings for
the Process Variable X
 The salt content, thickness, or crispness of a
bag of potato chips.
 The number of chocolate chips in a
container of chocolate-chip ice cream.
 The diameter of a bearing, or the center of a
gear.
 The waiting time at a fast-food restaurant or
at an airport check-in counter.
 The internal temperature of a “rare” steak
when it leaves a restaurant’s kitchen.
2
 Sampling
Over some period of time, take N samples with each
sample having n observations.
Example: During each of N=10 consecutive hours,
remove n=4 bags of cements from the filling process and
weigh them.
OBSERVATIONS
SAMPLE 1 199.98 200.37 200.94 200.80
SAMPLE 2 200.42 201.04 199.91 199.80
SAMPLE 3 199.59 200.08 199.04 198.47
SAMPLE 4 200.44 201.34 199.39 200.09
SAMPLE 5 199.80 199.37 200.41 196.63
SAMPLE 6 199.68 198.52 201.73 198.99
SAMPLE 7 199.83 201.68 198.53 200.33
SAMPLE 8 197.65 199.67 200.04 199.52
SAMPLE 9 199.11 200.75 200.86 199.76
SAMPLE 10 199.65 198.98 201.33 199.65
3
 Two Ways a Process Can be
Out-of-Control
Both of the processes below are out-of-control.
But in different ways! Can you see the difference?

OBSERVATIONS OBSERVATIONS
SAMPLE 1 20 10 30 SAMPLE 1 20 10 30
SAMPLE 2 40 30 20 SAMPLE 2 31 20 9
SAMPLE 3 40 50 30 SAMPLE 3 8 32 20
SAMPLE 4 50 40 60 SAMPLE 4 33 20 7
4
Two Ways to be Out-of-
Control (continued)

5
Establishing the
Control Chart’s UCL & LCL

Go to Excel Workbook

6
 Establishing the
Control Chart’s UCL & LCL (continued)

X-bar Control Chart Range Control Chart

220 40

Sample Range
Sample Mean

210 30
UCL UCL
200 X-bar-bar 20 R-bar
LCL LCL
190 10

180 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Sample Sample

7
 The Mean is out-of-control!
Sample Sample Data X-bar Range
31 203 198 191 212 201.000 21
32 205 188 207 197 199.250 19
33 199 199 205 197 200.000 8
34 211 200 208 202 205.250 11
35 197 194 203 199 198.250 9
36 187 200 193 205 196.250 18
37 195 214 216 193 204.500 23
38 218 207 223 205 213.250 18
39 199 193 208 195 198.750 15
40 208 201 201 195 201.250 13

X-bar Control Chart Range Control Chart

Out-of-control
215 40
210
Sample Range
30 UCL
Sample Mean

205 UCL
200 X-bar-bar R-bar
20
195 LCL LCL
190 Sample Mean 10 Range
185
0
180
31 32 33 34 35 36 37 38 39 40
31 32 33 34 35 36 37 38 39 40
Sample Sample
8
 The Range is out-of-control!
Sample Sample Data X-bar Range
31 190 199 198 199 196.500 9
32 224 207 195 192 204.500 32
33 186 199 199 209 198.250 23
34 211 204 194 202 202.750 17
35 217 200 188 200 201.250 29
36 204 202 184 195 196.250 20
37 193 200 201 205 199.750 12
38 211 208 212 173 201.000 39
39 205 205 202 211 205.750 9
40 188 198 178 207 192.750 29

X-bar Control Chart Range Control Chart

Out-of-control
215 40
210
Sample Range
30 UCL
Sample Mean

205 UCL
200 X-bar-bar R-bar
20
195 LCL LCL
190 Sample Mean 10 Range
185
0
180
31 32 33 34 35 36 37 38 39 40
31 32 33 34 35 36 37 38 39 40
Sample Sample
9
 Patterns to Investigate
Case #1
Why might this process be out-of-control?

Case #1

350
300
Sample Data

Upper
250
200 Middle
150 Lower
100
Data
50
0
31 32 33 34 35 36 37 38 39 40

Sample

10
 Patterns to Investigate
Case #2
Why might this process be out-of-control?

Case #2

350
300
Sample Data

Upper
250
200 Middle
150 Lower
100
Data
50
0
31 32 33 34 35 36 37 38 39 40

Sample

11
 Patterns to Investigate
Case #3
Why might this process be out-of-control?

Case #3

350
300
Sample Data

Upper
250
200 Middle
150 Lower
100
Data
50
0
31 32 33 34 35 36 37 38 39 40

Sample

12
 Patterns to Investigate
Case #4
Why might this process be out-of-control?

Case #4

350
300
Sample Data

Upper
250
200 Middle
150 Lower
100
Data
50
0
31 32 33 34 35 36 37 38 39 40

Sample

13
 The Process Control “Cycle”
Initialization. Take an initial set of N samples with n
observations, and use these to compute the initial lower and
upper control limits.

Step 1. Continue with periodic samples until the process goes

out-of-control. Look for an assignable cause.

Step 2. If possible, improve the process in a manner that

decreases the chance that the same assignable cause will
reoccur.

Step 3. After a process improvement, recalibrate the lower

and upper control limits by taking another set of N samples
with n observations. Return to Step 1.

14
 Another Type of Control Chart
We have discussed control charts in the context of a process
whose performance is measured by a continuous variable X.
For some processes, performance is measured by an binary
attribute – an attribute that is either present or not present.
Examples:
A product is either defective or non-defective.
A invoice either contains an error or is error-free.
A customer is either satisfied or unsatisfied.
To control a process measured by an binary attribute, you
need to use another type of control chart known as a p-chart.

15