MS 3322 LT 6

Statistical Quality Control I

X-bar chart and R chart

Dr. William Chung

City University of Hong Kong

1 1
 E1

Outline of Session

 Understanding Process
Variation
 Types of Process Variation
 How to “control” process
variation?

2
New product
Original weight of line of Cereal
a box Corn
Flakes is around 12oz. Due to
the size of family become
smaller, top management
decide to develop another
new product line (smaller box
around 10.7 oz. on average,
+/- 0.5 oz. acceptable
deviation). A new filling
machine is bought and the old
one is getting retired. In order
to make the new product line
start, there was a trial run
and we collect the following
3
statistics (weight of 75
 Cereal Problem
Time Obs 1 Obs 2 Obs 3 Obs 4 Obs 5
1 10.682 10.689 10.776 10.798 10.714
2 10.787 10.86 10.601 10.746 10.779
3 10.78 10.667 10.838 10.785 10.723
4 10.591 10.727 10.812 10.775 10.73
5 10.693 10.708 10.79 10.758 10.671
6 10.749 10.714 10.738 10.719 10.606 average = 10.728
7 10.791 10.713 10.689 10.877 10.603
8 10.744 10.779 10.11 10.737 10.75 SD = 0.1024
9 10.769 10.773 10.641 10.644 10.725
10 10.718 10.671 10.708 10.85 10.712
11 10.787 10.821 10.764 10.658 10.708
12 10.622 10.802 10.818 10.872 10.727
13 10.657 10.822 10.893 10.544 10.75
14 10.806 10.749 10.859 10.801 10.701 27
15 10.66 10.681 10.644 10.747 10.728
Process Variations
 We find that there are variations
of the weights.
 Standard deviation = 0.1024
If we assume the weight of cereal is normally distributed. That is,
P( - 1 < X <  + 1) = .6826
P( - 2 < X <  + 2) = .9544
P( - 3 < X <  + 3) = .9973  (10.4208, 11.0352)
where X is the weight of any cereal box produced by the new
machine.
 = 10.728g;  = 0.1024g

The acceptable range is (10.2, 11.2)

5
Why there are variations
We find that there are
variations of the weights.
 P( - 3 < X <  + 3) = .9973
  (10.4208, 11.0352)
 Can we remove these
variations?
 Can we reduce these
variations?
6
E
Funnel Experiment

Marble

Funnel
Apparatus Target paper
with bull’s eye

7
Deming Funnel Experiment

http://www.symphonytech.com/dfunnel.htm

8
Which rule you will select to reduce or remove the deviations?

9
 S

Two Basic Forms of Variation

 ____________________________is caused by factors
that can be clearly and possibly. (e.g., workers
not being equally trained.)
 ____________________________is inherent in the
service process. (i.e., random variation, caused
by type of equipment or service process design
e.g. service time of an operator)
 Which one we can be removed?

11
 E
Cereal Problem -
revisit

How do we know if these variations

If we assume the weight of cereal is (10.4208, 11.0352) only consists of
normally distributed. That is, “common” variation?
P( - 1 < X <  + 1) = .6826
P( - 2 < X <  + 2) = .9544
P( - 3 < X <  + 3) = .9973 How do we know if these variations
where X is the weight of any cereal
box produced by the new machine.
consists of both “common” and
 = 10.728g;  = 0.1024g ”assignable”?
variations = (10.4208, 11.0352)

14
That is, how can we
detect if there are
“assignable” variations in
the existing process?

15
 S

Cereal Problem – take samples

Time Obs 1 Obs 2 Obs 3 Obs 4 Obs 5
1 10.682 10.689 10.776 10.798 10.714 Sample no. 1
2 10.787 10.86 10.601 10.746 10.779
3 10.78 10.667 10.838 10.785 10.723what is the sample size =
4 10.591 10.727 10.812 10.775 10.73
5 10.693 10.708 10.79 10.758 10.671how many samples =
6 10.749 10.714 10.738 10.719 10.606
7 10.791 10.713 10.689 10.877 10.603
8 10.744 10.779 10.11 10.737 10.75
9 10.769 10.773 10.641 10.644 10.725
10 10.718 10.671 10.708 10.85 10.712
11 10.787 10.821 10.764 10.658 10.708
12 10.622 10.802 10.818 10.872 10.727
13 10.657 10.822 10.893 10.544 10.75
14 10.806 10.749 10.859 10.801 10.701
15 10.66 10.681 10.644 10.747 10.728
16
How to create control charts (X & R
charts) for the continuous
measurement, like weight, length,
volume, etc.

m = 10.728g;  = 0.1024g
x Chart Control Limits
 n=5
UCL = x  3
n
UCL = 10.866

LCL = x  3 LCL = 10.591
n

18
 How to create control
charts:
Time Obs 1 Obs 2 Obs 3 Obs 4 Obs 5 Avg Sample Average
1 10.682 10.689 10.776 10.798 10.714 10.732
2 10.787 10.86 10.601 10.746 10.779 10.755
3 10.78 10.667 10.838 10.785 10.723 10.759
4
5
10.591
x
10.693 10.727 10.812
x 10.682  10.689

10.708 10.79
10.775
...
10.758
 10 .
10.73
714
10.671
10.727
10.724
6 10.749 n 10.714 10.738 510.719 10.606 10.705
7 10.791 10.713 10.689 10.877 10.603 10.735
8 10.744 10.779 10.11 10.737 10.75 10.624
9 10.769 10.773 10.641 10.644 10.725 10.710
10 10.718 10.671 10.708 10.85 10.712 10.732
11 10.787 10.821 10.764 10.658 10.708 10.748
12 10.622 10.802 10.818 10.872 10.727 10.768
13 10.657 10.822 10.893 10.544 10.75 10.733
14 10.806 10.749 10.859 10.801 10.701 10.783

x
15

10.66
x 10.681 910.644

10.747
.110  10.755  ...  1010.728
.692 10.692

Number of subgroups 15 Averages 10.728 𝝁 19

 How to create control charts:
X-bar chart

the 2nd sample average = 10.755

10.900
UCL
10.850 =10.866

10.800

10.750
Means

Average
=10.728
10.700

10.650

10.600 LCL
=10.591
10.550
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sample

20
Why we need
to create R
chart?
How to create
it?

21
 How to create R chart:
Time Obs 1 Obs 2 Obs 3 Obs 4 Obs 5 Avg Range
1 10.682 10.689 10.776 10.798 10.714 10.732 0.116
2 10.787 10.86 10.601 10.746 10.779 10.755 0.259
3 10.78 10.667 10.838 10.785 10.723 10.759 0.171
4 10.591 10.727 10.812 10.775 10.73 10.727 0.221
5 10.693 10.708 10.79 10.758 10.671 10.724 0.119
6 10.749 10.714 10.738 10.719 10.606 10.705 0.143
7 10.791 10.713 10.689 10.877 10.603 10.735 0.274
8 10.744 10.779
R = xmax10.11
– xmin 10.737 10.75 10.624 0.669
9 10.769 10.773 10.641
=10.798-10.682 10.644 10.725 10.710 0.132
10 10.718 10.671 10.708 10.85 10.712 10.732 0.179
11
12
10.787
10.622
10.821 10.764
 R10.658 10.708
 10.818 10.872 10.727
R10.802
10.748
10.768
0.163
0.250
13 10.657 Number
10.822 10.893 of subgroups
10.544 10.75 10.733 0.349
14 10.806 10.749 10.859 10.801 10.701 10.783 0.158
15 10.66 10.681 10.644 10.747 10.728 10.692 0.103

Averages 10.728 0.220400

How to create R charts:
Determine Control Limit Formulas and Necessary Tabled Values

x Chart Control Limits Figure 12-8 (for A2 & D4)


UCL = x  3  x  A2 R n A2 D3 D4
n 2 1.88 0 3.27
 3 1.02 0 2.57
LCL = x  3  x  A2 R
n 4 0.73 0 2.28
5 0.58 0 2.11
R Chart Control Limits 6 0.48 0 2.00
UCL = D 4 R 7 0.42 0.08 1.92
8 0.37 0.14 1.86
LCL = D 3 R 9 0.34 0.18 1.82
10 0.31 0.22 1.78
11 0.29 0.26 1.74
Create control charts, x-bar, by
Fig.12-8

10.900

10.850 UCL

10.800

10.750
Means

Average
10.700

10.650

10.600
LCL
10.550
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sample
24
How to create control charts: R-bar
UCL = D 4 R ( 2.11)( 0.2204 ) 0.46504
LCL = D 3 R ( 0)( 0.2204 ) 0

0 .8 0 0

0 .7 0 0

0 .6 0 0

0 .5 0 0
UCL
R 0 .4 0 0

0 .3 0 0
Average
0 .2 0 0

0 .1 0 0

0 .0 0 0
LCL
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
S a mp le
25
Use of x-bar and R charts

26
 Is this process
under
control?

Is this process
under
control?
 Interpreting Control
Charts

 When a point is found to be outside of

the control limits, we call this an “out
of control” situation.
 When a process is out of control, the
variation is probably no longer
random.
 The figures in the next two slides
show different signals for concern
regarding “out of control” that are
sent by a control chart.
28
 Rule #1
Control Chart Evidence for Investigation

Rule #2

Rule #3a 29
Control Chart Evidence for Investigation
.

Rule #5
Rule #3b
Rule #4

Rule #6

30
 S
Interpreting Control
Charts
 When a process is out of control, the
variation is probably no longer random.
 It implies that there is “assignable”
variations in the process.
 What we will do?

31
E
Interpreting Control
Charts
 When a process is out of control, the variation is
probably no longer random.
 It implies that there is “assignable” variations in
the process.
 What we will do?

(1)find the factors causing the assignable

variations, e.g., workmanship problem,
after we fix it we can start to use the
same chart for quality control.
(2)when do we need to develop a new
control chart after removing the factors
of the assignable variations to the 32
current process?
S
Statistical Process Control Steps
to create control charts for new
processes

Produce Good
Start Provide Service No

Assignable.
Take Samples variations?
Yes
Inspect Samples Stop Process

Create
Find Out Why
Control Chart(s)

33
How do we know the
variations come from “normal
distribution”?

If not, what we can do?

35
 The Relationship Between
Population and Sampling
Distributions
Population distributions
Distribution of sample means
Beta
Mean of sample means x

Normal Standard deviation of x

x 
the sample n
means

Uniform

 3 x  2  x  1 x x   x  2  x  3 x
(mean)
95.5% of all x fall within 2σ
x
99.7% of all x fall within 3 x
Central Limit Theorem

If we plot all the average weight (sample

mean) to form a histogram, what it
would look like?

37
10.7 ±0.05
g/box

38
 Sampling Distribution of Means,
E and Process Distribution
Distribution of sample
means with /sqrt(n)

Population
distribution with 

x m
( mean )
39
Conclusion
 Statistical technique used to
ensure process is making
product to standard
 Objective: Identify assignable
causes
 Uses process control charts

40