Advanced Tecniques SPC

Advanced Statistical Quality
Control and Monitoring

Prof. Fugee Tsung
Hong Kong University
of Science and Technology
Quality Lab: http://qlab.ielm.ust.hk
Outline
Module A: Overview of SPC; Autocorrelated

SPC
Module B: Multivariate SPC; Profile
Monitoring
Module C: Integrating EPC and SPC; Run-to
Run Control
Module D: Multistage SPC; Multi-Charts
Module E: Categorical SPC; Nonparametric
SPC
Dr. Fugee Tsung, HKUST
How to Handle Variation

Ignore
Desensitize
Compensate
Reduce/
Eliminate
t
s
u
ob
n
g
i
s
De
C
P
E
S PC
Statistical Process Control (SPC):

Role in Six Sigma
Define
Control
Measure
Improve
Analyze
SPC is emphasized as the major tool used to

control and sustain a Six Sigma process
SPC: Basic Idea

yi ~ N ( 1 , 12 ) for i = 1,2,..., ;
yi ~ N ( 2 , ) for i = + 1,..., n.
Sample Mean
2
2
1.5
UCL=1.342
1.0
0.5
0.0
Mean=0
-0.5
-1.0
LCL=-1.342
-1.5
0
10
20
30
Sample Number
In most SPC applications it is assumed that the quality of a process or product

can be adequately represented by the distribution of a univariate quality
characteristic or by the general multivariate distribution of a vector consisting of
several quality characteristics (Woodall et al. 2004).
SPC for Profiles
SPC: A Different Idea

yi = f ( xi )
1.5
UCL=1.342
Sample Mean
1.0
0.5
0.0
Mean=0
-0.5
-1.0
LCL=-1.342
-1.5
0
10
20
30
Sample Number
In many practical situations, however, the quality of a process or product is

better characterized and summarized by a relationship between a response
variable and one or more explanatory variables (Woodall et al. 2004).
A Profile Example
Another Profile Example
SPC for engineering controlled

processes
A Engineering Controlled Process

1.5
UCL=1.342
Sample Mean
1.0
Process
0.5
0.0
Mean=0
-0.5
-1.0
EPC
Controller
LCL=-1.342
-1.5
0
xi = k P yi + k I yi + k D yi
10
20
30
Sample Number
j =1
kI
xi = k P yi +
yi + k D (1 B ) yi
1 B
11
Example: CMP Run-to-Run Control
Drift in removal rate during

processing
Variation in incoming thickness
Head-to-head variation in
processing
Unmeasurable tool
characteristics affect product
output
Incomplete product
measurements
Significant measurement delay
3
4
2
1
5
dt
xt
CMP
Process
EPC
Controller
yt
yt
12
SPC for Multistage processes
A Multistage Process
Stage 1
Stage k-1
Stage k
Stage N
1.5
UCL=1.342
yk-1
Sample Mean
1.0
yk
0.5
0.0
Mean=0
-0.5
-1.0
LCL=-1.342
-1.5
yk = Ak 1 yk 1 + Bk k + wk
10
20
30
Sample Number
Yk = Ck yk + vk
vk ~ N (0, v2 ), wk ~ N (0, w2 )
14
Example: Printed Circuit Board Manufacturing
Exposure
Black Oxide
Lay-up
From Ding, Shi, and Ceglarek (2002)
Hot Press
Cutting
Drilling
Inspection
.
15
Module A: Overview of Statistical

Process Control (SPC)/ Statistical
Quality Control (SQC)
Based on Montgomery, Introduction to

Statistical Quality Control, 4th edition, John
Wiley, 2001.
Outline
I.
What is SPC
II.
Common and special causes of quality

variation
III.
Statistical basis of control charts
IV.
Introduction of popular control charts

17
I. What is SPC
Statistical process control (SPC)
Control charts, plus other problem-solving

tools
Useful in monitoring processes, reducing
variability through elimination of special
causes
On-line technique
19
The Tools of SPC
The primary tool of SPC is the Control Chart

The Control Chart looks similar to a run chart, however the
addition of control limits plays an important role.
20
II. Common and special causes of

quality variation
Purpose for Using Control Charts
The Control Limits probabilistically separate

common cause and special cause variability.
22
Common and Special Causes of Quality

Variation
A process that is operating with only common

cause/chance causes of variation present is said to
be in statistical control.
A process that is operating in the presence of
special/assignable causes is said to be out of
control.
The eventual goal of SPC is reduction or
elimination of variability in the process by
identification of assignable causes.
23
Common and Special Causes of Quality

Variation
24
III. Statistical basis of control charts
Statistical Basis of the Control Chart

Basic Principles
A typical control chart has control limits set at

values such that if the process is in control,
nearly all points will lie between the upper
control limit (UCL) and the lower control limit
(LCL).
26

Basic Principles
27
The Out-of-Control Signal
The control limits define a probabilistic

level of occurrence of an extreme reading.
28

Out-of-Control Situations
If at least one point plots beyond the

control limits, the process is out of control
If the points behave in a systematic or
nonrandom manner, then the process could
be out of control.
29
Hypothesis Testing & Control Charting
30

Relationship between hypothesis testing and
control charts
We have a process that we assume the true
process mean is = 74 and the process standard
deviation is = 0.01. Samples of size 5 are taken
giving a standard deviation of the sample
average, x , as
0.01
x =
=
= 0.0045
n
5
31

control charts
Control limits can be set at 3 standard

deviations from the mean.
This results in 3-Sigma Control Limits
UCL = 74 + 3(0.0045) = 74.0135
CL= 74
LCL = 74 - 3(0.0045) = 73.9865
32

control charts
Choosing the control limits is equivalent to

setting up the critical region for testing
hypothesis
H0: = 75
H1: 75
33

Relationship between the process and the control chart
34

Important uses of the control chart
Most processes do not operate in a state of statistical

control.
Consequently, the routine and attentive use of control
charts will identify assignable causes. If these causes can
be eliminated from the process, variability will be
reduced and the process will be improved.
The control chart only detects assignable causes.
Management, operator, and engineering action will be
necessary to eliminate the assignable causes.
Out-of-control action plans (OCAPs) are an important
aspect of successful control chart usage.
35

Types the control chart
Variables Control Charts
These charts are applied to data that follow a

continuous distribution (measurement data).
Attributes Control Charts
These charts are applied to data that follow a

discrete distribution.
36

Type of Process Variability
Stationary behavior, uncorrelated data
Stationary behavior, autocorrelated data
Nonstationary behavior
37

Type of Variability
Most conventional control charts are
effective when the in-control process data
is stationary and uncorrelated.
38

Popularity of control charts
1) Control charts are a proven technique for
improving productivity.
2) Control charts are effective in defect prevention.
3) Control charts prevent unnecessary process
adjustment.
4) Control charts provide diagnostic information.
5) Control charts provide information about process
capability.
39
Choice of Control Limits

General model of a control chart
UCL = W + L W
Center Line = W
LCL = W L W
where L = distance of the control limit from the
center line
W = mean of the sample statistic, w.
W = standard deviation of the statistic, w.
40

99.7% of the Data
If approximately 99.7% of the data lies within 3

of the mean (i.e., 99.7% of the data should lie
within the control limits), then 1 - 0.997 = 0.003
or 0.3% of the data can fall outside 3 (or 0.3%
of the data lies outside the control limits).
(Actually, we should use the more exact value
0.0027)
0.0027 is the probability of a Type I error or a
false alarm in this situation.
41

Three-Sigma Limits
The use of 3-sigma limits generally gives good

results in practice.
If the distribution of the quality characteristic is
reasonably well approximated by the normal
distribution, then the use of 3-sigma limits is
applicable.
These limits are often referred to as action limits.
42
Sample Size and Sampling Frequency
In designing a control chart, both the

sample size to be selected and the
frequency of selection must be specified.
Larger samples make it easier to detect
small shifts in the process.
Current practice tends to favor smaller,
more frequent samples.
43

Average Run Length
The average run length (ARL) is a very
important way of determining the appropriate
sample size and sampling frequency.
Let p = probability that any point exceeds the
control limits. Then,
1
ARL =
p
44

Illustration
Consider a problem with control limits set

at 3standard deviations from the mean.
The probability that a point plots beyond
the control limits is again, 0.0027 (i.e., p =
0.0027). Then the average run length is
1
ARL =
= 370
0.0027
45

What does the ARL tell us?
The average run length gives us the length of
time (or number of samples) that should plot in
control before a point plots outside the control
limits.
For our problem, even if the process remains in
control, an out-of-control signal will be
generated every 370 samples, on average.
46

Average Time to Signal
Sometimes it is more appropriate to

express the performance of the control
chart in terms of the average time to signal
(ATS). Say that samples are taken at fixed
intervals, h hours apart.
ATS = ARL ( h )
47
Rational Subgroups
Subgroups or samples should be selected

so that if assignable causes are present, the
chance for differences between subgroups
will be maximized, while the chance for
differences due to these assignable causes
within a subgroup will be minimized.
48
Analysis of Patterns on Control Charts

Nonrandom patterns can indicate out-of-control
conditions
Patterns such as cycles, trends, are often of considerable

diagnostic value.
Look for runs - this is a sequence of observations of the
same type (all above the center line, or all below the
center line)
Runs of say 8 observations or more could indicate an outof-control situation.
Run up: a series of observations are increasing
Run down: a series of observations are decreasing
49
Analysis of Patterns on Control Charts

Western Electric Handbook Rules (Should be used
carefully because of the increased risk of false alarms)
A process is considered out of control if any of the
following occur:
1) One point plots outside the 3-sigma control limits.
2) Two out of three consecutive points plot beyond the 2sigma warning limits.
3) Four out of five consecutive points plot at a distance of 1sigma or beyond from the center line.
4) Eight consecutive points plot on one side of the center line.
50
IV. Introduction of popular control

charts
1. The Shewhart chart (X-bar & R

chart)
Notation for variables control charts

n - size of the sample (sometimes called a
subgroup) chosen at a point in time
m - number of samples selected
x i = average of the observations in the ith sample
(where i = 1, 2, ..., m)
x = grand average or average of the averages
(this value is used as the center line of the control
chart)
53
Notation and values

Ri = range of the values in the ith sample
Ri = xmax - xmin
R = average range for all m samples
is the true process mean
is the true process standard deviation
54
Statistical Basis of the Charts

Assume the quality characteristic of interest is normally
distributed with mean , and standard deviation, .
If x1, x2, , xn is a sample of size n, then the average of
this sample is
x1 + x 2 + L + x n
x=
n
x is normally distributed with mean, , and standard

deviation,
x =/ n
55
Statistical Basis of the Charts

The probability is 1 - that any sample mean will fall
between
+ Z / 2 x = + Z / 2
n
and
Z / 2 x = Z / 2
n
The above can be used as upper and lower control limits
on a control chart for sample means, if the process
parameters are known.
56
The Operating Characteristic Function
How well the x and R charts can detect process

shifts is described by operating characteristic (OC)
curves.
Consider a process whose mean has shifted from
an in-control value by k standard deviations. If
the next sample after the shift plots in-control,
then you will not detect the shift in the mean.
The probability of this occurring is called the risk.
57
The Operating Characteristic Function
The probability of not detecting a shift in the

process mean on the first sample is
= ( L k n ) ( L k n )
L= multiple of standard error in the control limits
k = shift in process mean (#of standard deviations).
58
59
2. The Cumulative Sum (CUSUM)

Charts
Basic Principles: The Cusum Control Chart for

Monitoring the Process Mean
The cusum chart incorporates all information in the
sequence of sample values by plotting the cumulative
sums of the deviations of the sample values from a target
value.
If 0 is the target for the process mean, x jis the average of
the jth sample, then the cumulative sum control chart is
formed by plotting the quantity
i
Ci = ( x j 0 )
j=1
61
The Tabular or Algorithmic Cusum for

Monitoring the Process Mean
Let xi be the ith observation on the process

If the process is in control then x i ~ N( 0 , )
Assume is known or can be estimated.
Accumulate derivations from the target 0 above the
target with one statistic, C+
Accumulate derivations from the target 0 below the
target with another statistic, CC+ and C- are one-sided upper and lower cusums,
respectively.
62
The statistics are computed as follows:

The Tabular Cusum
[
= max[0, (
Ci+ = max 0, x i ( 0 + k ) + Ci+1

Ci
k
)
x
+
C
0
i
i 1
]
]
starting values are C 0+ = C 0 = 0

K is the reference value (or allowance or slack value)
If either statistic exceed a decision interval H, the process
is considered to be out of control. Often taken as a H = 5
63
Selecting the reference value, K

K is often chosen halfway between the target 0 and the
out-of-control value of the mean 1 that we are interested
in detecting quickly.
Shift is expressed in standard deviation units as 1=
0+, then K is
1 0
K= =
2
2
64
3. The Exponentially Weighted

Moving Average (EWMA) Charts
3. The EWMA chart
66
67
68
69
Other Charts
Cuscore Charts (Box and Ramirez, 1992)

Reference-Free Cuscore Charts (Han and
Tsung, JASA, 2006)
General EWMA Charts (Han and Tsung,
Annals of Stat, 2004)
GLRT Charts
70
Related references
Han, D., and Tsung, F.(2004), " A Generalized EWMA Control Chart and its
Comparison with the Optimal EWMA, CUSUM and GLR Schemes," The Annals of
Statistics, 32.
Han, D. and Tsung, F. (2005), " Comparison of the Cuscore, GLRT and CUSUM
Control Charts for Detecting a Dynamic Mean Change," The Annals of the Institute of
Statistical Mathematics, 57.
Han, D. and Tsung, F.(2006) " A Reference-Free Cuscore Chart for Dynamic Mean
Change Detection and a Unified Framework for Charting Performance Comparison ,"
Journal of the American Statistical Association, 101, 368-386.
Ou, Y. J., Wu, Z. and Tsung, F. (2012), " Comparison Study of Effectiveness and
Robustness of Control Charts for Monitoring Process Mean," International Journal of
Production Economics, 135, 479-490.
71

Advanced Tecniques SPC

Uploaded by

Copyright:

Available Formats

Advanced Tecniques SPC

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced Tecniques SPC

Uploaded by

Copyright:

Available Formats

Advanced Statistical Quality

Control and Monitoring

Module A: Overview of SPC; Autocorrelated

How to Handle Variation

Dr. Fugee Tsung, HKUST

Statistical Process Control (SPC):

SPC is emphasized as the major tool used to

SPC: Basic Idea

In most SPC applications it is assumed that the quality of a process or product

SPC for Profiles

SPC: A Different Idea

In many practical situations, however, the quality of a process or product is

Dr. Fugee Tsung, HKUST

Another Profile Example

Dr. Fugee Tsung, HKUST

SPC for engineering controlled

A Engineering Controlled Process

Example: CMP Run-to-Run Control

Drift in removal rate during

Dr. Fugee Tsung, HKUST

SPC for Multistage processes

Example: Printed Circuit Board Manufacturing

From Ding, Shi, and Ceglarek (2002)

Module A: Overview of Statistical

Based on Montgomery, Introduction to

Common and special causes of quality

Statistical basis of control charts

Introduction of popular control charts

Statistical process control (SPC)

Control charts, plus other problem-solving

Dr. Fugee Tsung, HKUST

The Tools of SPC

The primary tool of SPC is the Control Chart

Dr. Fugee Tsung, HKUST

II. Common and special causes of

Purpose for Using Control Charts

The Control Limits probabilistically separate

Common and Special Causes of Quality

A process that is operating with only common

Common and Special Causes of Quality

Dr. Fugee Tsung, HKUST

III. Statistical basis of control charts

Statistical Basis of the Control Chart

A typical control chart has control limits set at

Dr. Fugee Tsung, HKUST

Statistical Basis of the Control Chart

Dr. Fugee Tsung, HKUST

The Out-of-Control Signal

The control limits define a probabilistic

Dr. Fugee Tsung, HKUST

Statistical Basis of the Control Chart

If at least one point plots beyond the

Dr. Fugee Tsung, HKUST

Hypothesis Testing & Control Charting

Dr. Fugee Tsung, HKUST

Statistical Basis of the Control Chart

Statistical Basis of the Control Chart

Control limits can be set at 3 standard

Statistical Basis of the Control Chart