CHAPTER 6

CONTROL CHARTS
FOR VARIABLES
CC01 - GROUP 05

Quality Management and Control Course

 GROUP 05 - MEMBERS

ĐINH NHƯ THUẦN

2153847
HỒ THY HÀ TRANG
2153906
LÊ NGỌC THUẬN
2153851
NGUYỄN HOÀNG ANH THY 2153880
NGUYỄN TUẤN ĐẠT
2
 6.1
INTRODUCTION

3
 What is variable ?

● Many quality characteristics can be expressed in terms of a numerical measurement.

● A single measurable quality characteristic, such as a dimension, weight, volume…
is called a variable.

Why Control charts for variables are important ?

● Control charts are one of the primary tools used in the Analyze and Control steps of
DMAIC (define, measure, analyze, improve, and control).
● When dealing with a quality characteristic that is a variable, it is usually necessary to
monitor both the mean value and the variability of the quality characteristic.

4
 Simple definitions
● The x control chart: Control chart for means (use to control the process average or
mean quality level).
● The s control chart: Control chart for the standard deviation (use to control the
process variability).
● The R control chart: Control chart for the range.

Example

5
 6.2
CONTROL CHARTS
FOR
–x
AND R

6
6.2.1 Statistical Basis of the Charts

7
8
9
Development of Equations 6.4 and 6.5

10
Development of Equations 6.4 and 6.5

11
Application of x and R Charts

After construct x and R Charts(using 6.4 and 6.5 equation), there are many cases will happen.
Divide it to 4 main cases:
Case 1: All points plot inside the control limits
● Points within control limits and with no obvious systematic behavior
→ Process was controlled in the past, and the trial control limits are appropriate for current
and future control.

Case 2: Points plot outside the control limits (assignable cause)

● Remove points with identified causes and recalculate control limits with the remaining
points. Upon recalculation, some points that were plot inside control limits may now fall
outside due to the tighter limits. Repeat until all points are plot inside control limits.

12
Application of x and R Charts
Case 3: Some points plot outside the control limits (undefined assignable cause)

Solution 1: Eliminate those points and do not require further explanation.

Solution 2: Retain those points based on considering whether the trial control limits are
suitable for the current situation.

● If the point represents an outlier, the resulting control limits will be too wide.
However, if there are only 1 or 2 points, it will not significantly distort the graph.
● If future samples still fall within control, unexplainable points can be safely
disregarded.

Overall, if both charts show points outside control limits, immediately create an R chart. If the
R chart also shows points outside control limits, the variability of the process is unstable, and
the upper control limit on the x chart is not trustworthy.

13
Application of x and R Charts

Case 4: Many points plot outside the control limits (undefined assignable cause)

● Focus on the pattern of points plot outside control limits of the chart, a pattern will
almost always exist .
● Usually, identifying the cause related to the pattern of the point outside the control
limits is easier. Remove this process issue leads to significant improvement in the
process.

14
6.2.2 Development and Use of x and R Charts
● Example 6.1
● Example 6.2

15
Example 6.1

A hard-bake process (see Section

5.3.1) is used in conjunction with
photolithography in semiconductor
manufacturing. We wish to establish
statistical control of the flow width
of the resist in this process using x
and R charts. 25 samples, each of size
five wafers, have been taken when we
think the process is in control. The
interval of time between samples or
subgroups is one hour. The flow width
measurement data (in microns) from
these samples are shown in Table 6.1.

16
Example 6.1
Because the control limits on the x chart
depend on the process variability, unless
process variability is in control, these
limits will not have much meaning.
⇨ Begin with the R chart.
● Center line of R chart:

n = 5 observations ⇨ D3 = 0; D4 = 2.114
(Table VI)
● Control Limits for R chart:
LCL = RD3 = 0
UCL = RD4 = 0.68749 17
 Example 6.1
Since the R chart indicates that process
variability is in control, we may now
construct the x chart.
Center line of x chart:

n= 5 observations ⇨ A2 = 0.577 (Table VI).

Control Limits for x chart:
LCL = x - RA2 = 1.31795
UCL = x + RA2 = 1.69325

18
Example 6.1

19
Estimating Process Capability
Fraction of non-conforming Process capability ratio (PCR)
product Cp
- x = 1.5056 microns
- n = 5 ⇨ d2 = 2.326 (Table VI)
^
- σ = R/d2 = 0.32521/2.326 = 0.1398 microns
^ ^
- Specification Limits = 1.5 ± 0.5 microns σ is usually unknown ⇨ Use σ ⇨ Result is Cp

^
Cp ∈ (LSL, USL) ⇨ A moderately small number of
⇨ About 0.035 percent [350 parts per million nonconforming wafers will be produced.
(ppm)] of the wafers produced will be outside
of the specifications.
20
Estimating Process Capability
The PCR Cp may be interpreted another way, that is using the Percentage of the
specification band that the process uses up - P.

For the hard-bake process:

⇨ The process uses up about 84% of the specification band.

21
03 cases of interest relative to the PCR Cp and
process specifications.

PCR Cp > 1 PCR Cp = 1 PCR Cp < 1

The process uses up The process uses up all the The process uses up more
much less than 100% of tolerance band. than 100% of
the tolerance band the tolerance band
⇨ Few non-conforming ⇨ A large number of non-
units will be produced conforming units
will be produced.

22
Revision of Control Limits and Center Lines

- The effective use of any control chart will require periodic revision of the control
limits and center lines.
+ Time periods: every week, month
+ Samples: Every 20, 50 or 100 samples
- Use m ≥ 25 samples/ subgroup (recommend 200–300 individual observations).

23
Revision of Control Limits and Center Lines

When R chart is in control, if:

+ Mean can be changed by simple adjustment of a manipulatable variable in the
process ⇨ The user will replace the center line of the chart with a target value
x0, shifting the process average to desired value.
+ Mean is not influenced by a simple process adjustment ⇨ Requires complex and
unknown function of several process variables.
When R chart is out of control:
+ Eliminate out-of-control points. Re-calculate R ⇨ New center line of R chart.
+ Determine new LCL/ UCL of x chart and R chart.

24
Phase II Operation of the x and R Charts
20 additional samples of wafers from the hard-bake process were collected after the control
charts were established. The sample values of x and R plotted on the control charts
immediately after each sample was taken.

25
Phase II Operation of the x and R Charts

After established CCs and used it to monitor

the process, it’s time to detect shift in
production.
Sensitizing rules (Western Electric rules) only
work best when m < 40 samples, or it’ll
greatly increase the occurrence of false alarms.
⇨ Use Tier chart (Tolerance diagram).
It shows a particular value of x or R was
produced by one or two unusual observations
in the sample, reveal some pattern of the data.
Construct using Minitab Box plot.

26
Phase II Operation of the x and R Charts

This chart does not indicate that the

out-of-control signals were
generated by unusual individual
observations, but instead, they
probably resulted from a shift in the
mean around the time that sample
38 was taken.
μnew > μold ⇨ A much higher
percentage of nonconforming
wafers will be produced at this new
mean flow rate.

27
Control Limits (CLs), Specification Limits (SLs), and
Natural Tolerance Limits (NTLs)

There’s no connection & relationship between Control Limits and Specification Limits.

Control Limits Specification Limits

Come from x and R charts Process

External
Internal
Determined (Management,
from (Natural Tolerance
manufacturing engineers,
Limits of the process)
customers, etc.)

28
 Rational subgroup

Defining a rational subgroup in practice may be easier with a clear understanding of the
functions of the two types of control charts.

x control chart R control chart

Monitors between - sample variability Measures within - sample variability

(variability in the process over time) (the instantaneous process variability at
⇨ Select sample so that maximizes the a given time)
chances for shifts in the process average ⇨ Select sample so that variability
to occur between samples; show up as within samples measures only chance or
out-of-control points on the x chart. random causes.

29
Guidelines for the Design of the Control Chart

To design economically optimum control chart, the analyst detailed information about both
the statistical characteristics (sample size, control limit width, frequent of sampling) of the
control chart tests and the economic factors (cost of sampling, costs of investigating and
possibly correcting the process in response to out-of-control signals, costs associated with
producing a product that does not meet specifications) that affect the problem (Ch10, s10.2).

30
Guidelines for the Design of the Control Chart

R chart (s6.3):
- n small (n = 5) ⇨ R chart insensitive.
- n large (n ≥ 10-12) ⇨ Use a control chart for s or s2 instead of the R chart.

The problem of choosing the sample size and the frequency of sampling is one of allocating
sampling effort. If the interval between samples is too great, too much defective product will be
produced before another opportunity to detect the process shift occurs.
- From economic view: High cost of producing defective items ⇨ Choose samples with small
sample n, high frequent f.
- From production view: High rate of production (> 50,000 units/hour) ⇨ Choose samples with
large sample n (If per unit inspection and testing costs are not excessive), high frequent f.
31
Changing Sample Size on the x and R Charts

Sample size n is not constant between each sample m.

● Situation 1: Variable sample size on control charts
Each sample may consist a different number of observations. Normal x and R charts leads to
changing Center Line on R chart.

⇨ Use x and s chart.

● Situation 2: Permanent (or semi-permanent) change.
Because of cost/ Because the process has exhibited good stability and fewer resources are
being allocated for process monitoring.
⇨ Recompute new Control Limits from the old ones.

32
New control limits for x chart New control limits for R chart

In which,

33
Example 6.2

To illustrate the above procedure,

consider the and R charts developed for
the hard-bake process in Example 6.1.
These charts were based on a sample
size of 5 wafers. Suppose that since the
process exhibits good control, the
process engineering personnel want to
reduce the sample size to 3 wafers.
Set up the new control charts.

34
Example 6.2

From Example 6.1:

nold = 5 Rold = 0.32521

From Appendix Table VI:

d2, old = 2.326
d2, new = 1.693

35
Example 6.2 - Solution
New Control Limits on x chart:

New Control Limits on R chart:

36
Example 6.2 - Solution

The effect of
reducing the
sample size is to
increase the width
of the limits on
the x chart and to
lower the center
line and the upper
control limit on
the R chart.

37
6.2.3 Charts Based on Standard Values
● If the process mean and variance are known or can be specified, then control limits can be
developed using these values:

With: (constant that depends on n)

With:
d2: mean of distribution of relative range
d3: standard deviation of distribution of relative range

● Constants are tabulated in Appendix VI.

38
6.2.4 Interpretation of x and R Charts

Patterns of the plotted points will provide useful diagnostic information on the process, and
this information can be used to make process modifications that reduce variability.

Cyclic pattern x-bar chart: temperature, operator

fatigue, regular rotation of operators
and/or machines, etc.
R chart: maintenance schedules,
operator fatigue, or tool wear

39
6.2.4 Interpretation of x and R Charts
Mixture pattern

A mixture pattern is generated by

two (or more) overlapping
distributions generating the process
output.

- Mixtures result from “overcontrol”

- Occur when output product from several sources (such as parallel machines) is fed into a
common stream that is then sampled for process monitoring purposes.

40
6.2.4 Interpretation of x bar and R Charts
A shift in process level

Shifts may result from the introduction of new workers; changes in

methods, raw materials, or machines; a change in the inspection method
or standards, etc.

A trend
Trends are usually due to a gradual wearing out or deterioration of a tool
or some other critical process component; from human causes and
seasonal influences.
Regression control chart: monitoring and analyzing processes with
trends
Modified control chart: process exhibits tool wear 41
6.2.4 Interpretation of x bar and R Charts

Stratification pattern

One potential cause of stratification is incorrect calculation of control limits.

Result when the sampling process collects one or more

units from several different underlying distributions within each subgroup.
42
6.2.5 The Effect of Non-normality on x bar and R

● In general, the x chart is insensitive (robust) to small departures from normality.

● An assumption in performance properties is that the underlying distribution of quality
characteristics is normal.
- If underlying distribution is not normal, sampling distributions can be drive and exact
probability limits obtained.
● Usual normal theory control limits are very robust to normality assumption.
● In most cases, samples of sizes 4 or 5 are sufficient to ensure reasonable robustness to
normality assumption for x chart.
● The R chart is more sensitive to nonnormality than the x bar chart
● For 3-sigma limits, the probability of committing a type I error is 0.00461 on the R-chart.
(Recall that for x, the probability is only 0.0027).
● Assumptions of normality and independence are not a primary concern in Phase I.

43
 6.2.6. The Operating-Characteristic Function

Operating-characteristic (OC) curves: describes the ability of the x and R charts to detect
shifts in process quality

Φ: the standard normal cumulative distribution function

44
6.2.6. The Operating-Characteristic Function

45
 6.2.6. The Operating-Characteristic
Function The case of three-sigma limits (L = 3)

To construct the OC curve for the x chart

Plot the β-risk against the magnitude of the shift

(we wish to detect)
expressed in standard deviation units
for various sample sizes n

46
6.2.6. The Operating-Characteristic
Function
The probability that the shift will be detected on the rth subsequent sample is simply 1 - β times
the probability of not detecting the shift on each of the initial r - 1 samples:

In general, the expected number of samples taken before the shift is detected is simply the
average run length

Average run length: The expected number of samples taken to detect a shift

47
6.2.6. The Operating-Characteristic
Function
To construct the OC curve for the R chart
The distribution of the relative range W = R/σ
is employed.

Suppose that the in-control value of the

standard deviation is σ0. Then the OC curve
plots the probability of not detecting a shift to
a new value of σ,
σ1 > σ0 — on the first sample following the shift.

48
6.2.7. The Average Run Length for the x Chart

or

for the in-control ARL

for the out-of-control ARL

49
6.2.7. The Average Run Length for the x Chart
● Detect a shift of 1.5σ (k = 1.5) using a sample
size of n = 3, then the average number of
samples required will be ARL1 = 3.

● We could reduce the ARL1 to approximately 1 if

we increased the sample size to n = 16.
● Distribution of run length for a Shewhart
control chart is geometric and that this can be a
very skewed distribution, so the average (that is,
the ARL) may not be the best measure of a
typical run length.
● The computations for a specific control chart
are usually based on estimates of the process
parameters. This results in inflation of both
ARL0 and ARL1.
50
6.2.7. The Average Run Length for the x Chart
● The average time to signal (ATS): the number of time periods that occur until a signal
is generated on the control chart.
● If samples are taken at equally spaced intervals of time h, the ATS is

● The ARL in terms of the expected number of individual units sampled I rather
than the number of samples taken to detect a shift. If the sample size is n, the
relationship between I and ARL is

51
6.2.7. The Average Run Length for the x Chart

Detect a shift of 1.5σ, an x chart

with n = 16 will require that
approximately 16 units be
sampled, whereas if the sample
size is n = 3, only about 9 units
will be required, on the average

52
 6.3
CONTROL CHARTS
FOR
–x
AND s

53
6.3.1. Construction and Operation of x and s Charts

54
 6.3.1. Construction and Operation of x and s Charts

Consequently, the parameters of the s chart with a standard value for σ given become

55
6.3.1. Construction and Operation of x and s Charts

● If no standard is given for σ => estimated by analyzing past data.

● Suppose that: m preliminary samples are available, each of size n, and si is the
standard deviation of the ith sample.

The average of the m standard deviations is

56
6.3.1. Construction and Operation of x and s Charts

57
6.3.1. Construction and Operation of x and s Charts
❖ The parameters of the s chart:

Note that B4 = B6 /c4 ; B3 = B5 / c4

❖ When s/c4 is used to estimate σ

58
6.3.1. Construction and Operation of x and s Charts

Then the sample standard deviation is:

59
60
Example 6.3. x and s Charts for the Piston Ring Data

Construct and interpret x and

s charts using the piston ring
inside diameter
measurements in Table 6.3.

61
Example 6.3 - Solution

62
Example 6.3 - Solution

(a) The x chart with control limits (b) The s control chart
based on s
63
6.3.1. Construction and Operation of x and s Charts

Estimation of σ

● s/c4 is an unbiased estimate of σ

● c4 = 0.9400 for samples of size 5,
⇒ Our estimate of the process standard deviation is

64
6.3.2. The x and s Control Charts with Variable Sample Size

In case, the sample sizes are variables

We should use a weighted average approach in calculating x and s
If ni is the number of observations in the ith sample, then use

As the center lines As the center lines

on the x control chart on the s control chart

65
Example 6.4: x and s Charts for the Piston Ring Data

Consider the data in Table 6.4,

which is a modification of the piston-ring
data used in Example 6.3.
Note that the sample sizes vary from
n = 3 to n = 5.
Use the procedure described on page
255 to set up the x and s control charts.

66
Example 6.4 - Solution
The weighted grand mean and weighted average standard deviation are computed from
equations 6.30 and 6.31 as follows:

Therefore, the center line of the x chart is x = 74.001 and the centerline of the s chart is
s = 0.0103. The control limits may now be easily calculated.
67
Example 6.4 - Solution

We have used the values of A3, B3, and B4

for n1 = 5. The limits for the second
sample would use the values of these
constants for n2 = 3.
The control limit calculations for all
25 samples are summarized in Table 6.5.

68
Example 6.4 - Solution

● The limits for the x chart are

● The control limits for the s chart are

69
6.3.2. The x and s Control Charts with Variable Sample Size

● An alternative is to base the control limit calculations on an average sample

size n . If the ni are not very different, this approach may be satisfactory in
some situations; it is particularly helpful when using in a presentation to
management.
● Since the average sample size ni may not be an integer, a useful alternative is
to base these approximate control limits on a modal (most common) sample
size.

70
6.3.2. The x and s Control Charts with Variable Sample Size
Estimation of σ
Average all the value of si for
which ni = 5 (the most
frequently occurring value of
ni).

Where the value of c4

used is for samples of
size n = 5

71
6.3.3 The s2 Control Chart
● s preferable to R for moderate to large sample sizes.
● Some practitioners recommend a control chart based directly on the sample variance s2,
the s2 control chart.
The parameters for the s2 control chart are

: the upper
and lower ∝/2 percentage points of the Chi-
square distribution with n - 1 degree of
freedom
s2: an average sample variance obtained
from the analysis of preliminary data.

● A standard value σ2 could be used in this equation instead of s2 if one were available.
● Note that this control chart is defined with probability limits.
72
 6.4
THE SHEWHART CONTROL
CHART FOR INDIVIDUAL
MEASUREMENTS
73
There are many situations in which the sample size used for
process monitoring is n = 1

For examples:

- Data become available relatively slowly,and it is inconvenient to allow sample sizes of

n > 1 to accumulate before analysis. The long interval between observations will cause
problems with rational subgrouping.

- Repeat measurements on the process differ only because of laboratory or analysis error,
as in many chemical processes.

=> In such situations, the control chart for individual units is useful

74
6.4. The Shewhart control chart for individual measurements

In many applications of the individuals control chart,we use the moving range
of two successive observations as the basis of estimating the process variability.
The moving range is defined as

It is also possible to establish a moving range control chart.

The procedure is illustrated in some examples

75
Example 6.5 Loan Processing Costs

The mortgage loan processing unit of

a bank monitors the costs of
processing loan applications. The
quantity tracked is the average weekly
processing costs, obtained by dividing
total weekly costs by the number of
loans processed during the week. The
processing costs for the most recent 20
weeks are shown in Table. Set up
individual and moving range
control charts for these data.

76
Example 6.5 - Solution
To set up the moving range chart, we use
❖ D3 = 0 and D4 = 3.267 for n = 2.
Therefore, the moving range chart has center line,
❖ Average MR = 7.79, LCL = 0,
❖ UCL = D4 ✕ Average MR = 3.267 ✕ 7.79 = 25.45

In minitab, notice that no points are out of control.

77
Example 6.5 - Solution

For the control chart for individual measurements, the parameter are

78
For the control chart for individual measurements
If a moving range of n = 2 observations is used, then d2= 1.128.
For the data in Table, we apply formula in previous slide:

There are no out-of-

control observations on
the individuals control
chart.

79
 Phase II. Operation and Interpretation of the Charts
Table below data on mortgage application processing costs (weeks 21 – 40) plotted indicates an
upward shift in cost around week 39.

80
 Phase II. Operation and Interpretation of the Charts

The chart for individuals shows a clear "shift in process level" pattern, followed by an out-of-control
signal at week 40. The moving range chart also reacts, displaying a single large spike at week 39, aiding in
pinpointing the shift in the mean. Caution is advised in interpreting moving range chart patterns due to
correlated moving ranges, while x chart assumes uncorrelated individual measurements. 81
 Average Run Lengths (ARL)

● Crowder (1987b) studied the average run length (ARL) of the combined control chart for
individuals and moving range.
=> ARL0 is much less than its of Shewhart which using conventional three-sigma limits.
=> Employing three-sigma limits on individuals chart and calculating the upper control limit
on the moving range chart, get a better result

82
 Average Run Lengths (ARL)
● The ability of the individuals control chart to detect process shifts by looking at the OC
curves or the ARL curves

Operating-characteristic Average run length (samples) for the chart

curves for the chart with three- with three-sigma limits, where the
sigma limits. b = P process mean shifts by ks
83
Average Run Lengths (ARL)

The poor ability of the individuals control chart to detect small shifts
Example: Continuous chemical process, take samples hourly with the mean of
about one standard deviation occurs
=> Potentially devastating economic consequences
=> Limits the usefulness of the individuals control chart
❖ Suggestion:
- The cumulative sum control chart
- Exponentially weighted moving average control chart

84
Normality

Borror, Montgomery, and Runger (1999) found that non-normal data

significantly impact the in-control average run length (ARL).
Example
With three-sigma limits:
- ARL0 for gamma-distributed, data ranges from 45 to 97
- ARL0 for t-distributions, varies from 76 to 283
(DEPENDING on DEGREES OF FREEDOM)

85
Normality

● If the process shows evidence of moderate departure from normality

=> Inappropriate control limits for phase II process monitoring

● Approaches:
- Determine the control limits of individuals control chart based on the
percentiles Histogram (large sample)
Probability distribution fit to the data

- Conduct normally distributed transformation the original variable, then

apply control charts
86
Example 6.6 The Use of Transformations
Table 6.8 presents consecutive measurements on the resistivity of 25 silicon wafers after an epitaxial
layer is deposited in a single-wafer deposition process.
Construct an individuals control chart for this process.

87
Example 6.6
Construction of a normal probability plot of the resistivity measurements

The distribution of resistivity has a MORE APPROXIMATELY New variable

LONG TAIL to the RIGHT by a NORMAL DISTRIBUTION
=> Log transformation
=> Produce a distribution
closer to normal. 88
Example 6.6
The last column of Table shows the moving ranges of the natural log of resistivity

We use Minitab to present the individuals and moving range control charts for the
natural log of resistivity
89
Example 6.6
Figure 6.24 presents the individuals and moving range control charts for the natural log
of resistivity

=> Notice that there is no indication of an out-of-control process 90

 6.5
SUMMARY OF PROCEDURES
–
FOR x ,
R, AND s CHARTS
91
6.5. Summary

Formulas for x, R, and s charts when standard values for μ and σ are given

92
 6.5. Summary
The corresponding summary when no standard values are given and trial control limits
must be established from analysis of past data.

The constants given for the s chart assume that n – 1 is used in the denominator of s

93
 6.6
APPLICATIONS OF
VARIABLES
CONTROL CHARTS

94
6.6. Applications of variables control charts

● Improve Suppliers’ Processes

● Purchase a Machine Tool
● Short-Run Job-Shop
● Transactional and Service Businesses
● Selecting Rational Subgroups
● …

95
THANK YOU
FOR YOUR
ATTENTION!
GROUP 05