Lean Six Sigma Green Belt

Certification Course
Statistical Process Control
 Learning Objectives

By the end of this lesson, you will be able to:

Explain the basics of Statistical Process Control

Choose appropriate traditional control charts

Differentiate between CUSUM and EWMA Charts

Statistical Process Control (SPC) Basics
 Statistical Process Control Basics

Walter A Shewhart developed statistical process control in 1924.

SPC

Visual
Monitoring

Inputs Process and its Parameters Outputs

Statistical Statistical
measures measures
 Benefits of SPC Basics

Special causes of variability

SPC separates
Common causes of variability

recognizes Unexpected changes in the process

output

identifies stable
zones Variables with unknown specification
limits

useful external
information Continuous process improvement and
online monitoring

The concept of Process Control can be used in the Measure Phase, when we check for
data stability.
 Features of Common Cause Variation

➢ Small variations in raw materials

➢ Variations due to manual interventions in manual

processes

➢ Responses from machines or systems, etc.

Aware of
common cause Frequency Predictability
variation

Huge investment and time to identify

Repetitive
and remove them

Part of the Process
Within tolerance or specification limits
 Statistical Process Control Basics

➢ Machine or system crash

➢ Delay in supply of raw materials

➢ Huge variations in the raw materials, etc.

Frequency Predictability

Less investment and time to

Part of the Process

Non-repetitive identify and remove them

Variation affects the flow of the process due to which the
defects appear
 Statistical Process Control Basics

Subgroup or Sample → A collection of units

that are produced or created under the same
set of conditions

Rational Subgroup
Represents the process at a particular point in time
Measurements must be taken within a short span of
time but should also be independent of each other
 Statistical Process Control Basics

Approach 1 Approach 2

Random for all

Every other group of process output over
Sample Sample
10 consecutive units the sampling interval

➢ Used to make decisions about the acceptance of a

➢ Used to detect process shifts product

➢ A process shift happens over time and is reflected in ➢ Detecting the shifts of the output variable to an out
the variation of the output variables of control state and back into an in-control state
between samples

Example: Example:

Process 10 consecutive units Process 100 parts per hour

sample sample

every other group 5 randomly selected parts every hour

 Statistical Process Control Basics

Rational subgrouping refers to the selection of subgroups or samples in a way that if special
causes are present, chance for differences between subgroups will be maximized and chance
for differences due to special causes within a subgroup will be minimized.

The interval bars represent the within-subgroup variation

The red line that connects the averages of the consecutive subgroups represents the between-
subgroup variation

http://support.minitab.com/en-us/minitab/17/topic-library/quality-tools/control-charts/data/within-subgroup-and-between-subgroup-variation/
 Statistical Process Control Basics

Walter Shewhart, 1920s

Characteristics of control charts are as follows:

❑ Similar to run or trend charts, with an addition of a control limit line and an average or center
line
❑ Can be used with discrete or continuous data
❑ Control limits (UCL and LCL) are typically set at approximately three standard deviations from
the center line
❑ Specification limits (USL and LSL) normally do not appear on them
Statistical Process Control Basics
 Statistical Process Control Basics

Most of the points fall

Control Charts Tracking Process Statistics Process is in
within the bounds of the
over time control
control limits

Detecting the presence of Points do not display any

Special Causes non-random patterns

A process is in control when most of the points fall within the bounds of the
control limits and the points do not display any non-random patterns.
 Statistical Process Control Basics

Uses control σmean from the data’s grand

Standard Control
Chart limits average (ന
X, average of the
sample averages, or μ).

The probability of an out-of-control point when the

process has not changed → 0.27%

2σ the chance of type I or alpha error

4σ the chance of type II or beta error

Walter Shewhart had set 3σ limits on control charts with the

belief that when the process goes beyond these limits, it needs
correction.
 Statistical Process Control Basics

An Out-Of-Control (OOC) condition is indicated if one of the following is true:

p(f) = 0.27% p(f) = (0.5)8 = 0.39%

1 point is 8 consecutive
outside the points are
Control Limits above the
(either above Center Line (CL)
UCL or below or consecutively
LCL) below the CL

2 out of 3 6 to 8 points
points are are
within 1 σmean of consecutively
either the UCL increasing or
or the LCL decreasing
3!
p(f) = (2!1!)(0.023)2(0.477) = 0.08% for p(f) = (0.5)6 or (0.5)8 =
1.6% to 0.39%
one side
 Tips

Identify the purpose for data collection and try to determine what kind
of data may be needed for measurement

If you don’t adapt data to filter

Identify measures that are any noise factors from the
Some Tips process, the control chart will
used daily
show you wrong results
Traditional Control Charts
 Choosing an Appropriate Control Chart

I-MR Chart
Individual
Data Points (Depicts the variability of individual
characteristics over time)
(Pulling one
sample at
Continuous fixed
data frequency)
sampling X and R Chart
ഥ
Subgroups (If n is between 2 and 9)
(Taking
X and s Chart
ഥ
periodic group
data) (When standard deviation is calculated
and n≥10)
 Choosing an Appropriate Control Chart

np Chart
Constant Subgroup Size
(Number of Units Rejected)
Defectives
p Chart
Varying Subgroup Size (Percentage of Units
Rejected)
Discrete
Data
c Chart
Constant Subgroup Size
(Number of Defects)
Defects
u Chart
Varying Subgroup Size (Average Number of
Defects per Opportunity)
 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ
 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ

❑ It is the plot of the means of the

subgrouped data

❑ It shows inter-subgroup or between-

subgroup variation

❑ The control limits are calculated based on

mean of means, range or standard
deviation, and other factors
 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ

❑ It is the plot of the value of subgroup range

❑ The R chart shows intra subgroup

❑ One of the most sensitive charts to track

and identify special causes of variation

❑ Can be plotted with any type of data

 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ

❑ It is the plot of the standard deviation of

the subgroup range

❑ One of the most sensitive charts to track

and identify special causes of variation

❑ Can be plotted with any type of data

 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ

❑ It is of the same subgrouped data as the

X and s chart
ഥ

❑ One chart is the X bar and the other is the R

chart

❑ Can be plotted with any type of data

 ഥ Chart Principles
𝐗

X → average of each subgroup of data

ഥ

X chart → the subgroup average data will be plotted

ഥ

X chart
ഥ s chart X and s chart
ഥ

R chart X and R chart

ഥ

❑ It is of the same subgrouped data as the

X and R chart
ഥ

❑ One chart is the X bar and the other is the s

chart

❑ Can be plotted with any type of data

 Control Limit Formulas

Control
Limit
Formulas

Control
Chart
Constants
 UCL and LCL in ഥ
X and R Chart

UCLXҧ = ന
X + A2 Rഥ
LCLXҧ = ന
X - A2 R
ഥ

UCLR = D4 R
ഥ
LCLR = D3 R
ഥ

X → grand average and R

ന ഥ → average of the range.

A2, D3, and D4 are values from the control chart table
 UCL and LCL in ഥ
X and s Chart

UCL = ന
X + A3 Sത
LCL = ന
X - A3 Sത

UCL = B4 Sത
LCL = B3 Sത

❑ S → Standard deviation of each subgroup data

❑ The data is divided into subgroups.
❑ Standard deviation is calculated for each subgroup.
Values for A3, B3, and B4 are constant and are taken from the control chart
table. ഥ
X and s charts are used to track process variation where the
subgroup sample size ≥ 9.
 X and R and Subgroup Data
ഥ

Q
Establish 1 σ process limits for the data set shown. Use the table of
control chart constants for values of A2, D3, and D4.

X Chart
ഥ
Table for control chart constants

n A2 D3 D4

2 1.88 0 3.27

3 1.02 0 2.57

4 0.73 0 2.28

5 0.58 0 2.11

6 0.48 0 2.00
 X and R and Subgroup Data
ഥ

A ● In ഥ
X and R chart, point SG 6 is the point of change in the process from
below the center line to above the center.
● No points are outside control limits in the given process; examine points 6
and 7 on ഥX chart, and points 10 and 11 on the R chart for rule #4 (If 2 out of
3 points are within 1 σmean of either the UCL or the LCL).
 X and s and Subgroup Data
ഥ

Q
The data in subgroups with 10 samples in each subgroup is given here along with
the ഥ
X chart. Using this data, find out if the process is in control.

Table for control chart constants X Chart

ഥ

n A3 B3 B4

2 2.659 0 3.267

3 1.954 0 2.568

4 1.628 0 2.266

5 1.427 0 2.089

6 1.287 0.030 1.970

7 1.182 0.118 1.882

8 1.099 0.185 1.815

9 1.032 0.239 1.761

10 0.975 0.284 1.716

 X and s: Constructing Chart
ഥ

A
● The ഥ
X chart point SG 10 is the variation of the point from the mean.
● Also, points 4, 10, and 23 have more variation from the center. These points can be
analyzed further.
● The points are within the limits, and hence the process is in control.
 I-MR Chart Principles

I-chart Plot of the individual

data points

I-MR Charts

MR-chart Plot of the moving

Used range of the previous
individuals

❑ When subgroup variation = 0 or no subgroups exist; and

❑ With data points from destructive testing or batch
processing, or summary data from a time period
 I-MR Chart Principles

I-MR charts → sensitive to trends, cycles, patterns, and normality.

Control limits of the I-MR chart are calculated using a similar method as the ഥ
X and R chart.
 I-MR and Individual Data: Example

Q
The QC department measures the strength of its milk cartons once in every
hour. Is the process in control?
● Since the data is individual data, the I-MR chart will be used here.
● This is an example of a destructive test.

M
I
L
K

Strength of the milk carton

 I-MR and Individual Data: Constructing Chart
A
● Moving range is the absolute value of difference between the last two data points.
● In I-chart, point 16 is close to the upper limit (analysis required).
● No points are out of control in the process.
 I-MR Chart: IT/ITES Example

Given is the data used to study the number of calls handled per hour in
call center operations. This data was studied using I-MR charts to check
 if the process is in control. The data, I-MR chart, and analysis are as
follows:
Call Center
Data: I-MR Chart:
Analysis of I-chart
❑ All points are closer
to mean values
❑ Process is well within
control

Analysis of MR-chart
❑ Few points are closer
to LCL.
❑ No points are outside
of the control limits
 Control Charts for Attribute Data

Based on sample size and data type (defects or defectives), the following types of
control charts can be selected:

np chart c chart p chart u chart

If the sample size is

If the sample size is consistent and the data If the sample size is If the sample size is
consistent and the data type available changes inconsistent and the inconsistent and the
type is defectives from defectives to defects data type is defectives data type is defects

Control limits may be constant, such as ഥ

X and R charts (for np and c
charts), or vary depending on sample size (for p and u charts.)
 np Chart Principles

The np chart is used to measure the non-conforming proportions or number of defectives

within a standardized group size. Some of the principles of np chart are as follows:

The expectation is that the same proportion exists

in each group

The np chart follows binomial distribution

Large subgroups are required (50 minimum) for

this chart

Subgroup size must be constant

Control limits will be constant for an np chart

 np Chart: Formulae

Important formulae of np chart are as follows:

D
● Proportion of p =
n

D
● np = n ∗ =D
n

● Control Limits = np ± 3√np (1−𝑝 ̅)

where, D = Defectives
 np Charts and Uniform Subgroup Size

Q
The sourcing department at Java Coffee House Worldwide measures 125
purchase orders daily and records the number of entry errors in them.
The tabulated data is given here. Is the order entry process in control?

● Since the data has a constant subgroup size (orders processed) of

defectives, an np chart will be used.

● Assumption is that there is only one error per order possible.

Purchase Order

------
-----

Java House Coffee

 u Chart: Constructing Chart

A ● In np chart, point 12 is beyond the control limit of three standard deviations.

Analysis must be done to find the reason and take corrective action if necessary.
● Hence, point 12 is out of control in the process.
 p Chart Principles

The p chart is used to measure the non-conforming proportion or defectives.

Some of the principles of p chart are as follows:

The expectation is that the same proportion exists

in each group

The p chart follows binomial distribution

The subgroup size should at least be 50

Subgroup size need not be constant

Control limits may vary from subgroup to subgroup

based on the subgroup size.
 p Chart: Formulae

Important formulae of p chart are as follows:

𝑝ҧ 1−𝑝ҧ
Control Limits = 𝑝ҧ ± 3
𝑛

Note: When n changes, control limit

also changes.
 p Charts and Varying Subgroup Size

Q
The sourcing department in Java Coffee House Worldwide measures
the number of entry errors on a daily basis. The tabulated data is
presented here. Is the order entry process in control?

● Since the data has varying subgroup sizes (orders processed) of

defectives, a p chart will be used

Order Entry
----
----
----

Java House Coffee

 p Charts and Varying Subgroup Size

A
● In a p chart, point 12 has gone beyond the limit of 3 sigma level. Analysis must be
done to find the reason and take corrective action if necessary.
● Hence, point 12 is out of control in this process.
 c Chart Principles

To form a c chart, measure the number of occurrences of non-conforming

defects. Some of the principles of the c chart are as follows:

The c chart follows a Poisson distribution.

The sample size is fixed or the area of opportunity
is constant
Used to identify attribute data for the sample

Each count is a subgroup of samples

The control limits will be constant

The subgroup size should at least be 20

! Control Limits = 𝑐ҧ ± 3√𝑐ҧ

 c Chart Principles

Q
Final inspection grades the tinted glass on the number of white
specs. The product is priced by grade. White specs are defects, not
defectives, and are measured over a constant sample area; so c chart
will be used. Is the process in control?

● Since the data is for defects, c chart will be used.

Defects

Tinted Glass
 c Chart: Constructing Chart

A
● Points 2, 3, 4, 12, 13, 16, and 17 are out of control in this process; additionally, points 7,
9, 18, and 19 break rule #4.
● In this c chart, the process is not stable and many points go beyond 3 sigma control
levels. Analysis must be done to find the reason and take corrective action.
● The process is not in control.
 u Chart Principles

The u chart is used to measure the non-conforming proportion or defectives.

Some of the principles of u chart are as follows:

The u chart follows a Poisson distribution.

Used to identify attribute data for the sample

Sample size is not fixed

Control limits may vary

The subgroup size should at least be 20

 u Chart: Formulae

Important formulae of u chart are as follows:

ഥ
𝑢
Control limits = 𝑢ത ± 3√( )
𝑎

Where, a = area of opportunity

 u Chart

Q The plastics operation counts defects after a “run” which is undetermined

in length (once started, it continues until all material is used). Is the process
in control?

Since the count of defects has a varying area of opportunity and the length
of runs is not constant, u chart will be used.

Plastics Operation
Defects
 u Chart: Constructing Chart

A ● In this u chart, point 18 has gone beyond the 3 sigma level. Analysis must be
done to find the reason and corrective action must be taken if necessary.
● Point 18 is out of control in this process.
CuSum and EWMA Charts
 CuSum Charts

If µ0 → the target for the process mean

𝑋ത𝑗 → the average of the jth sample

Then, the cumulative sum control chart is formed by plotting the quantity as follows:
𝑖

𝐶𝑖 = ෍(𝑋ത − µ0 )
𝑗=1
 EWMA Charts

To plot data to detect small Example: Used in Stock Modeling

To monitor process mean
shifts over a small period Software Packages
or variance
of time

EWMA chart –
time weighted
control chart

Predict performance in the

More weight on recent
next period of change or
observations
instability
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits

❑ Weight is applied to the most

recent rational subgroup average

❑ λ is between 0 and 1

❑ Usually λ is selected between

0.05 and 0.25
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits

❑ Multiple of the rational subgroup

standard deviation to set control limits

❑ Typically set to 3 to match other control

charts

❑ Can be reduced for if λ is small (if 0.1 ≥

λ , 2.6 < L < 2.8)
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits

This is represented by T or the target

value of the quality characteristic
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits

T → the estimates of the long-term process mean

S → standard deviation established
n → the number of samples in the rational
subgroup

The limits widen for each

successive rational subgroup
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits

Where,

xi → current rational subgroup average

zi – 1 → running average of all preceding

observations
 EWMA Parameters

L– Control
λ – Weight Centerline Plotted Stat Example
Multiplier Limits
 Key Takeaways

SPC aids in the visual monitoring of a process and controlling

its parameters

A rational subgroup represents the process at a particular

point in time

Control chart plots and processes the input and output data
over a period of time

Out-Of-Control (OOC) condition indicates special cause for

variation

p chart is used for defectives which have a varying subgroup

and is represented as a percentage of units rejected
 Key Takeaways

c chart is used for defective which have a constant

subgroups size and is represented as number of defects

u chart is used for defects that have a varying subgroup size and
is represented as an average number of defects per opportunity

An I-chart is a plot of the individual data points

MR chart is a plot of the moving range of the previous individuals

A CuSum chart plots the cumulative sums of the deviations

of the sample values from the target value

EWMA stands for Exponentially Weighted Moving Average Chart

Knowledge Check
Knowledge
Check While monitoring a process using a control chart a team observed six consecutive data
1 points increasing. What does this mean about the process?

A. The process is not stable but it is in control

B. The process is stable and in control

C. There is not enough data to determine

D. The process is not stable and not in control

 Knowledge
Check While monitoring a process using a control chart a team observed six consecutive data
1 points increasing. What does this mean about the process?

A. The process is not stable but it is in control

B. The process is stable and in control

C. There is not enough data to determine

D. The process is not stable and not in control

The correct answer is D

The probability of observing 6 data points consecutively increasing is 1.6% and therefore not likely which indicates
that the process is not stable or in control
Knowledge
Check What type of control chart should be made if measuring defectives and the subgroup
2 size changes?

A. X and R
ഥ

B. p chart

C. u chart

D. c chart
 Knowledge
Check What type of control chart should be made if measuring defectives and the subgroup
2 size changes?

A. X and R
ഥ

B. p chart

C. u chart

D. c chart

The correct answer is B

If defectives are measured with a subgroup size that is not constant a p chart should be used.
Knowledge
Check
Which of the following is NOT true of special causes of variation?
3

A. Difficult to eliminate

B. Occur sporadically in the process

C. Can be identified with help of a control chart

D. Is not a part of the process

 Knowledge
Check
Which of the following is NOT true of special causes of variation?
3

A. Difficult to eliminate

B. Occur sporadically in the process

C. Can be identified with help of a control chart

D. Is not a part of the process

The correct answer is A

It takes less investment to remove special cause variations than to remove common cause variation. Hence, special
cause variations are not difficult to eliminate.
Knowledge
Check
What is true of a control chart?
4

A. Does not assume normality of data

B. Typically the UCL and LCL are set to 2 standard deviations away from the mean

C. The probability of having an out of control condition is 0.27%

D. Typically show spec limits

 Knowledge
Check
What is true of a control chart?
4

A. Does not assume normality of data

B. Typically the UCL and LCL are set to 2 standard deviations away from the mean

C. The probability of having an out of control condition is 0.27%

D. Typically show spec limits

The correct answer is C

From the options presented the only choice that is true about control charts is C, in which the control limits are set at
3 standard deviations from mean because it provides a 0.27% of an out of control condition.
Knowledge
Check A team wants to measure defective items in which the subgroup size is constant. What
5 type of control chart could be used?

A. I-MR

B. c-chart

C. u-chart

D. np -chart
 Knowledge
Check A team wants to measure defective items in which the subgroup size is constant. What
5 type of control chart could be used?

A. I-MR

B. c-chart

C. u-chart

D. np-chart

The correct answer is D

Since we are dealing with attribute data, and looking for defectives with constant subgroup size the best chart to use
would be the np-chart.
Knowledge
Check
What is the benefit of using CuSum or EWMA charts over traditional control charts?
6

A. Detects large shifts

B. Detects small shifts

C. Easier to calculate

D. More accurate
 Knowledge
Check
What is the benefit of using CuSum or EWMA charts over traditional control charts?
6

A. Detects large shifts

B. Detects small shifts

C. Easier to calculate

D. More accurate

The correct answer is B

The CuSum and EWMA charts are typically used to plot the data to detect small shifts over a small period of time.