4-8 SPC - Variables Data

Statistical Process Control for Variables Data
(SPC)
MQPM
© 2001 ConceptFlow 1
Module Objectives
By the end of this module, the participant will be able to:

• Apply SPC rules
• Interpret run and trend patterns in control charts
• Create and interpret
• Xbar-R Charts
• I-MR charts
• Target I-MR charts
MQPM
Why Learn About SPC for Variables?
SPC for variable data will:

• Keep process centered
• Minimize variation
• Reduce excursions
• Validate improvements
• Focus Six Sigma® process activity
MQPM
What is SPC for Variables?
SPC for variable data is

• Industry standard control language
• Reliable, easy method of determining
• Common cause variation
• Special cause variation
• Graphical communication
• Set of statistical tools for analyzing variables performance data
MQPM
Introduction to SPC
MQPM
SPM and SPC
• Statistical Process Control before six sigma

• Statistical Process Monitoring
• Usual focus of SPC tools
• Looks at output
• Corrective action after output is out of control
• Statistical Process Control after Six Sigma
• Same tools, additional focus
• Focus on inputs
• Corrective action on inputs prior to output out of control
MQPM
Sources of Variation
Statistical Environment Statistical

Process Process
Control Monitor
KPIVs Process KPOVs
Machines Places
Materials Process Policies

Control
Methods Procedures
Operators People
Measurements
MQPM
SPC Defined
Statistical Process Control

• Is application of statistical tools and methods to provide feedback
• Sets limits of variation
• Provides trigger for action
MQPM
SPC Function
SPC Charts
• Used to monitor and control process under local responsibility
• Require process owners to
• take measurements
• Plot and interpret data
• Take action
• Provide a history of the process
MQPM
Components of a Control Chart
Random Variation Region

Observation 10
10
9
Upper Control
Limit
Observation value
8
7
6
5 Mean
4
3
2 Lower Control
1
Limit
0
0 5 10 15 20
Observation number
Nonrandom Variation Region
MQPM
Statistics of a Control Chart
Random Variation Region
10
UCL
9
+ 3
Observation value
8
7
6
5 Mean
4
3
2 LCL
1
0 - 3
0 5 10 15 20
Observation number
Nonrandom Variation Region 99.73%

area
MQPM
Establishing Process Control Limits
Control limits are

• Are statistical limits set +/- 3 standard deviations from the mean
• Set when process is in control
• Fixed at baseline value
• Adjusted for improvements
• Never widened
• Control limits are not related to specification limits
Control Limits are not specification limits

MQPM
Definition of Control
In control is
• A statistical term for process variation
• Within three standard deviations of the mean
• That is random without cause
• That does not show run patterns
• That does not show trend patterns
• No assignable cause variation
MQPM
Western Electric Rules for Control
Overview
• Any point outside control limits
UCL • 7 consecutive points on same side of
A centerline
B • 7 consecutive points increasing or
decreasing
6 4 2
C
• 2 of 3 points in same zone A or
C
beyond
B • 4 of 5 points in same zone B or
A beyond
LCL
• 14 consecutive points alternating up
and down
• 14 consecutive points in either zone
C
Established rules for
run and trend
analysis
MQPM
Nelson Tests for Special Causes
UCL 1. Any point outside control limits

A
2. 9 consecutive points on same
B side of centerline
C 3. 6 consecutive points increasing
6s 4s 2s or decreasing
C
4. 2 of 3 points in same zone A or
B
beyond
A
LCL 5. 4 of 5 points in same zone B or
beyond
6. 14 consecutive points
alternating up and down
Tests proposed by Lloyd
7. 15 consecutive points in either
Nelson (1984) and used by zone C
MINITAB™ for run and trend 8. 8 points in a row outside zone
analysis C, same side of centerline
MQPM
Which Tests are Better?
Western Electric Nelson

• Any point outside control limits • Any point outside control limits
• 7 consecutive points on same side of • 9 consecutive points on same side of
centerline centerline
• 7 consecutive points increasing or • 6 consecutive points increasing or
decreasing decreasing
• 2 of 3 points in same zone A or beyond • 2 of 3 points in same zone A or beyond
• 4 of 5 points in same zone B or beyond • 4 of 5 points in same zone B or beyond
• 14 consecutive points alternating up and • 14 consecutive points alternating up
down and down
• 14 consecutive points in either zone C • 15 consecutive points in either zone C
• 8 points in a row outside zone C, either
side of centerline
MQPM
False Alarm Rates are the Key
Nelson False Alarm Rate

• Any point outside control limits .0027
• 9 consecutive points on same side of centerline Approx .003
• 6 consecutive points increasing or decreasing Approx .003
• 2 of 3 points in same zone A or beyond .00305
• 4 of 5 points in same zone B or beyond .0043
• 14 consecutive points alternating up and down Approx .004
• 15 consecutive points in either zone C Approx .003
• 8 points in a row outside zone C, either side of Approx .003
centerline
The Nelson tests are designed so that the false alarm

rates for all tests are approximately the same. The
Western Electric rules do not have this property.
MQPM
Nelson Test 1
Rule 1: Any point outside control limits
MQPM
Nelson Test 2
Rule 2: 9 consecutive points on same side of centerline
MQPM
Nelson Test 3
Rule 3: 6 consecutive points increasing or decreasing
MQPM
Nelson Test 4
Rule 4: 2 of 3 points in same zone A or beyond
MQPM
Nelson Test 5
Rule 5: 4 of 5 points in same zone B or beyond
MQPM
Nelson Test 6
Rule 6: 14 consecutive points alternating up and down
MQPM
Nelson Test 7
Rule 7: 15 consecutive points in either zone C
MQPM
Nelson Test 8
Rule 8: 8 points in a row outside zone C, either side
MQPM
Nelson Tests in MINITAB
Stat>Control Charts>Xbar-R
MQPM
Control Chart Roadmap
Variable
No
Xbar-s
N<10
chart
Yes
Yes No
I-MR N=1 Xbar-R
Chart Chart
MQPM
Xbar-R: Average, Range Charts
MQPM
Xbar-R Chart Principles
Xbar-R Charts (and Xbar-s) are two separate charts of the same
subgroup data
• Xbar chart is a plot of the subgroup means
• R chart is a plot of the subgroup ranges (or if s, plot of subgroup
standard deviation)
• Most sensitive charts for tracking and identifying assignable cause of
variation
• Based on control chart factors that assume a normal distribution within
subgroups
• Establish three sigma process limits
MQPM
Xbar-R and Subgroup Data
X1 X2 X3 X4 X5 • You measure the revenue per FA

SG 1 43.8 43.7 47.2 46.3 44.4 earned from 20 complexes over five
SG 2 44.7 43.2 45.7 45.8 44.4
days.
SG 3 45.3 43.8 44.3 46.2 46.6
SG 4 45.4 44.1 44.6 45.3 45.0 • Is the process in control?
SG 5 43.8 45.6 44.6 44.8 45.0 • Since the data is subgroup data an
SG 6 45.7 46.0 45.6 45.9 46.5 Xbar-R chart will be used
SG 7 46.5 45.6 45.7 46.9 45.6
SG 8 46.1 45.8 45.5 45.9 45.1
SG 9 44.5 44.0 45.4 45.8 44.7
SG 10 47.8 43.6 44.5 46.0 44.5
SG 11 45.5 45.4 42.8 47.0 45.1
SG 12 46.8 43.5 43.4 46.0 45.0
SG 13 44.2 44.7 46.1 44.5 45.8
SG 14 44.6 44.7 45.2 43.0 45.5
SG 15 46.0 46.0 45.0 44.5 47.2
SG 16 46.3 43.7 44.8 46.0 45.4
SG 17 43.2 43.0 45.6 44.8 45.4
SG 18 45.2 45.1 46.9 45.0 44.8
SG 19 44.6 44.5 44.6 43.7 45.1
SG 20 45.6 44.2 46.0 43.5 45.9
MQPM
Constructing an Xbar-R Chart Graph
X1 X2 X3 X4 X5 Xbar Range
SG 1 43.8 43.7 47.2 46.3 44.4 45.1 3.5
SG 2 44.7 43.2 45.7 45.8 44.4 44.8 2.6
SG 3 45.3 43.8 44.3 46.2 46.6 45.2 2.8
SG 4 45.4 44.1 44.6 45.3 45.0 44.9 1.3
SG 5 43.8 45.6 44.6 44.8 45.0 44.8 1.8
SG 6 45.7 46.0 45.6 45.9 46.5 45.9 0.9
SG 7 46.5 45.6 45.7 46.9 45.6 46.1 1.3
SG 8 46.1 45.8 45.5 45.9 45.1 45.7 1.0
SG 9 44.5 44.0 45.4 45.8 44.7 44.9 1.8
SG 10 47.8 43.6 44.5 46.0 44.5 45.3 4.2
SG 11 45.5 45.4 42.8 47.0 45.1 45.2 4.2
SG 12 46.8 43.5 43.4 46.0 45.0 44.9 3.4
SG 13 44.2 44.7 46.1 44.5 45.8 45.1 1.9
SG 14 44.6 44.7 45.2 43.0 45.5 44.6 2.5
SG 15 46.0 46.0 45.0 44.5 47.2 45.7 2.7
SG 16 46.3 43.7 44.8 46.0 45.4 45.2 2.6
SG 17 43.2 43.0 45.6 44.8 45.4 44.4 2.6
SG 18 45.2 45.1 46.9 45.0 44.8 45.4 2.1
SG 19 44.6 44.5 44.6 43.7 45.1 44.5 1.4
SG 20 45.6 44.2 46.0 43.5 45.9 45.0 2.5
Average 45.13 2.36
MQPM
Defining the Xbar-R UCL and LCL
UCLX  X  A2 R
LCLX  X  A2 R
UCLR  D4 R
UCLR  D3 R
A2, D3 and D4 are Shewhart control constants

MQPM
Calculating the Xbar-R UCL and LCL
Shewhart Control Chart Constants

n D4 D3 A2
2 3.27 0.00 1.88
3 2.57 0.00 1.02
4 2.28 0.00 0.73
5 2.11 0.00 0.58
6 2.00 0.00 0.48
7 1.92 0.08 0.42
8 1.86 0.14 0.37
9 1.82 0.18 0.34
n is the subgroup size
UCLR  D4 R  2.11* 2.355  4.97

UCLR  D3 R  0 * 2.355  0
Calculated values UCLX  X  A2 R  45.13  0.58 * 2.355  46.49

agree with Minitab LCLX  X  A2 R  45.13  0.58 * 2.355  43.76
MQPM
Xbar-R Charts in Minitab
Step 1
• Copy or enter the data by subgroups into the worksheet
• Open file SPC VARIABLE XBAR.MTW
MQPM
Step 2
Stack the data
MQPM
Step 3
Stat>Control Charts>Xbar-R
MQPM
Step 4
MQPM
Step 5
MQPM
Xbar-R Class Exercise
Using “Xbar Charts Data” tab of file SPC Variable Class Exercises.xls
1. Find Xbars, Xdbar and Rbar
2. Determine applicable Shewhart constants
3. Calculate UCL and LCL for Xbar and R
4. Copy the data into Minitab
5. Stack the data
6. Verify your calculations
7. Determine if process is in control
MQPM
Also called X-MR
I-MR: Individual, Moving Range Charts
MQPM
I-MR Chart Principles
I-MR Charts are two separate charts of the same data

• I chart is a plot of the individual data
• MR chart is a plot of the moving range of the previous individuals
• I-MR charts are sensitive to trends, cycles and patterns
• Used when subgroup variation is zero or no subgroups exist
• Destructive testing
• Batch processing
MQPM
I-MR and Individual Data
Revenue
2.38
2.06
2.46
Once a day the office measures
1.96 the revenue generated by its’
2.22
2.44 FA’s.
2.16
2.13
1.97
Is the process in control?
2.29
2.07
1.97
2.09
2.16
2.57
2.83
2.04 Since the data is individual data an I-MR
2.13
2.55 chart will be used.
2.39
MQPM
Constructing an I-MR Chart
Graph
Revenue MR (2)
1 2.38
2 2.06 0.32
3 2.46 0.40 Revenues
4 1.96 0.50
5 2.22 0.26
6 2.44 0.22
7 2.16 0.28
8 2.13 0.03
9 1.97 0.16
10 2.29 0.32
11 2.07 0.22
12 1.97 0.10
13 2.09 0.12
14 2.16 0.07
15 2.57 0.41
16 2.83 0.26
17 2.04 0.79
18 2.13 0.09
19 2.55 0.42
20 2.39 0.16
Ave 2.244 0.270
Note: calculated for a

moving range of 2
MQPM
Defining the I-MR UCL and LCL
Revenues
UCLX  X  E2 R
LCLX  X  E2 R
UCLMR  D4 R
UCLMR  D3 R
E2, D3 and D4 are Shewhart control constants

MQPM
Calculating the I-MR UCL and LCL
Revenues Shewhart Control Chart Constants

n D4 D3 A2 E2
2 3.27 0.00 1.88 2.66
3 2.57 0.00 1.02 1.77
4 2.28 0.00 0.73 1.46
5 2.11 0.00 0.58 1.29
6 2.00 0.00 0.48 1.18
7 1.92 0.08 0.42 1.11
8 1.86 0.14 0.37 1.05
9 1.82 0.18 0.34 1.01
n is the data or moving range subgroup size
UCLMR  D4 R  3.27 *0.270  0.88

UCLMR  D3 R  0 *0.270  0
UCLX  X  E2 R  2.244  2.66 *0.270  2.96

Calculated values
agree with Minitab LCLX  X  E2 R  2.244  2.66 *0.270  1.52
MQPM
I-MR Charts in Minitab
Step 1
• Copy or enter the data by subgroups into the worksheet
• Open file SPC VARIABLE IM.MTW
Revenues
MQPM
Step 2
Stat>Control Charts>I-MR
Revenues
MQPM
Step 3
Revenues
MQPM
I-MR Charting an Improvement in Process
A process “improvement” has been made to increase

revenues. Is it real?
Revenues
Stack the data
MQPM
I-MR Shows Two Populations
Recalculating limits based upon improved

statistics show clearly that “old” process is
significantly different.
MQPM
I-MR Class Exercise
Using “IMR Charts Data” tab of file SPC Variable Class Exercises.xls,
1. Find Xbar, Xdbar and Rbar
2. Determine applicable Shewhart constants
3. Calculate UCL and LCL for Xbar and R
4. Copy the data into MINITAB
5. Stack the data
6. Verify your calculations
7. Determine if process is in control
MQPM
Target I-MR Charts
MQPM
Target I-MR Chart Principles
Target I-MR Charts are two separate charts of the same data
• Individuals plotted as difference from target
• MR chart is a plot of the moving range of the previous individuals
• I-MR charts are sensitive to trends, cycles and patterns
• Useful when trying to predict widely varying “parent” individuals
• Inventory levels
• Forecasting
MQPM
Target I-MR and Actual Data
Actual
132
96
127
The marketing department uses a
177 demand forecasting process for
126
120
weekly revenue forecasting.
133
185 Is their demand forecasting process
152
148 in control?
189
148
163
139
131
111
143
166 Since the data is individual data
134 an I-MR chart is used.
135
MQPM
I-MR Chart of Actual data
Looks like the forecasting is in

control, but dig a little deeper…
MQPM
Need for Target I-MR Charting
Actual Target Delta

132 138 6
96 99 3 Demand forecasting does not
127 127 0
177 175 -2 produce “actual” – clients do.
126 128 2
120 123 3 Demand forecasting produces a
133 135 2
185 166 -19 demand target. The difference
152 154 2 between the forecast and the
148 154 6
189 186 -3 actual is the true measure of the
148 153 5
163 161 -2
process.
139 143 4
131 136 5
111 133 22
143 143 0
166 171 5 I-MR chart the difference
134 138 4
135 135 0
MQPM
Target I-MR Chart
Demand forecasting process is not in control.

Possible area for Six Sigma project work!
MQPM
Module Key Learning Points
MQPM
Objectives Review
The participant will be able to:

• Apply SPC rules
• Interpret run and trend patterns in control charts
• Create and interpret
• Xbar-R Charts
• I-MR charts
• Target I-MR charts
MQPM

4-8 SPC - Variables Data

Uploaded by

Document Information

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

4-8 SPC - Variables Data

Uploaded by

Copyright:

Available Formats

Statistical Process Control for Variables Data

By the end of this module, the participant will be able to:

SPC for variable data will:

SPC for variable data is

• Statistical Process Control before six sigma

Statistical Environment Statistical

Materials Process Policies

Statistical Process Control

Random Variation Region

Nonrandom Variation Region

Random Variation Region

Nonrandom Variation Region 99.73%

Control limits are

Control Limits are not specification limits

UCL 1. Any point outside control limits

Western Electric Nelson

Nelson False Alarm Rate

The Nelson tests are designed so that the false alarm

Rule 1: Any point outside control limits

Rule 2: 9 consecutive points on same side of centerline

Rule 3: 6 consecutive points increasing or decreasing

Rule 4: 2 of 3 points in same zone A or beyond

Rule 5: 4 of 5 points in same zone B or beyond

Rule 6: 14 consecutive points alternating up and down

Rule 7: 15 consecutive points in either zone C

Rule 8: 8 points in a row outside zone C, either side

X1 X2 X3 X4 X5 • You measure the revenue per FA

Average 45.13 2.36

A2, D3 and D4 are Shewhart control constants

Shewhart Control Chart Constants

UCLR  D4 R  2.11* 2.355  4.97

Calculated values UCLX  X  A2 R  45.13  0.58 * 2.355  46.49

Stack the data

I-MR: Individual, Moving Range Charts

I-MR Charts are two separate charts of the same data

Note: calculated for a

E2, D3 and D4 are Shewhart control constants

Revenues Shewhart Control Chart Constants

n is the data or moving range subgroup size

UCLMR  D4 R  3.27 *0.270  0.88

UCLX  X  E2 R  2.244  2.66 *0.270  2.96

A process “improvement” has been made to increase

Stack the data

Recalculating limits based upon improved

Looks like the forecasting is in

Actual Target Delta

Demand forecasting process is not in control.

The participant will be able to:

You might also like