|
 |
Understanding Statistical Process Control
Donald J. Wheeler
David S. Chambers |
Copyright © 1992 SPC Press, Inc.
5908 Toole Drive Knoxville, Tennessee 37919
(615) 584-5005 Fax (615) 588-9440
ISBN 0-945320-13-2
Contents
Dedication
Table of Contents
Foreword by W. Edwards Deming
Preface to Second Edition
Preface to First Edition
Chapter One - Two Approaches to Variation
1.1 The Engineering Concept of Variation
1.2 The Shewhart Concept of Variation
1.3 Two Ways to Improve a Production Process
1.4 Dr. W. Edwards Deming
1.5 The Two Alternatives
1.6 The Necessity of Control Charts
1.7 The Uses of Shewhart's Charts
Chapter Two - Summarizing Data
2.1 Measures of Location
2.2 Measures of Dispersion
2.3 Histograms
2.4 Stem and Leaf Plots
2.5 Running Records
2.6 Summary
Chapter Three - Shewhart's Control Charts
3.1 The Logic of Control Charts
3.2 Using Subgroups to Monitor the Process
3.3 Average and Range Charts
3.4 Limits for Individual Values
3.5 Other Charts for Subgrouped Data
3.6 Control Charts With Subgroup Size One
3.7 Choice of Scale for Control Charts
3.8 What is a Reasonable Degree of Statistical Control?
3.9 Summary
Chapter Four - The Whys and Wherefores of Control Charts
4.1 Charts Done Right
4.2 Why Three Sigma Limits?
4.3 What if the Data are Not Normally Distributed?
4.4 Myths About Shewhart's Charts
4.5 Four Foundations for Shewhart's Charts
Chapter Five - Using Control Charts Effectively
5.1 Patterns in the Running Record
5.2 Simple Run Tests
5.3 More Complex Run Tests
5.4 Four Rules for Defining a Lack of Control
5.5 Other Patterns in the Running Record
5.6 Rational Subgrouping
5.7 Questions Regarding Data
Chapter Six - Capability, Stability, and World Class Quality
6.1 The Capability of a Stable Process
6.2 Capability Confusion
6.3 Converting Capabilities Into Fraction Nonconforming
6.4 What Can Be Said for Unstable Processes?
6.5 The Hypothetical Capability of an Unstable Process
6.6 Short Term Capability Studies
6.7 World Class Quality
6.8 Summary
Chapter Seven - Using Control Charts for Continual Improvement
7.1 A Flowchart for Using Control Charts
7.2 Continual Improvement
7.3 But Will This Work in North America?
7.4 Summary
Chapter Eight - Setting the Process Aim
8.1 The Difference Between Aim and Consistency
8.2 The Necessity of Process Stability
8.3 Setting the Process Aim Using a Sequence of Values
8.4 Setting the Process Aim Using Multiple Measurements
8.5 Summary
Chapter Nine - Special Topics Concerning Control Charts for Measurements
9.1 Inadequate Measurement Units
9.2 Individual and Moving Range Charts Done Right
9.3 When Should One Use an XmR Chart?
9.4 Control Charts for Moving Averages
9.5 Three-Way Control Charts
9.6 Revising Control Chart Limits
9.7 Updating Control Chart Limits
9.8 Charts for the Subgroup Median and Subgroup Range
9.9 From Where Do Those Control Chart Constants Come?
Chapter Ten - Control Charts for Data Based on Counts
10.1 A Simple Approach for Attribute Data
10.2 Charts for Data Based on Binomial Counts
10.3 Charts for Proportions Based on Binomial Counts
10.4 Problems With Binomial Charts
10.5 Charts for Data Based on Poisson Counts
10.6 Charts for Nonconformities per Unit Area
10.7 Summary
Chapter Eleven - Using Attribute Data Effectively
11.1 Three Characteristics of Attribute Data
11.2 Using Attribute Data Effectively
11.3 Summary
11.4 Afterword
Chapter Twelve - Getting Started
12.1 Flowcharts
12.2 Cause and Effect Diagrams
12.3 Pareto Charts
12.4 Summary
Chapter Thirteen - Further Topics
13.1 Interpreting Skewness and Kurtosis
13.2 Enumerative Studies versus Analytic Studies
13.3 The Characterization of Product
13.4 The Problem of Modified Control Limits
13.5 The Fallacy of Acceptance Sampling
13.6 Estimating the Fraction Nonconforming
13.7 The Transformation of Data
13.8 The Effect of Variation on Balanced Systems
Appendix
Glossaries of Terms and Symbols
Lists of Examples and Figures
Bibliography
Answers to Exercises
Tables
Index
Foreword
by W. Edwards Deming
It is an honour to have the privilege to write a foreword for this book by my friend Dr. Donald J. Wheeler. The reader may wish to remember that Dr. Shewhart perceived two kinds of variation.
- Variation from constant causes, the same causes from hour to hour, lot to lot, worker to worker. Dr. Shewhart's term was chance causes.
- Variation from a special cause.
How did the problem arise? The management of the Western Electric Company, the Hawthorne Plant, Chicago, sought to achieve uniformity, so that a telephone company that bought their product could depend upon it. The aim was noble. Their methods, though, were folly. They took action, made some kind of a change at every sign of departure from uniformity. They were smart enough and honest enough to observe that their actions only made this worse. They sought help. The problem went to Dr. Shewhart in the Bell Telephone Laboratories, 463 West Street, New York, newly formed.
There are obviously two kinds of mistake to make in efforts to achieve uniformity.
Mistake 1. Attribute an outcome to a special cause of variation when actually it came from common causes of variation.
Mistake 2. Attribute an outcome to common causes of variation when actually it came from a special cause.
Both mistakes are costly. Anyone may set for himself a perfect record from this hour henceforth, never to make Mistake 1. Simple: attribute any outcome to common causes. In doing this, though, he will maximize his loss from Mistake 2. Likewise, anyone may set for himself a perfect record from this hour henceforth, never to make Mistake 2. Simple: attribute any outcome to a special cause. In doing this, though, he will maximize his loss from Mistake 1.
It would be good never to make Mistake 1 and never to make Mistake 2. This unfortunately is impossible.
Dr. Shewhart settled on a different aim: make Mistake 1 now and then. Make mistake 2 now and then, but regulate the frequencies of the two mistakes to achieve minimum economic loss from both mistakes. To this end, he gave to the world the control chart, with 3-sigma limits. The control chart does a marvelous job under a wealth of applications. It works.
Statistical control may be achieved by hunting down and identifying each special cause as a point goes outside the control limits, and taking appropriate action.
I need not emphasize here the advantage of having a process in statistical control. Costs are predictable with a high degree of belief. Limits of variation are predictable.
In closing, it is fitting to add my deep appreciation for the mathematical achievements of Dr. Wheeler. His understanding of theory, and his application, are guided by mathematical knowledge. It has been my privilege to learn from him.
W. Edwards Deming
Washington 8 June 1992 |
