Statistical Process Control

A Pragmatic Approach

Stephen Mundwiller

• Publisher: CRC Press
• Year: 2017
• eBook ISBN: 9781351248495
• Edition: 1
• Topic: Business
• Subtopic: Decision Making
• Index: Business

People with minimal math skills, and even those with advanced math skills, have difficulty grasping the intuitive concepts behind Statistical Process Control (SPC). Many practitioners do not understand the concepts behind Control Charts, the differences of out of control and out of specification, and the process variation on Control Charts. This book will explain these concepts by using a simple methodology that will bring a much greater level of understanding to those that use it by providing a detailed description of the method, using common language, real-world examples to illustrate the concept, and instructions on easy implementation.

1         

Why Statistical Process Control?

 

Control without action is simply a hobby.

Kaoru Ishikawa

Why? Because it will still work and it works great when used properly, with appropriate training. Just because something is old, does not mean it no longer provides value. In this Six Sigma world, there are other methods to monitor, control, and improve processes. Since Six Sigma is basically a project management methodology, it can be used to reduce or eliminate issues identified by statistical process control (SPC) charts. Note that as you read this book, I am a pragmatic fanatic when it comes to product or service quality.

SPC does not solve anything. SPC charts provide data, and extremely meaningful data if one understands what they are seeing when they look at the control chart. That data provides a picture of products that have been manufactured or a service that has been provided. At this point, I must state that I am a manufacturing guy. While I’ve consulted in the service industry, approximately 98% of my background is in manufacturing. So this book, the examples, and the war stories provided are from manufacturing scenarios.

Years back, I worked as a quality manager in a high-speed liquid, consumer product manufacturing factory. Line speeds were generally 80–150 bottles per minute depending on the size and the viscosity of the liquid. The lines were mostly automated with a crew of three to five people per line. It was a three-shift operation with occasional weekend work, although not all lines ran on second or third shift. The hourly workers were all trained in SPC, mostly by me. The first thing I did when I arrived at my desk in the morning was to turn on my PC and review the SPC data from the previous second and third shifts. With the SPC software used and essentially any SPC software on the market, one can isolate the data to a production line, a shift, or any time period. Based on the charts, I could predict their shift’s level of quality, their approximate case output, and what mistakes, if any, were made, as well as the level of operator frustration. But, then, I’m highly talented. Although with time and training almost any professional could do the same. I’m just pointing out the value of accurate SPC data with accurate control charts.

During this time, I could identify remedial training needs by evaluating the SPC control charts. An example would be when the primary line leader was on vacation and the backup employee was running the line. Sometimes, it was a case of nerves. Other times, they had clearly forgotten some of their SPC training. While not a catastrophe, I could easily address these special or assignable causes of variation. Later in this book, I will discuss what I call problem variation, which I consider the most important type of variation to be able to recognize.

My SPC training classes started in this organization with the first-shift production lines of which there were about a dozen, as all lines usually ran on first shift. After first shift was trained, I then worked second shift to provide training to the smaller second shift, which only ran a few lines based on demand. Their training was followed by the third shift. This training was a two-year ordeal, as training was not the primary focus of my job. Like many production scenarios, taking hourly direct labor personnel out of producing product is almost unheard of and not something done routinely.

With the first shift trained, there was a noticeable reduction in variation as shown on the SPC control charts. Two main reasons: data entry errors were able to be edited, and I constantly preached to avoid machine or line adjustment unless it was absolutely necessary. If the process was in control and the bottle fill levels, as well as the cap off-torque, met or were close to the minimum, then my policy was to let the line run. More on this later, but as stated, adjusting a process is adding a variance, and things will get worse before they get better.

So when untrained second shift operators, who tended to be younger and male, took over the production line from the more experienced and SPC-trained first‑shift operators, who tended to be female, their first thought was, “I can get out more cases on my shift than she did!” So what did they do? They increased the line speed, which added a variation and was an assignable cause. The filler became out of synchronization, the capper and labeler started jamming, and the case packer would malfunction. So they would have to stop the line many times. Their case output would be about 60% of the first shift. Why? They added variance to the process by making adjustments. Not really their fault, as they did not know any better until I had trained them. I will repeat this multiple times in this book. When making changes to a process, things will always get worse before they improve, albeit sometimes for a short period of time. Once I had second and third shifts trained, these types of competitive issues went away. On some lines, engineering and I had the adjustments locked into a specific setting where no adjustments could be made except by maintenance. Overall, case output went up significantly for the entire operation.

I had other very significant accomplishments at this facility. Eventually, they got into very serious financial trouble. They had to cut staff and as in many instances, quality assurance (QA) is the first to go. Such is life for the hard-working quality manager.

Other special or assignable causes that are seen on the control charts would fall into one of the six categories as listed on a Cause and Effect Diagram or Fishbone Diagram.^∗ These are environment, method, materials, measurement, manpower, and machine. This would provide projects for quality, maintenance, engineering, or sometimes human resources. Without the control charts, it would have been much more difficult to identify these special causes of variation unless there was clear catastrophic failure. One thing that should be noted is that in high-speed manufacturing like this, stopping the production line is detrimental to the shift case output and adds unnecessary variation to the process. Yes, stopping a production line is adding variance to the process.

At other times, we used the SPC control charts to identify more long-term projects to reduce the common cause variation inherent in the process. While I will discuss this more in Chapter 4 on variation, think of it as the background noise of the process. Life is not perfect, manufacturing is not perfect, service is not perfect, and control charts are not perfect in appearance. This is shown on the control chart as the common cause variation.

Another advantage of having an SPC program in place is that it enables one to quantitatively measure the reduction in variation after an improvement is implemented. Again, always keep in mind that when a change or changes are made to a process, that is introducing a new variable. Things will get worse before they get better. This will be shown on the control charts. So don’t panic. Be patient! Again, later in this book I will discuss what I call problem variation.

For example, a new machine is added to a production line to replace an old one that needs excessive maintenance. The new machine is also capable of higher capacity. When the production line is restarted after that new machine has been installed, production will be slower and quality will likely be worse. The new machine has to be “fine-tuned.” It has to be adjusted to be compatible with other pieces of equipment on the production line. Other machines on the production line may need adjustment to be more compatible with the new machine. Then let the production line run. Let things start to synchronize. Don’t panic and start over adjusting and blaming the new machine. Take it slow and easy. Gradually, the process will get better. The control charts will graphically show this slow improvement as a slow reduction in common cause variation. The variation caused by putting in the new machine is special or assignable cause variation. If the common cause variation levels out with no real overall improvement, then the new machine was not the answer to reduce the common cause variation. However, there is still an advantage due to the reduction in maintenance and downtime. All this is seen by using SPC and viewing the control chart. Note again that every time a production line stops and then starts there is an introduction of special cause variation which could be a source of problems.

While an engineer or quality professional may accumulate data and construct a control chart manually, the best means to track the SPC data is with software. Whether the measures are taken manually or automatically, those measures are entered onto an SPC control chart or into an SPC system. Manual SPC charts may be used to evaluate a specific set of data over a brief period of time if SPC software is not available. Clearly, software is the best option for a production scenario. Every time new data is entered, the software does the calculations … we’ll discuss these calculations in Chapter 3. The software will hold tens of thousands of pieces of data. The data present can cover years of production. Or one can isolate and just look at any specific time period. This is very useful to evaluate the differences between shifts and operators, identify improvement projects, or measure the value of improvement projects and the level of improvement.

For all practical purposes, SPC does not work in a real world sense unless SPC software is used. It should be noted that the software will immediately perform all of the various calculations and update the important numerical indices, as well as the control charts. This provides the user, management, and any other interested parties with an immediate picture of the process performance.

Today, we have Six Sigma, Lean, and Lean-Six Sigma as the solution to all things in the quality and business world. I consider Lean or Lean Manufacturing to be a cultural change and Six Sigma to be a project management ­methodology. As I’ve discussed in this chapter and will in more detail in the next, SPC is not new and it is actually rather old in the time frame of the modern manufacturing world. But, it still works and will work extremely well. It can be used in conjunction with modern tools such as Lean and Six Sigma to significantly provide business improvement. While improvements can be made without SPC as I’ve done as a consultant many times, it is quite useful when used together with Lean and Six Sigma or other methodologies. As shown in Chapter 4, SPC can be used to identify projects or processes needing improvement and then when completed, assess the level of improvement that was accomplished. It should be noted that SPC is a monitoring tool. It is something that provides warnings or clear indications when there is a problem that needs to be addressed.

When an organization goes through a cultural change to Lean, it is truly a change in the way an organization functions. There are eight classic wastes identified that need to be eliminated or greatly reduced as an organization goes through this cultural change. Reducing or eliminating waste releases capacity. One of these classic wastes to work on is to eliminate or reduce defects. Consider how many problems defects cause. Scrap, overtime, rework, excessive standard labor hours, excessive engineering hours, excessive quality control hours, and reduced efficiency are the main consequences of defects. While SPC cannot be applied to all activities involved to reduce the defect rate, it can be used to monitor a process to determine if the common cause variation, assignable cause variation, and what I call problem variation is in fact reduced or eliminated.

Today, we truly live in a Six Sigma world. Every organization wants belts. Green Belts, Black Belts, and Master Black Belts are typical quality and engineering job requirements. SPC can definitely and should be used with Six Sigma projects. Whether by manually calculated results and manually ­plotted control charts or by using SPC software, SPC is a great tool to measure the progress and success of a Six Sigma project. Not all Six Sigma projects, but certainly some, can be used with SPC for improving a process or ­business. Unfortunately, not very many Six Sigma professionals use or understand SPC. I’ve seen multiple holders of Six Sigma black belts that could not ­interpret a simple control chart. Remember, SPC in itself does not solve problems. SPC is only used to support a process. While that may be for long-term monitoring and just keeping an eye on things, or to identify projects to work on, or to monitor the progress of an improvement, SPC remains a valuable tool, albeit a rather old one.

2

A Brief History

You cannot inspect quality into a product.

Harold F. Dodge

In 1939, while working for Bell Telephone Laboratories, Inc., Dr. Walter A. Shewhart, PhD, wrote Statistical Method from the Viewpoint of Quality Control.^∗ The book (currently available from Dover Publications) is based on a series of four lectures Shewhart gave on the subject to the Graduate School of the Department of Agriculture. In the preface, Shewhart noted that statistical methods of research had been highly developed in the field of agriculture. He similarly noted that statistical methods of control had been developed by industry for the purpose of attaining economic control of quality of product in mass production. Shewhart stated that it was reasonable to expect that much is to be gained by correlating as far as possible the development of the two types of statistical techniques.

In his book, Shewhart explained the statistical mathematics behind control and the methodology for plotting data on a control chart. The calculations for the process average, the upper and lower control limits were described and demonstrated. He also included such concepts as the minimum and maximum or the range of the data, subgroup average, standard deviation, and the quantity of data.

Contributors to Dr. Shewhart’s book included Dr. W. Edwards Deming, PhD,^† and Harold F. Dodge.^∗ Dr. Deming always referred to control charts as Shewhart charts in reference to the original developer.

For decades, Dr. Demi...

Table of contents

• Cover
• Half-Title
• Series
• Title
• Copyright
• Dedication
• Contents
• Acknowledgements
• About the Author
• 1 Why Statistical Process Control?
• 2 A Brief History
• 3 A Teaching Methodology That Works
• 4 ariation in the Real World
• 5 War Stories
• 6 Now What?
• 7 Problems with Solutions
• 8 If It Ain’t Broke … Don’t Fix It
• Glossary
• Notes
• Index