Chapter 5
Removing the Product
Variability Barrier with
Statistical Process Controls
While the previous chapter on real-time process control was relatively
straightforward, this section is relatively complex, tedious, and never end-
ing. In fact, getting all of the known hardware problems fixed is merely a
prerequisite for the main event: improving the quality of your product. This
is where statistical process control (SPC) takes over. SPC enables you to dig
much deeper to find the root causes of product variability and is therefore
an essential element of modern manufacturing.
I must preface this section on SPC with an emphatic disclaimer. I am
not a statistician—not even an apprentice statistician. I will only say that I
have been exposed to statistical methods and have seen them work mar-
vels, so you could consider this to be more of a testimonial than a treatise.
I hope my hands-on experiences will enable me to show you some of the
tools and how you might go about applying them. However, when we are
through, if you decide to try some of it in your plant, you will need consid-
erably more knowledge than what is presented here. Dr. Deming provided a
wealth of that knowledge, there for the taking, in Quality, Productivity, and
Competitive Position (QP&CP) (MIT Center for Advanced Engineering Study,
1982). And, of course, there are many otherne resources out there, includ-
ing Appendix 2 from Statit
. Just do not spend all your time reading books
52 ◾  Removing the Barriers to Efcient Manufacturing
instead of rolling up your sleeves and getting to work. Eventually, you may
find you need the services of a professional outside your organization.
Sampled Data versus Continuous Monitoring
To begin, let us establish a fundamental concept that every product or work
in process, whether a physical object or a service, has characteristics that
define its value to the customer. These are generally referred to as product
attributes. And, while there may be some exceptions, generally these attri-
butes cannot be measured on a continuous basis. They can only be sampled
and measured as shown in Figure5.1.
How you perform the sampling and what you do with the results will deter-
mine your success or failure as an enterprise.
In the previous chapter, process variables were measured continuously in
real time and with an accuracy as high as 1%. If we displayed those variables
on strip chart recorders, each variable would be connected to a colored pen
that would begin tracing its value as a continuous timeline. At any point in
time, you could see the pens move almost instantaneously in response to
changes in the process. If a variable is constant, its pen moves little and traces
a straight line. On the other hand, if a variable is noisy, the corresponding
pen will oscillate rapidly, making a much wider line. (This is just for illustra-
tion. Remember, we supposedly fixed all of those in the previous chapter.)
Sample and Measure
Control Chart
Identify and Correct
Assignable Cause
Figure 5.1 The control chart process. (Courtesy of the Data and Analysis Center
for Software [DACS] Gold Processes, http://goldpractice.thedacs.com/practices/spc/

Get Removing the Barriers to Efficient Manufacturing now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.