|
Statistical Process Control
The
objective of statistical process control (SPC) is to
bring the process into statistical control so that we
can determine its capability and breakthroughs can be
identified. This is accomplished primarily by the use
of statistical control charts. A process is said to be
in a state of statistical control when the variation
present in the process is consistent over time. A
process in statistical control is stable. That is, the
distribution of the output is consistent over time and
the output has the same location, spread and shape.
When in statistical control, only “common causes” of
variation are present in the process.
There are many advantages of having a process in
statistical control. The output of the process is
predictable. The capability of the process can be
determined. The cost of “non-quality” can be
predicted. The effect of process changes can be
measured accurately. Output will be at its best given
the current in-control process. Process improvement
will therefore require changing the current process.
Finally, when a process is in statistical control,
common cause variation can be investigated through
designed experiments and typically changes can be made
to reduce the variation in output.
There are two broad categories of causes for variation;
common and special causes. The fundamental purpose of a
control chart is to separate the two. The basic
strategy for variance reduction is to first eliminate
special causes and, hence, bring the process into
statistical control. Second, through use of designed
experiments, reduce the common causes and fundamentally
improve the system. Finally, design the process to
anticipate variation using robust design techniques.
Robust design serves to make the process less responsive
to uncontrolled sources of variation.
In
some cases, production may have a high degree of product
mix and low volume on individual products. High mix/low
volume manufacturing can take an unacceptably long time
for standard control charts to be established on
individual products. An alternative is called “short
run SPC”. The fundamental difference with short run SPC
is that data transformation is used to be able to plot
multiple products on the same control chart. The
advantage of this is faster establishment of the chart
and control of the process.
When
a process is in a state of statistical control, the
process capability can be measured. There are several
commonly accepted metrics for process capability. To be
valid, they all depend on processes which are stable and
predictable over time. Many erroneous assessments of
process capability have been made by not first
establishing statistical process control.
|
