Page 50 - Design for Six Sigma a Roadmap for Product Development

28 Chapter Two

and applied to processes. In addition to all quality methods, Six Sigma
uses many existing methods for process modeling and process analy-
sis. These methods include process mapping, value stream mapping,
and process management.

2.3 Process Capability and Six Sigma
2.3.1 Process performance and
process capability
Process performance is a measure of how well a process performs. It is
measured by comparing the actual process performance level versus the
ideal process performance level. For a power supply unit, its perfor-
mance may be measured by its output voltage, and its ideal performance
level could be 6 V. For a customer billing process, the performance could
be measured by the number of errors per month; in this case, the ideal
performance level is “zero error.”
For most processes, performance level is not constant. For example,
a customer billing process may have very little error in some months,
but somewhat more errors in other months. We call this variation the
process variability. If process performance can be measured by a real
number, then the process variability can usually be modeled by normal
distribution, and the degree of variation can be measured by standard
deviation of that normal distribution.
If process performance level is not a constant but a random variable,
we can use process mean and process standard deviation as key per-
formance measures. Mean performance can be calculated by averaging
a large number of performance measurements. For example, in the
customer billing service, we can collect the number of errors per month
for the last 3 years and take the average value, and this is the process
mean performance. For the power supply unit, we can measure a large
number of units and average their output voltages to get a process
mean performance level.
If processes follow the normal probability distribution, a high per-
centage of the process performance measurements will fall between
3 of the process mean, where is the standard deviation. In other
words, approximately 0.27 percent of the measurements would natu-
rally fall outside the 3 limits and the balance of them (approxi-
mately 99.73 percent) would be within the 3 limits. Since the
process limits extend from 3 to 3 , the total spread amounts to
about 6 total variation. This total spread is often used to measure the
range of process variability, also called the process spread.
For any process performance measure, there are usually some perfor-
mance specification limits. For example, if the output voltage of power
supply unit is too high or too low, then it will not function well. Suppose