|
Contents
| Quality Encyclopedia
| Discussion Blogs
Short Run Techniques
See
also:
Defining
Control Limits
Short Run analysis combines
data from several runs into a single analysis.
Short Run is typically used
to analyze processes with an insufficient amount of data available from
a given product or service classification to adequately define the characteristics
of the process. In manufacturing, for instance, you may only produce thirty
units of a given part number, then reset the machine for a different part
number. Although the process is fundamentally the same (if it is acted upon
by the same causal system), the first part number may be one inch in diameter,
plus or minus 1/8 inch, and the second part number five inches in diameter,
plus or minus 1/8 inch. This difference in nominal size prevents you from
charting the raw measurements from the different part numbers on the same
chart. Likewise, in a service application, the amount of time to resolve
a customer complaint may be influenced by the type of complaint, such as
one day for incorrect item shipped versus five days for incorrect billing.
In either case, we are interested in statistically significant changes to
our system, relative to either a nominal value (which we define) or an average
value (which the system defines).
Thus, if we assume that the
process is influenced by a common set of causes, regardless of the run (i.e.
part number, complaint type, etc.), then we could use a single control chart
to define the operating level for all runs. In order to do this, we must
standardize each observation based on the properties of its run. Standardization
may be performed a number of ways, as explained below (Montgomery, 1991;
Pyzdek, 1992a).
Nominal Control Charts:
These charts are created by simply subtracting the Nominal value of the
run from the observation. Here, the nominal value is usually the midpoint
of the specification limits, the target value, or the historical mean (average)
value observed from past studies. However, the Nominal charting method MUST
only be used if it can be safely assumed that each run has the same amount
of variation. This method of standardization is useful for any subgroup
size, and subgroup size may vary.
Stabilized Control Charts:
As mentioned above, the Nominal control chart is only valid when each run
has the same amount of variation. In manufacturing, even though two parts
are produced by the same process, the effects of different raw material
may increase the process variation for one part. Likewise, the variation
in time to resolve a billing complaint may be much larger than the incorrect
shipment, since more departments may be involved. When the level of variation
is not similar for all runs, then we must also standardize relative to the
variance. This is done in much the same manner as standardizing the mean.
|
