Once the control limits have been set, they should be revised only:
1. When there has been a change in the process
Initially, the data is plotted against the existing control limits.
However, if any changes are made to the operation of the process, which may
have a material effect on the outputs, then new limits should be
calculated.
2. When they are trial limits based on data
from the first few samples only
It is often desirable to establish temporary limits as early as
possible, so the data from the first 5 or so subgroups can be used to
establish trial limits. Once sufficient
new data is available (usually about 25 subgroups), the control limits
should be recalculated.
3. When there is reason to suspect that they may
have been distorted by a Special Cause
Once the special cause has been identified and eliminated, new
control limits should be calculated for the process.
4. If the size of an individual sample varies
by more than 20% from the target sample size
Remember that u-charts work with variable sample sizes. Technically,
new limits should be calculated every time the sample size changes.
However, this would mean calculating new limits for almost every sample,
and would become very tiresome. In practice, as long as the sample
sizes don't vary by more than 20% from the target size, the same limits
can be used.
Regular revision of the control limits, such as recalculating them every time a chart is completed, is unnecessary. Furthermore, it may counter-productive, as it could obscure certain behaviours in the process, such as drift in the average, or increased variability.