Implementing Auto-Calculated SPC Limits
Plant Applications offers an option to automatically calculate control limits based on a configurable sample size. The implementation affects four existing SPC-based alarm rules and adds these five additional SPC rules:
-
n out of m points beyond two sigmas from the mean
-
n consecutive points alternating up or down
-
n out of m points beyond one sigma from the mean
-
n consecutive points within +/- one sigma from the mean
-
n consecutive points outside +/- one sigma from the mean
NOTE: The SpecificationSetting Site Parameter applies to values derived from a calculated sigma for existing rules and all new rules except "n consecutive points alternating up or down."
Rules are selected at the General tab in the Alarm Configuration screen as described at Configuring Alarms. For each attached variable, a sampling size can be specified at the Attached Variables tab.
The sampling size is the number of data points desired to calculate the mean, standard deviation (sigma), and control limits for the data set. When the sample size is zero, control limits from a specification are used to evaluate alarms. In this case, a delta of six sigma between the lower control limit (LCL) and upper control limit (UCL) is assumed. The mean is calculated as (LCL + UCL) / 2, and sigma calculation is based on an assumed normal data distribution.
When the sample size is greater than zero (a maximum sample size of 100 is permitted), the mean and sigma are calculated according to three scenarios. In these, control limits are set as three standard deviations (3 sigma) from the mean for process control assuming normally distributed data. The scenarios are:
-
A new or existing product has fewer data points than the specified sampling size. Specification control limits are used if defined to derive the mean and sigma for evaluating alarms.
-
Data collection for a product reaches the sampling size. The mean and sigma are automatically calculated and stored when the sample size is reached and control limits are derived. The mean and sigma are recalculated for each additional data point and based on the sample size. This approach generates a rolling mean and sigma to evaluate alarms.
-
Control limits are manually set by an override action during production. The mean and sigma are derived from the new control limits for alarm evaluation. Once the specified number of data points are collected with a new specification in effect, the rolling calculation of the mean and sigma automatically starts based on sampling size to derive control limits and evaluate alarms.
The General tab in the Alarm Configuration screen provides a Disabled Products feature to specify one or more products for which you wish to disable alarms.