Abstract: A method for evaluating an alarm system for continuous or semi-continuous processes that gives an indication of the actual loading on the operator. A transformations is applied to a data set consisting of alarms and associated times, resulting in the data being much more amenable to common techniques of statistical quality control. The technique requires only a history of alarm events, and the transformation has a strong intuitive appeal.

Abstract: A method for determining the performance of model predictive controllers requiring only closed-loop data and an estimate of the process deadtime. A numerical measure is calculated which is a comparison of the actual variation of the process to the variation that would be expected if the process were controlled by a model predictive controller with no move suppression and no model mismatch. The measure accounts for the fact that the disturbance model employed in the Model Predictive Controller may be different from the actual process disturbance.