Run length-based control performance monitors: a comparison between two techniques

This work compares two techniques of controller performance monitoring. Both are based on the sequences (or run lengths) of signs of the actuating errors of a single controlled loop variable. One technique defines run length as a state, and transition probabilities between states are then modeled as a Markov chain. The other technique observes the distribution of run lengths. Both techniques were tested with experimental data and found to be equivalent in their ability to detect poor control loop performance.