PATTERN-BASED ADAPTIVE PROCESS-CONTROL

A controller that has an adaptation strategy based on the patterns exhibited in the recent history of the controller error and manipulated input variable is detailed. The method of approach is to analyze all controller error transients to extract features including overshoot and damping. The deviation of these features from desired values is then used to update the parameters of an internal controller model. When the error transient is caused by a disturbance, the patterns exhibited in the manipulated input variable are also considered to determine whether the disturbance has actually changed the character of the process, or if the error patterns are simply reflecting the nature of the disturbance itself. The strategy is general to a number of model-based controllers and is demonstrated here using both a long range predictive controller and a PI controller with Smith predictor. Details of the strategy are presented along with a demonstration of the method on process simulations constructed to present challenging applications.