Compensation of dominant and variable delay in process systems

A technique is proposed for compensating dominant and variable delay in process systems. The basic structure of the Smith predictor is adopted, while the delay parameter in the process model is continuously tuned on-line by minimizing a performance index function. The average magnitude difference function, the expectation of the absolute difference between process output and process model output, is shown to be a good performance index function. The effectiveness of the proposed method is demonstrated through applications to two typical processes.