Who knows something about Smith Predictor ?


Thread Starter


Dear Friends,

I'm a brazilian post-graduate student and i'm studying THE SMITH PREDICTOR, so i just like to have some help. If you know any sites or person that could help me in this work, i'd be glad. Thanks in advance.

Eng. Luis Mauro Oquendo Pereira

Rick Jafrate

The 1996 article depicts a rolling mill application. Because the speed of the rolling mill is not constant the delay is not constant so the Smith Predictor must continuously compensate for these variations.

In the steel and aluminum industry we typically translate the algorithm from the time domain to a distance traveled domain. The delay is then measured in feet and inches instead of minutes and seconds. In this way the mill speed is irrelevant (expect for speed=0 when you have to
make sure there are no divide by zeros).

I had to use a Smith Predictor once because my colleague and I failed to recognize that we were solving a distance domain problem versus a time domain problem. Inserting a Smith Predictor was easier, at that point, than re-casting the whole thing as a distance domain control.

If you hove occasion to make a comparison between these two methods please let me know. I would be interested in the results.


Rick Jafrate
A company out of Mississauga, Ontario, Canada formerly known as Entech and recently purchased by Emerson, has done a great deal of work using Smith Predictor to determine process variability and the extent to which it can be controlled versus the need to make process changes such as piping configuration, valve replacements, and so on. The general manager is Ross Elliott and the former founder and Smith Predictor brains of the operation is Bill Bialkowski. You can probaly reach them by first contacting Emerson Process Management.