Smith Predictor Approximation for Industrial Control Applications with Genetic Algorithms

Smith predictors are powerful controllers for time delay systems, but difficult to implement in industrial applications due to the proliferation of the standardized PID (proportional-integral-derivative) controllers in parallel form. This paper offers a reduction method based on genetic algorithms and time response matching that approximates a Smith controller with a PID one. The optimization criteria take into account both controller and closed loop responses, with stability constraints. The case study of a time delay air blower is presented with very good results, including stability margins. The method produces PID-class controllers with behaviour similar to that of formally designed Smith predictors.