Development of an intelligent self-tuning PID controller

Abstract

This paper presents the design of a practical intelligent self-tuning proportional-integral-derivative (PID) controller. It has a start-up mode where the system is `cold-started' using relay control during which the ultimate gain and period of the process will be computed from the resultant limit cycle. The PID controller is then switched in with PID parameters computed using the Ziegler-Nichols tuning formula. Under normal operation, the controller injects a small pseudo-random-binary-sequence (PRBS) test signal and applies the cross-correlation method to continuously fine tune its PID parameters. The intelligence built into the controller includes heuristic logics to check for instability, in which case the controller will execute a `cold-start'.