Design and implementation of a fuzzy supervisor for on-line compensation of nonlinearities: An instability avoidance module

Unstable behavior of a control loop is the most feared situation in control engineering. Since PID controllers are linear in nature, but still used in 90% of control loops in the process industry, their use in nonlinear processes increases the risk of closed-loop unstable response. This paper presents the design of an instability avoidance module for SISO PID control loops to automatically adjust the controller's tuning parameters when undesired oscillatory behavior is monitored. This module is a Fuzzy-based Supervisor composed of a dynamic parameter identification module that operates on-line, and a Fuzzy Inference System (FIS) with fuzzy rules to modify PID tuning. Tests in nonlinear process models are performed, demonstrating that this Fuzzy Supervisor—Type 1 (non-intrusive) provides the PID controller the ability to adapt its tuning to eliminate observed oscillatory behavior.