Abstract:
Auto-tuning of multi-input multi-output (MIMO) controllers of a bulk tailing treatment (BTT) surge tank is presented. Two controllers are selected for optimisation. The first controller is a decentralised proportional-integral (PI) controller that controls a plant simulated using a linear model. The second
controller is a multivariable inverse PI controller that controls a plant simulated using a non-linear process model. Objective functions are designed to promote set point tracking and disturbance rejection. The search domain constraints are determined by intuitively expanding the search domain around the tuning parameters of the reference controller. Results show that Bayesian optimisation is successful in improving the performance of the set point tracking and disturbance rejection controllers for the surge tank process.
