SENSITIVITY OF MIQP-BASED MPC DESIGN IN COMPENSATING ACTUATOR BACI(LASH UNDER LOCATTON UNCERTAINTY

The novel method of Mixed-Integer Quadratic Programming (MIQP)-based design of
Model Predictive Controller (MPC) that simultaneously addressesa ctuator saturation and
backlash have shown considerable improvement in the closed-loop performance in
compensating backlash. However, implementing the controller in a multi-variable system
may introduce some problems if the acfual backlash location is unknown to the MIQPbased
MPC controller. In this paper, the sensitivity of the controller is investigated for
uncertainty in the backlash location. Simulation study using MATLAB/SIMULINK
interfacing with GAMS solver for a model developed using an industrial FCCU case
study shows that the MIQP-based MPC is unable to retain its good performance when
wrong location is detected. Also, over-compensation gives sluggish perfonnance which is
extremelyu ndesirable.