MODEL PREDICTIVE CONTROL DESIGN FOR LINEAR MULTIVARIABLE SYSTEMS UNDER CONTROL VALVE STICTION

Product quality and production costs are dependent on optimal control of the process. Product variability and oscillation of the process would indicate poor control. Such is the consequence of nonlinearity in the response of the control valves towards controller instructions. The most common cause of nonlinearities is static friction (stiction) in the mechanical assembly of the valve. The ability of a hybrid model predictive control (MPC) formulation to compensate for stiction was to be tested. The formulation was previously shown to be able to compensate for backlash by solving a mixed integer quadratic programming (MIQP) problem.
Simulation studies were conducted using a model of a paper machine headbox model in Simulink. The hybrid MPC formulation was updated to run on current software versions and the Choudhury stiction model was integrated into the system. Due to errors in the simulation, the ability of hybrid MPC to compensate for stiction finally could not be determined. The errors encountered are documented as well as recommendations to overcome the shortcomings of the simulation.