Evaluation of MPC and PI control on a Tribometer : Design and comparison of control methods for a Pin-on-Disc Tribometer

Abstract: This thesis compares Model Predictive Control (MPC) and Proportional-Integral (PI) control in a pin-on-disc tribometer. The tribometer consists of a geared Direct Current (DC) motor driving a flywheel and a voice coil actuating a pin onto the surface of the flywheel. The contact surface of both pin and disc can be replaced with the desired material. The system is first modeled, and both conntrollers are derived from the models and Matlab and Simulink. A PI-cascade controller is used for the disc while a single PI controller is used for the pin. The PI controller is derived using the pole placement, the MPC is derived using the Simulink MPC-designer.  The control methods are evaluated using the Root Mean Sqaure Error (RMSE), percentage overshoot and rise time of step responses from two cases: The ideal case and the operational case. The ideal case contians minimal disturbance as only one actuator is operating at a time. The operational case intends to depict the future use case of the test rig where both actuators are operating simultaneously.  Statistical analysis is conducted as a three was Analysis of Variance (ANOVA) between controller (MPC/PI), case (ideal/operational) and actuator level (40/80 N or 100/200 rpm). In conclusion the MPC has a faster rise time but a higher RMSE than the PI controller in both the ideal and operational case when controlling the disc application of a tribometer. No significant difference in overshoot is established in the disc application of the tribometer. Considering the pin system of the tribometer, no conclusion can be drawn. The main factor being the large impact of the noise at the sampling frequency. This causes failure of the controllers in the operational case.