Nonlinear model predictive control of a hydraulic actuator
Thesis DisciplineMechanical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
The main objective of this thesis is the development and implementation of a nonlinear optimal controller for a hydraulic positioning system. The controller is able to respond rapidly as well as take care of the changing dynamics within the hydraulic system. The necessary attributes for a hydraulic actuator controller are determined by analysing the problems generally associated with hydraulic drives and reviewing the control methods that have been applied in the past. It is concluded that while significant advancements have been made in disturbance rejection, little effort has been placed on the optimal, or minimum time specifications which are frequently demanded by positioning systems. It is also noted that high perfoffi1ance hydraulic drives are prone to cavitation and a controller must necessarily avoid this. The design of a hydraulic test rig is discussed and a novel valve drive circuit that allows direct digital control is presented. The ability of the rig to demonstrate typical control problems is established by experimental testing. The purpose of the test rig is to aid in the modelling process and for controller testing. Power Bond Graphs are used to model the experimental rig and a comparison between a nonlinear model and experimental data shows good correlation. A linear model is also considered and shown to be ineffective at representing the rig dynamics over a range of inputs. By formulating an idealised model, valuable insights into the dynamic characteristics are obtained and the directional dependent gain of single ended rams explained. The performance capabilities of the hydraulic rig are benchmarked by calculating the minimum time response of the hydraulic system subject to constraints on the actuator pressures, load velocity and position. A number of test cases are examined. The research objectives of high performance and flexible constraint handling make model predictive control (MPC) an ideal approach. Model predictive controllers have been successfully applied within the chemical process industry but their application to robotics is hindered by the excessive computational requirements of the algorithm. Furthermore they are typically linear and so in their present form unsuitable. By simplifying the optimisation procedure involved in the MPC algorithm an implementable, nonlinear version of the controller has been tested. The controller is able to constrain the values of pressure, velocity and position within prescribed boundaries, thus eliminating the need for extra hydraulic components. Moreover, the speed of response is comparable to the theoretical optimum. The work reported in this thesis contributes to the field of hydraulic systems control as it presents a novel, nonlinear optimal controller for a hydraulic positioning system. The controller differs from others reported in the literature in that it allows for the plant nonlinearities and forces the system to operate within prescribed boundaries on the state variables. As will be shown this eliminates the need for extra hydraulic components.