불확실성과 외란에 영향을 받는 전기 유압 시스템의 능동 외란 제거 제어에 관한 연구

Abstract

By virtue of extremely high force/torque output capability, smooth motion generation, great durability, and less maintenance requirement compared to electrical actuators, hydraulic actuation systems have been widely employed in heavy-duty industrial applications including hydraulic manipulators in manufacturing factories, hydraulic presses, construction machines, and so on in recent years. As one of the control problems for such hydraulic systems, motion tracking control has become an interesting research topic and attracted great attention from the research community in both academia and industry. Nonetheless, achieving high-accuracy tracking performance of hydraulic actuators is still a major concern that requires further consideration due to their high-order dynamics, nonlinear characteristics, model uncertainties, and unknown time-varying payload. Therefore, several attempts to enhance the trajectory-following ability of the electro-hydraulic systems (EHSs) have been made and presented in this thesis. Firstly, control algorithms based on the nominal system model of the studied EHS, which was offline identified in advance, were proposed. To overcome the problem of velocity measurement shortage, a Levant's differentiator was utilized to precisely attain the angular speed of the actuator based on the position information. Besides, to lower the negative impacts of disturbances resulting from the mismatch between the nominal system model and the actual system dynamics, uncertain nonlinearities, and unknown time-varying payload on the closed-loop system, dual disturbance observers (DOBs) with very few tuning parameters were established. Based on that, improved performance was achieved in comparison with some extended state observer (ESO)-based controllers. Secondly, as an alternative solution to the well-known ESO design, extended sliding mode observers (ESMOs) were first developed to cope with both the non-existence of measurement mechanisms and disturbances. By using the discontinuous function of the error between the measured signal and its estimate, the constructed ESMOs are able to react better against the fast-changing disturbances and estimate immeasurable system states more accurately compared to the ESOs. The stability of the suggested ESMOs was confirmed by using the Lyapunov theory. Subsequently, an ESMO-based controller was developed based on the traditional backstepping framework for an EHSS. Numerical simulations were conducted to demonstrate the effectiveness of the ESMO-based control approaches in comparison with ESO-based controllers. Thirdly, to avoid the time-consuming and painstaking identification process, adaptive neural networks (NNs) were developed to approximate unknown dynamics, the so-called unstructured uncertainties, that are functions of system states and control input in the system model. Different from the existing methods in the literature, in which only partial state-dependent model uncertainties were assumed to be unknown, the problem of completely unknown dynamics were considered in this research. To surmount this obstacle, multiple NNs are utilized for tracking control of systems suffering from completely unknown dynamics. In addition, imperfections of NN-based approximations are solved by using disturbance observers. Based on the sliding mode theory, a composite controller was synthesized to not only increase the robustness of the closed-loop system but also avoid the computational complexity of the standard backstepping control. The boundary layer approach was adopted to effectively decrease the chattering that naturally exists when applying conventional sliding mode controllers. The obtained results indicated the effectiveness of the suggested method and unveiled its applicability to real applications subject to completely unknown dynamics. Finally, in addition to the servo-valve-controlled EHSs, the tracking control for pump-controlled EHSs was also investigated in this study. For the sake of cost-effectiveness, an output feedback robust control was developed based on an ESMO. To address the computational burden of the traditional back-stepping because of the analytic derivative calculation of a virtual control law at each back- stepping iteration, a dynamic surface control (DSC) technique was adopted. Additionally, compared to the existing method in which only mismatched uncertainties in the pressure dynamics were considered, both total matched and mismatched disturbances were sufficiently estimated by the integration of multiple ESOs then they were compensated by feedforward compensation in the synthesized control laws. Subsequently, the composite control approach was introduced and the stability of the recommended method was confirmed through Lyapunov theory. Numerous comparative experiments were conducted to verify the efficacy of the developed controller in comparison with some reference methods.