Research On Robust Adaptive Output Feedback Control For Electro-hydraulic Servo System Of Rolling Mill

The problem of high accuracy hydraulic servo control was investigated quiteintensively in recent years, especially for state feedback control under the assumptionthat all the states are measurable. However, these studies do not provide much attentionto output feedback control. In most cases, there exists unmeasurable state in theelectro-hydraulic servo system (EHSS), such as the speed and acceleration of plunger.Therefore, the research on output feedback control for EHSS is very necessary. Thisresearch originates from the project of National Natural Science Foundation (NNSF) ofHebei Province "Research on adaptive robust control for cold-strip rolling mill AGCsystem driven by EHSS" and NNSF of China "Research on multi-model switch adaptivecontrol for EHSS of rolling mill with input saturation". Focusing on the EHSS of rollingmill, this thesis concerns the development of robust adaptive output-feedback control forthe EHSS with unmeasurable state, uncertain parameters, nonlinearities, unknownexternal load force, input saturation, and delayed measurement. The main contents of thisthesis are as follows:First, focusing on the uncertain parameters and unknown load force in EHSSposition control of rolling mill, a reduced-order unknown input observer based robustoutput-feedback control (UIOROFC) algorithm is proposed. A reduced-order unknowninput observer is first constructed to estimate unmeasurable state and unknownperturbation, in which the perturbation containing unknown external force is regarded asan unknown input. Then, a robust output feedback controller is developed. In comparisonwith traditional observer based robust output-feedback controller (OROFC), the proposedUIOROFC algorithm has better transient performance and smaller steady state error.Second, a high gain observer (HGO) and parameter estimator based adaptive outputfeedback control algorithm is proposed for the EHSS of rolling mill with the assumptionthat the upper bounds of uncertain parameters are known. By choosing the appropriatehigh gain and design parameters, the proposed algorithm can guarantee that all signals ofthe closed loop system are ultimately bounded, the system state and its estimation error converge to a neighbourhood determined by the high gain of the origin. On the otherhand, if the upper bounds of the uncertain parameters are unknown, the EHSS can beregarded as an uncertain nonlinear system with unknown control coefficient andunknown linear growth rate. A dynamic HGO based adaptive output feedback controlalgorithm is proposed for this class of systems. By means of a linear combination, thedynamics of the resulting generalized error does not depend on the control input, and byintroducing a dynamic high gain, the unknown linear growth rate as a difficult issue issolved. It can be proved that all the signals of the closed-loop system are bounded, andthe system state and estimation errors ultimately converge to zero. Moreover, theproposed controller is applied to the EHSS, the effectiveness of the algorithm is validatedby example simulations.Third, an anti-windup robust dynamic output feedback (DOF) control algorithm isproposed to solve the input saturation problem in EHSS. Without consideration of inputsaturation, a robust DOF controller is first designed for a class of uncertain nonlinearsystems. By using Finsler's lemma, a sufficient condition of stability of the closed-loopsystem is transferred into a linear matrix inequality (LMI) condition, and the controllerparameter matrices are obtained by solving the LMI. Then, a static anti-saturation robustDOF controller, based on anti-windup method, is designed. It can be proved that theclosed-loop is robust stable and has robust H∞performance.Finally, an observer based output-feedback robust predictive control algorithm isproposed to solve the delayed output measurement problem in hydraulic automatic gaugecontrol (HAGC) of rolling mill. Through proper transformation, a system withoutdelayed output measurement is obtained, and an observer is constructed to reconstruct thesystem state. Then, a robust predictive controller via output-feedback is developed. Toreduce on-line computational burden, an off/on-line algorithm is adopted. Theeffectiveness of the proposed algorithm is validated by simulations.