Study On Data-driven Motion Control Strategies For AC Servo System

High-speed and high-precision motion control strategy of the AC(Alternating Current)servo system is a key technique and research focus in the industrial automation field.Its performance will directly determine the operating efficiency and control accuracy of the mechatronics equipment.However,the dynamic tracking and stability are influenced by the structural nonlinear and fractional-order characteristics,model uncertainties,time-varying features and various disturbances.Considering the existence of unmodeled dynamics and modelling errors,the traditional model-based control strategies are difficult to meet the practical requirements of the AC servo system.Combining with the National Natural Science Foundation,data-driven control theory is adopted to improve the control performance of the AC servo system.The research concentrates on the controller structure,the anti-disturbance ability and the constraint indexes.To provide theoretical guidance and experimental verification for the AC servo system,data-driven motion control strategies are proposed with respect to the fractional-order controller,disturbance compensation and multiple performance constraints.The main content and innovations are as follows:The vector control principle and velocity loop structure are analyzed for the AC servo system.After the limitation analysis of the traditional models,a motion control strategy route is designed employing data-driven virtual reference feedback tuning method under unknown system model information.Next,the performance evaluation indexes are studied including step response,deviation integral and statistical criterion.The data-driven fractional-order motion control strategy is introduced,where the traditional integer-order control structure is replaced by the fractional-order controller.Then,the controller tuning problem is derived using the received input/output data,followed by the devised data-driven off-line optimization algorithm to tune the fractional order and control parameters.Regarding the time-consuming and poor convergence issues,an improved Just-In-Time-Leaning-based technique is proposed for the controller online adjustment,which utilizes the similarity criterion to limit the database capacity.Therefore,the corresponding data containing the current system dynamics can be used to update the control parameters with shortening calculation time and faster convergence.Moreover,the above-mentioned method can be applied to tune both the integer-order controller and fractional-order controller.For the data-driven motion control strategy with disturbance rejection,the estimation of the data disturbances is incorporated into the controller design to improve the anti-disturbance ability and system robustness.The analysis of the data disturbances is conducted so that the expression of the input/output data is established.A data-driven re-weighted iterative motion control algorithm is proposed integrating feedback control and unbiased disturbance compensation.To this end,the information of data dropouts and noises are estimated periodicity by means of tuning residual kurtosis-based iterative learning,which ensures the optimal gain scheduling of the designed controller.The convergence and closed-loop stability are guaranteed theoretically.Simulation results verify that the system robustness is greatly increased under data disturbances.Aiming at enhancing the comprehensive performance,a data-driven motion control strategy is proposed with multiple performance indexes,which consist of the following aspects.By exploiting the frequency response data,the analytical relationships between the control parameters and the performance constraints are derived,including the stability and frequency-domain evaluation.Then,a closed-loop Bode ideal transfer function is selected as the fractional-order reference model to supply time-domain constraint and sensitivity function.Moreover,the amplitude of the input signal is limited smoothly based on the natural constant.Thus,the data-driven model reference adaptive control algorithm can be constructed while the fractional-order controller is designed by variable-step particle swarm optimization algorithm.Simulation results show that the controlled system behaves closely as the reference model achieving optimal control performance.Finally,the experimental test scheme is put forward.Both software and hardware design are carried out on the developed bus-type AC servo drive.Then,a six-joint industrial robot system is built,as well as a flexible swing arm system and a two-inertia elastic AC servo system.The presented data-driven motion control strategies are implemented on these platforms.The experimental results demonstrate the practicability,effectiveness and superiority of the proposed methods.