Research On Data-driven Active Vibration Control

Whether it is passive vibration control or active vibration control,the control of vibration has always been the focus of scholars.In vibration control,it is often a difficult problem to build the accurate model of the complex system.However,traditional control methods rely heavily on accurate system models,and the model has a great influence on control effect.The application of traditional control methods in vibration control has significant limitations.With the development and application of technologies such as artificial intelligence,machine learning and big data,some novel ideas have been brought to the research of vibration control.For example,the controller design is based on the collected system response data which can avoid the influence of modeling errors.In this paper,the data-driven adaptive dynamic programming algorithm is taken as the main line.Taking high dimensions liquid rockets,airfoil with actuator input constraints and high-rise building structures under seismic excitation as applicaion objects,the data-driven vibration control studies are carried out on these typical vibration phenomena.The scope of application of data-driven ideas has been expanded in the field of vibration control.According to these three practical vibration problems,the data-driven adaptive dynamic programming algorithm is improved and adapted.The superiority and applicability of this data-driven control method is verified by the simulation and experimental results of the active controller.The work of this paper is summarized as follows:(1)The adaptive dynamic programming method is reviewed and three data-driven adaptive dynamic programming control design mehods are proposed for three types problems in vibration control.These three types of problems are the vibration control of discrete system,the vibration control of nonlinear system with input constraints,and the vibration control of systems subject to external disturbances.(2)Active suppression of the Pogo vibration of the liquid rocket has been carried out.Traditional liquid rocket Pogo vibration research is limited to passive control by adding pressure accumulators,and there are few active control researches.In this paper,a two-propellant and single-engine liquid rocket is studied,five independent unit models are assembled to establish a propulsion system model,and then combined with the structural model of the liquid rocket to obtain a Pogo analysis model.The active controller was designed for the reduced Pogo model,and the effect of the variation of three typical control parameters on the controller effect was investigated.The simulation analysis illustrates that the data-driven active control algorithm has a good control effect,which can avoid the Pogo phenomenon and ensure the safe operation of the liquid rocket during the launch process.(3)The flutter control problem of a typical airfoil wing section is studied.Different from the previous passive control methods that change the structural parameters of the wing section or increase the mass,in this study,the control surface deflection of the trailing edge of the wing section is used to actively suppress the flutter of the wing.With considering the input constraints of the actuator,a novel data-driven adaptive dynamic programming algorithm is proposed to design the active flutter controller.By introducing a non-quadratic term in the cost function,the optimla controller is still obtained with the actuator input constraints.The controller is compared with the linear quadratic control controller,and the simulation results show that the data-driven active control method has better control effect.By designing the flutter active controller based on the vibration response data of the system,it not only can get rid of the modeling and avoid the influence of modeling errors on the controller,but also has better prospects for practical applications.(4)Active vibration control of high-rise building structures under seismic excitation has been investigated.Obviously different from the mass tuned damper,the vibration control is performed by adding a mass damper that can be actively adjusted on the top of the structure.In this paper,the control problem subject to external perturbations is converted to a zero-sum game strategy solution,and a novel data-driven off-policy adaptive dynamic programming algorithm is proposed to find the optimal solution of the zero-sum game problem.Based on the collected vibration response of the high-rise building structure system under external excitation,a controller that can effectively suppress the structural vibration is designed without relying on the system model information.The damping effect of the controller is verified by the active masstuned damping experimental equipment produced by Quanser,and the experimental results show that the data-driven controller is very effective for active vibration damping of the structure subject to different seismic excitaitons.The research results of this paper are an inspiration and a theoretical guide to introduce the idea of data-driven controller design into the active vibration control of mechanical systems.