The Study On Vibration Control Of High-rise Structure Based On Finite State Feedback Genetic BP Algorithm

In the vibration control strategy of high-rise buildings,some classical control methods with good control effect need to arrange sensors with full degrees of freedom on all floors of the building structure to feedback information.From the perspective of engineering application and promotion of the control method,it is not only economical to arrange sensors on all floors of a high-rise building,but also may not have installation conditions on some floors of the building.Structural vibration decentralized control system has stronger reliability and robustness than centralized control,but the problems of information sharing and control coordination among subsystems lead to complex algorithm design,and the overall control of structural dynamic response by decentralized control The effect is not as good as centralized control.In view of the above problems,this paper selects the Benchmark model of the third-generation 20-story steel structure of the international vibration control general platform as the research object,and carries out the following research:1.The 6-layer feedback linear quadratic regulator(LQR)algorithm is proposed based on the finite independence method.On the 20-story Benchmark model,the whole floor layout of sensors is optimized to only 6 floor layouts(each floor sensor only feeds back the displacement and velocity information of the floor).Taking the control effect of the full state feedback LQR algorithm as the standard,the numerical calculation results verify that The control method based only on the effective state feedback of 6 floors has a good control effect on the inter-story displacement and acceleration response.2.The genetic BP(Back Propagation,BP)algorithm controller with effective state feedback is established.The algorithm controller is applied to the 20-story Benchmark model.On the premise that the number of sensors is 6,the floor positions of the sensors are optimized,and the BP neural network that can output the control force of the whole floor based on the effective state feedback is selected.The vibration control rate and energy consumption of 6-layer feedback LQR algorithm and genetic BP algorithm are analyzed.The results show that: based on the premise that the maximum output control force of each floor of the building is equal,the genetic BP algorithm is better than the 6-story feedback LQR algorithm in controlling the peak displacement and acceleration between floors.3.The influence of the parameters of the genetic BP algorithm(the number of floors with sensors and feedback information)on the control effect under centralized control is studied.Taking the control effect of the 6-layer feedback LQR algorithm as the standard:One is to not change the feedback information,only the number of sensor layout floors is changed from 6 to 4 and 2;the second is to not change the number of sensor layout floors(maintain 6),and consider the feedback information as "displacement + velocity + seismic acceleration","displacement" and "velocity" three working conditions.The research shows that the control effect of the genetic BP algorithm will be weakened with the decrease of the number of floors where the sensors are arranged,and the feedback information on the basis of displacement and velocity plus seismic acceleration has little effect on the control effect,but if the velocity or displacement information is missing,it is significantly reduced.4.The application of finite state feedback genetic BP algorithm in decentralized control is studied.First,the 20-story Benchmark model is divided into three working conditions: 2,4 and 5 sub-structures.Referring to the preset number of sensor layout floors when the genetic BP algorithm is used under the centralized control,the subsystem sensors of the three decentralized working conditions are analyzed.The number of layout floors is taken as 5,4 and 3(each floor sensor only feeds back the displacement and speed information of this floor).Then,the genetic BP algorithm is used to optimize the sensor arrangement position,and the BP neural network required for each subsystem is selected,and the control effect of the three working conditions is compared with the total feedback LQR control effect under centralized control.The research results show that: in decentralized control,the finite state genetic BP algorithm proposed in this paper can not only optimize the sensor layout of all subsystems and determine the BP neural network required by each subsystem,but also the system control effect is close to centralized control.