The Study On Vibration Control Of Decentralized Neural Network For Building Structure Under Earthquake

Building is an important infrastructure for human to survive and develop,and its destruction will bring huge disasters and property damage to mankind.At present,China is in an active period of earthquakes,so vibration control of building structures is an important measure to improve the earthquake resistance of buildings and ensure the safety of people's lives and property.However,the traditional structural vibration control algorithm requires an accurate mathematical model,and its real-time control of data processing and control force implementation will produce time-delay,and its adaptability of different structures and seismic action is poor.In the face of the increasing complexity of the architectural structure and the difficulty of earthquake prediction,the traditional active control algorithms are becoming more and more difficult to apply to vibration control of the building structure.What's more,nowadays most of the structural vibration control strategies are limited to centralized control,but for high-rise buildings with multi-degrees of freedom,centralized control has disadvantages such as large calculation cost and lack of reliability.Therefore,in this paper,the vibration control of the decentralized neural network derived from the combination of the decentralized control theory and the neural network theory is studied and numerically simulated,and the research work is mainly has the following aspects:1.Theoretical analysis of decentralized control strategies,neural network control algorithms,and fuzzy neural network control algorithms was conducted.The BP neural network control algorithm,RBF neural network control algorithm,and Takagi-Sugeno fuzzy neural network control algorithm were mainly discussed.A decentralized neural network control scheme derived from the combination of the decentralized control strategy and the above three intelligent control algorithms is proposed.2.Aiming at the problems that are easy to occur in the central control of neural network,the vibration Benchmark model of 20-story building structure is divided into three conditions of 2,4 and 5 substructures,and the BP and RBF neural network controllers are respectively designed for each substructure.Then the numerical simulation of two kinds of decentralized neural network vibration control is carried out by the MATLAB software.The vibration control effect and control force of decentralized control and centralized control of two kinds of neural networks are respectively compared.3.For the different characteristics of BP neural network and RBF neural network,the vibration control rate and energy consumption for the 20-story building structure were compared with the vibration control algorithm of two decentralized neural networks.4.Based on the advantages of the fuzzy neural network,a fuzzy neural network controller is designed for each layer of a 6-story building structure,and numerical simulation and analysis are carried out.Then the control effect of the vibration control of the decentralized fuzzy neural network is analyzed.Secondly,the modal analysis of the 6-story building model is carried out,and the four floors with the largest earthquake response in the first four modes are controlled.Then the control effect and the control force between the decentralized fuzzy neural network and the modal control is compared.The research results in this paper show that the decentralized neural network vibration control method proposed in this paper can effectively reduce the structural seismic response,achieve the same control effect of neural network centralized control,and improve the reliability of the vibration control of the building structure.Because of the simplification of the network structure of the subsystem controller,the neural network training time is shortened and the time-delay is eliminated,so the decentralized vibration control of the neural network is more easily applied to practical project.