Seismic response of high-rise buildings using long short-term memory intelligent decentralized control system

Abstract

Due to the difficulty in adopting the conventional centralized control method for seismic response control of high-rise buildings with complex structures and large sizes, the decentralized control method is applied. The challenge is that the control model of the high-rise structure divided into multiple subsystems changes as large nonlinear deformation under strong earthquakes. Therefore, with an application of the long short-term memory neural network, the long short-term memory intelligent decentralized control method is proposed for seismic response control of high-rise buildings. On the basis of the decentralized control theory for high-rise buildings, a long short-term memory network deep-learning framework is established to construct different types of decentralized controllers, and to determine the sufficient conditions for the stability of the decentralized controllers using the Lyapunov stability theory. The long short-term memory intelligent decentralized control system of a 20-story benchmark building mode is simulated, and its fault tolerance is studied. The simulation results show that the decentralized control method can reduce the complexity of the structure model by dividing the high-rise building structure into multiple subsystems. Compared with the centralized control method, the long short-term memory intelligent decentralized control method can effectively avoid the overall failure of the control system. The long short-term memory intelligent decentralized control method can still have a satisfactory performance under sensor noise and control devices failures. This verifies that the long short-term memory intelligent decentralized control system has a better fault tolerance and can provide an innovative solution for the decentralized control of high-rise buildings.