Active control of building structure using lattice probabilistic neural network based on learning algorithm

Abstract

Active control of building structure using lattice probabilistic neural network (LPNN) employing the gradient descent method (GDM) for learning to increase control capability is proposed. With the lattice pattern of the state vector used as the training data, LPNN calculates the control force using only the adjacent information of input, thus, response is greatly faster. Three story building under El Centro earthquake is used to train the LPNN. Northridge earthquake is used to verify the proposed method. In the numerical simulation of the building structure control, the control results of the LPNN are compared with the uncontrolled results. The proposed LPNN algorithm can effectively reduce the response of the building structure under earthquakes.