Study On The Application Of Neural Network In Semi-active Control Of Civil Engineering Structure

Earthquake can constitute a extremely serious threat to security of human life and property. Structural vibration control of civil engineering, especially semi-active control, can effectively reduce the building structural destruction caused by earthquake. Neural network, which has the abilities of self-learning, self-adaption and parallel distributed processing, especially radial basis function(RBF) neural network, which has the characteristic of fast convergence speed and the high accuracy, can be applied to semi-active control of civil engineering structure and be of importance to earthquake mitigation.In this paper, radial basis function neural network is applied to semi-active control of civil engineering structure, neural network controller is designed and the earthquake mitigation effect of controller is simulated and compared.Firstly, according to characteristic of earthquake wave, simulated earthquake wave is chosen and used for simulation. Simulink model of building structure without control is builded, MATLAB is used to simulate building structure without control under simulated earthquake wave and seismic response of structure is then got. Secondly, seismic response of building structure in the next moment for time-delay is predicted using RBF neural network and compared with original seismic response. The results show that RBF neural network can predict seismic response of building structure effectively and its learning convergence speed is fast. Thirdly, MR damper is set in the building structure, building structure with MR damper is builded, RBF neural network semi-active controller is then designed and its control effect is simulted and compared with seismic response of building structure without control. The results show that RBF neural network semi-active control can effectively reduce the structural response of displacement, velocity and acceleration and has a very good effect of earthquake mitigation. Therefore, neural network control is an effective semi-active control method.