What is Training Algorithm

For the active control of three dimensional (3D) irregular structures subjected to seismic load, a new nonlinear model is discussed in this chapter. As well as geometric nonlinearity, material nonlinearity is also considered with a neuro-controller training algorithm, which is applied to a structure of multi degrees of freedom. For the control model, a dynamic assembly of two different motions is considered such as coupling between torsional and lateral responses of the structure and interaction between the structural system and the actuator. The training algorithm and the proposed control system of the structure are evaluated by the response simulation of the structure under the excitation of El-Centro 1940 earthquake. With linear and nonlinear stiffness, a 3D three story building structure is controlled by a trained neural network as an example. As additional parameters for the simulation control time delay, the incident angle of earthquakes is considered. The results show that the proposed control algorithm is efficient in the control of the structural vibration.