Neural networks for the identification of linear dynamical model of a five story building

Presents the results of utilizing neural networks to provide an efficient computational model for a dynamical system. Neural networks are used for parameter identification of multistorey buildings. The networks were trained and tested using experimental data measured on building models. The measurements consist of acceleration time-histories taken at the base of the buildings and at the various floor levels. It is demonstrated that, once the initial learning phase is completed, the network can provide instantaneous identification of system parameters when presented with different acceleration records