Abstract
A basic analysis is carried out on the identification, prediction and control on a single degree of freedom system. First, the responses of the system excited by an active control device installed on the system are effectively utilized to identify the dynamic properties of the system which is modeled by a multi variate ARMA model. Then, general modes of a instantaneous optimal prediction control rule are formulated in terms of the identified components of the coefficient matrix ofthe ARMA model and the weights included in the control objective function. Based onthe formulation, a neural network is derived whose links have physically meaningful weights.
