Prediction Control of SDOF System
Abstract
General discussions are presented on an instantaneous optimal prediction control, which includes a series of identification, prediction, and control on a single degree of freedom (SDOF) system. First, a method for the identification of the dynamic properties of the system, which is modeled by a multivariate autoregressive moving average (ARMA) model, is investigated with the responses of the system excited by an active control device. Then general modes of an instantaneous optimal prediction control rule are formulated in terms of the identified components of the coefficient matrix of the ARMA model and the weights included in the control objective function. The prediction control rule is interpreted as an equivalent neural-network model whose links have physically meaningful weights.