The Determination of Multivariable Nonlinear Models for Dynamic Systems Using neural Networks

Modelling and identification is one of the major areas of control engineering and the theory and practice of linear system identification is now well established [1] [2]. During the past decade, efforts have been focused on developing coherent and concise methods of nonlinear modelling and identification [3]-[7] and more recently, artificial neural networks have been applied to complex nonliear dynamic systems [8]-[12]. There are two basic components in any system identification problem: determining the model structure and estimating or fitting the model parameters. These two tasks are critically influenced by the kind of model employed. Parameter estimation is relatively straightforward if the model structure is a known priori but this information is rarely available in practice and has to be learnt.....