Comparison of Identification Techniques for Nonlinear Systems

The problem of identifying a dynamic process from experimentally derived data has received much attention in the technical literature. Classically, a least squares based method is used, having first established a suitable structure, to estimate the parameters of a linear discrete time model. Whilst recognising that such linear models are not suitable for all applications, they have gained widespread acceptance, since, in the main they are used with controller design algorithms which require models of that form.