| Summary: | The work in this thesis is mainly concerned with the practical aspects of the identification package of nonlinear system. A survey of papers published by several researchers on various aspects of nonlinear systems such as linearization, equivalent circuits and the classical solution of nonlinear systems is presented. Various NARMAX models are introduced. State representations, threshold, affine and polynomial models are introduced. The focus of attention here is on polynomial representations of single input and single output forms of the NARMAX model. Various methods of structural detection are discussed and in particular the correlation test is discussed in great detail. Theprocedure developed by Brilliant and George for the correlation test is employed in this work. Model and signal validations are dealt with and, being the most important aspect of nonlinear identification, a great detail of mathematical theory is given. Six cases of model validations are given. Various essential data transformations and memory management problems in a PC based implementation identification work are discussed, and three methods of data transformation are given. Software aspect of Random Memory Management are given and the associated computer source code is provided. The important aspect of the user-computer interaction in modeling is discussed, and software aspects of windowing interactrion are discussed. This in-house developed nonlinear identification software facility puts most emphasis on the use this of this tool as a modeling aide.
|
|---|