Practical issues in non-linear system identification

In this thesis, the use of chemical engineering principles in characterization of the industrial chemical processes is explored. Utilizing the black box methodologies, these tools have been used for applications related to process modeling and soft sensor estimation when exact knowledge of the system is unavailable. In certain situations, these methods have provided a robust alternative to the complex physical modeling approach. Besides describing the theory, the applications of these techniques have also been investigated.; A number of issues in experimental design including designing an appropriate input sequence are studied. Extensive simulations and experiments are used to establish the superiority of multi level uniformly distributed input signal over other popular excitation signals.; The model on demand approach has been employed in identification of the pilot scale CSTR reactor. The domain of applicability of this approach has been extended to a multi rate system. The simulations and laboratory experiments are included wherever appropriate. (Abstract shortened by UMI.)...