Date of Graduation
Obtaining information on the dynamic behavior of a system or process for ultimate use in control strategies is essential. Identification of a system in real time is divided into two areas, parametric and non-parametric. In the non-parametric method a model is not assumed for the system, while in the parametric method a prior assumption of a given model form with unknown parameters is necessary. The objectives of this research are: (a) Detail studies of various methods of parametric identification in transfer function form, developing and extending some of these methods and a new method for multivariable discrete-time systems. Many criterions are considered here, such as: (1) direct mean square, generalized least square, instrumental variables, extended matrix, maximum likelihood and extended instrumental methods for parameter estimation; (2) determinant ratio, polynomial test, correlation test, statistical F test, residual error test, final prediction error and deterministic methods for order estimation; (3) correlation analysis, performance measure, frequency approach and convergence methods for time delay estimation. (b) It is desirable to have a computer software package for identification purposes which is interactive and simple for the user. On the other hand, it should be of sufficient quality to yield a good representation of the actual system. An interactive software package for parametric recursive identification of a multi-input multi-output linear discrete-time system in matrix transfer function form in the presence of correlated measurement noise is presented. The algorithm identifies the order(s), unknown parameter(s), and time delay(s) in the subjected system in a single computer run. (c) Application of the identification software package to an actual experimental coal gasification system is demonstrated in two system operation modes. One mode is normal system operation. The second mode is one in which no coal is fed to the plant but all other inputs are established as they would be in normal system operation.
DELAVARI, ALI REZA, "MULTIVARIABLE SYSTEM IDENTIFICATION COAL GASIFICATION PROCESS." (1984). Graduate Theses, Dissertations, and Problem Reports. 8736.