Identification and Control of Systems Subject to Abrupt Changes

Abstract: Many different recursive identification methods for time varying systems have been suggested in the literature. An assumption that the variations in the system parameters are slow is common for almost all the methods. When using the methods on systems with faster variations one is forced to compromise between alertness to parameter variations on one hand and noise sensitivity on the other. The topic ofthe thesis is to investigate how such a compromise can be avoided fora certain class of systems.The systems considered are such that their dynamic changes between some different typical modes. As an example one can think of the different "flight cases" of a supersonic aircraft. The philosophy behind the approach taken in the thesis is that the observations of such a system can be separated into different sets corresponding to the different states of the system. The parameters ofthe different modes can then be estimated from the separated data sets.Technically, this parallel modelling is achieved by describing the system parameters as the realizations of a Markov-chain. An estimation algorithm for time varying systems based on this parallel modet approach is given in the thesis.The behaviour of the algorithm is analysed and problems connected to it are illustrated through simulations. The analysis and the simulations show that a major problem is the initialization of the algorithm without sufficient a priori information. Based on the analysis a procedure is given that makes it possible to use the algorithm with a minimum of a priori information.It is further shown how this recursive identification algorithm can be utilized for adaptive control. As an illustration of this, the method is used for adaptive control of a mode! of a cold rolling mill.