Characterization of Dynamical Systems Via Syntactic Pattern Recognition

In many identification problems, the characterization of the system, i.e. the determination of the nature of the model from some specified response, is an important point. For an automatic characterization, a syntactic pattern recognition approach is related in this paper, which is based on the use of regular grammars and of a new grammatical inference algorithm. Linear dynamical systems with known impulse response are considered here, especially eight classes of models which are commonly encountered in the modelling of actual flow in chemical or biochemical reactors. The good results obtained for this application indicate the validity of this approach.