Construction of Composite Models from Large Data-Sets

Abstract: Most processes of realistic complexity cannot be described by simple linear relationships. In this thesis we describe a method of how to construct composite models from observed data. It is assumed that the dynamics of the process changes with som 'operating-point vector', which is assumed to be a measurable quantity. Based on input-output measurements, and measurements of the operating-point vector a composite model is constructed, which consists of piece-wise linear models. The dynamics of the different linear models are determined from the data, as well as the boundares in the operating point space which determine the dependence of the dynamics on the operating point. The basic idea is to utilize a method for recursive identification, which is able to track slow and as well as rapid dynamic changes. A classification procedure is then applied to the models produced by this identification procedure, and finally borders are created between the different classified models. Techniques for supervised patters recognition are used for the latter step. The whole construction procedure is illustrated with a number of examples.