On the representation and identification of structure systems

Publisher Website
Google Scholar
Abstract
In a number of recent papers a conceptual framework for structure modelling has been developed by Klir (1975, 1976). It is based upon a hierarchy of epistemological levels of systems. Also an operational procedure for solving the problem of structure identification has been proposed (see Klir, 1976, Klir and Uyttenhove 1976 a, b). It is the purpose of this paper to examine the applicability of information theory, or, rather, constraint analysis to the problem of structure identification. It will be shown that so-called structure candidates may conveniently be represented by means of the concept of total constraint, which is partitioned in terms of constraints existing within elements constituting the structure candidate. Since the decomposition of a structure system occurs in terms of subsystems (elements) which contain sets of variables that are not necessarily disjoint, a description of the decomposition is derived which may be considered as a generalized version of the way of partitioning a system by information theoretic means in terms of disjoint subsystems, which is usually encountered in the literature. For the identification of structure candidates an information theoretic measure of conformation is proposed which is compared with Klir's measure, called distance. Finally, the identification of a six-variable structure system illustrates the applicability of information measures to the identification procedure as proposed by Klir (1976).

This publication has 6 references indexed in Scilit: