Reconstruction analysis-based algorithm to decompose a complex system into subsystems

Abstract

Two previous papers [Mirats et al. (2002a) On the selection of variables for Qualitative Modelling of Dynamical Systems, International Journal of General Systems 31(5) pp. 435–467; Mirats et al. (2002b) Variable selection procedures and efficient suboptimal mask search algorithms in Fuzzy Inductive Reasoning, International Journal of General Systems 31(5), pp. 469–498] were devoted to the selection of a set of variables that can best be used to model (reconstruct) a given output variable, whereby only static relations were analysed. Yet even after reducing the set of variables in this fashion, the number of remaining variables may still be formidable for large-scale systems. The present paper aims at tackling this problem by discovering substructures within the whole set of the system variables. Hence whereas previous research dealt with the problem of model reduction by means of reducing the set of variables to be considered for modelling, the present paper focuses on model structuring as a means to subdivide the overall modelling task into subtasks that are hopefully easier to handle. The second and third sections analyse this problem from a system-theoretic perspective, presenting the reconstruction analysis (RA) methodology, an informational approach to the problem of decomposing a large-scale system into subsystems. The fourth section uses the fuzzy inductive reasoning (FIR) methodology to find a possible structure of a system. The study performed in this paper only considers static relations.