Qualitative system identification with the use of on-line genetic algorithms

Abstract

The major problem in building qualitative models via on-line qualitative system identification is how to filter the spurious constraints that are generated from the qualitative reasoning technique. This paper proposes a solution to this problem by an integration of genetic algorithms (GA) and qualitative reasoning. The paper will demonstrate the use of qualitative reasoning to partition the input quantity space into different subsystems, and implementation of GAs to filter and optimize the predicted constraints. The proposed method is verified by simulated examples that suggest the algorithm converges to the optimal point with high speed.