Adaptive Reconfiguration of Complex System Architecture

In this paper, an approach for adaptive reconfiguration of architecture of a complex system using computational intelligence techniques is proposed. This paper establishes that the simulated change in the relative importance of rules reflecting the significance of customer's key performance attributes could be used to affect architectural evolution. The proposed approach was demonstrated on a sample system. It was also extended to a general system. The evolving system architecture gave the system an adaptive feature in the sense that it accepted alternative components based on the simulated changes in the environment. Architecture alternatives were generated through genetic algorithms (GA), while fuzzy logic was used to determine the fittest architectures based on the ambiguous multi-dimensional performance attributes provided by the customer modeled in quality function deployment (QFD). The weights of fuzzy associative memory (FAM) rules were randomly changed to simulate results of environmental effects and demonstrate how the system would self-adapt.