Architecture generation and evaluation are critical points in complex systems design. System architecting starts with the exploration of a set of potential solutions that is progressively focused towards the most promising ones. In theory, these solutions are identified through their potential ability to reach system requirements. However, in early design stages, data supporting design choices are fuzzy and uncertain, making difficult the evaluation a priori of the future system architecture performance. In this paper, we propose a method using Bayesian Networks (BN) and Constraint Satisfaction Problem (CSP) to first generate potential system architectures, and then select the best ones regarding system requirements. The association of these two approaches allows to enhance architecture evaluation through integration of component placement optimization. This approach is demonstrated and implemented in radar antenna architecture generation. In the end, we also discuss some of the limits of the proposed approach as well as future research directions.

This content is only available via PDF.
You do not currently have access to this content.