Product Architectures Generation Under Uncertainty: Comparison Between Two Methods

Complex product architecture definition involves technological and architectural choices in order to reach defined system performances. These choices form a wide combinatorial design space whose complete exploration requires a computational method. The latter is made difficult because of the lack and the fuzziness of data and knowledge in preliminary design. To introduce this type of uncertainty, we have proposed an approach based on Bayesian nets: a Bayesian net architecture generation and clustering method is proposed. However, in recent research, lots of conceptual design problems were addressed with Constraint Satisfaction Problem (CSP). The purpose of this paper is to compare these two methods and advantages and challenges in view to design situations under uncertainty. The comparison consists in modeling a sample problem with both methods. The modeling process of each method is described, providing preliminary highlights on advantages and disadvantages of both methods. Then, the methods are evaluated in terms of modeling capabilities and easiness. The number of generated architectures and the execution time of each simulation are also analyzed with regard to the influence of uncertainty introduction in the models. The final objective is to determine which method seems to be the more appropriate to help designers in finding new innovative designs.