A Multi-Attribute Fuzzy Frontier Approach for System Architecture Selection under Uncertainty

Yu Ye

Abstract

System architects often face the problem of having to make selections that involve multiple attributes spanning across many domain areas. Multi-Attribute Utility Method (MAU) and the Analytic Hierarchy Process (AHP) thus are commonly adopted by system architects to analyze systems that involve multiple stakeholders. Unfortunately, the assignment of utilities and weights in these approaches are often subjective and contentious, and requires stakeholders from different agencies to concur on a common set of utilities/weights, which can make stakeholders feel unimportant and antagonized. Therefore, this paper presents an alternative approach to facilitate system architecture analysis and decision without subjective scoring of attributes.

This approach starts with a thorough and systematic analysis to formulate the problem through mathematical programming. Next, with the combined use of Fuzzy Pareto Optimality concept and S-D domain filtering technique (Smaling, 2005), the set of feasible designs will be reduced to a smaller set of potential “good” designs, while a reasonable design diversity is maintained. Then, the robustness of each system design is measured by the number of Fuzzy Pareto sets of “good” designs it lies in. Finally, the selection is made after carefully considering both the performance measures and robustness of each design.

The thesis ends with a conclusion that the proposed approach provides the system architects with a cohesive analysis framework that can be applied throughout the system design process. This process involves minimum participation of stakeholders, thus is able to generate results in a relatively object way. In addition, this framework can be continuously evolved as the system moves through the different stages of the development process.