Despite years of research efforts developing methods and decision support tools, architecting complex engineered systems remains a challenging task. Improvements in computational power and optimization algorithms have made it possible to explore large design spaces, but making sense of such datasets is difficult due to their scale and complexity. Various knowledge discovery tools and data-driven methods have been developed in the past to help system designers analyze and make use of such complex data. However, most of the currently available tools do not fully exploit the data that is generated during design space exploration and instead consider the mapping between design decisions (inputs) and objectives (outputs) as a blackbox function. In this paper, we introduce a new method that utilizes not only the design inputs and outputs, but also intermediate variables that are generated during the evaluation of each design. The tool stores all intermediate variables in a database, and then feeds them into a data mining algorithm to extract useful and human understandable features in the form of if-then rules. We show how the use of intermediate variables leads to new insights that could not be discovered with the blackbox approach, and improved knowledge discovery in the sense of features that are more compact and/or with higher predictive power. The method is demonstrated on a real-world system architecting problem of a constellation of Earth observing satellites.

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