LOST IN DESIGN SPACE: INTERPRETING RELATIONS/STRUCTURE BETWEEN DECISIONS AND OBJECTIVES IN ENGINEERING DESIGN
In the early phase design of engineering systems, it has become increasingly popular to use a system model and an optimizer to generate a Pareto Frontier of designs that satisfy certain objectives or criteria (e.g. cost, performance, risk). A decision maker then considers this design space and considers which design appears to best satisfy a stakeholder’s requirements. While the generation of alternatives is relatively straightforward, the stakeholder must still make sense of the potentially thousands of alternatives. Additionally, stakeholders often want to consider the possibility of changing priorities, market segmentation via design families and decision sensitivity for further investigation of system priorities. In practice, stakeholders also have many criteria (5 or more), which can be difficult to visualize or represent simultaneously. As such, we aim to create new tools and techniques to extract or aid in finding this information from large multi-criteria decision problems. We approach this with three techniques. The first, Cityplot, uses a virtual reality representation of the design space by use of dimension reduction to represent both the engineering decisions that create a design and the criteria that a stakeholder would care about. A human study is performed to show the validity of the approach by having human subjects perform basic tests and evaluations of real and synthetic design problems. We find the Cityplot allows the intuitive visualization of decisions and a large number of objectives. The second, MOMS-CFCs, provides an interpretation of a single linkage tree and matching procedure to find rules of thumb regarding system dependencies and decision sensitivities. While promising, a few key issues prevent the approach from being applicable in most practical applications. The third, MAPSA, models the shape of the Pareto frontier with a mean plane which parameterizes the space and a model of the residuals which characterizes the overall shape of the frontier. This enables characterization of the tradeoffs and summarization of key features of the frontier. We show some simple mathematical results and create a rough model of how such an approach can be useful for decision making.
Design Space Exploration; Design Statistics; Engineering Design; Multi-Criteria Decision Making; Multi-Objective Design Analytics; Visualization
Selva Valero, Daniel; Frazier, Peter; Sridharan, Karthik
Ph. D., Mechanical Engineering
Doctor of Philosophy
Attribution 4.0 International
dissertation or thesis
Except where otherwise noted, this item's license is described as Attribution 4.0 International
Showing items related by title, author, creator and subject.
Hiller, Jonathan (2011-05-31)The work presented here contains the first known comprehensive consideration of digital materials. Digital materials rely on a fundamentally new paradigm of manufacturing: Physical objects are composed of many discrete, ...
Characterization of BIB Designs for Constructing Minimal Fractional Combinatorial Treatment Designs Raghavarao, Damaraju; Federer, Walter T.; Cornell University. Biometrics Unit.; Cornell University. Dept. of Biometrics.; Cornell University. Dept. of Biological Statistics and Computational Biology. (2000-06)