Research Papers

A Computationally Assisted Methodology for Preference-Guided Conceptual Design

[+] Author and Article Information
Garrett J. Barnum

Department of Mechanical Engineering, Brigham Young University, Provo, UT 84602garrett@burley.com

Christopher A. Mattson1

Department of Mechanical Engineering, Brigham Young University, Provo, UT 84602mattson@byu.edu


Corresponding author.

J. Mech. Des 132(12), 121003 (Nov 23, 2010) (9 pages) doi:10.1115/1.4002838 History: Received June 28, 2010; Revised October 13, 2010; Published November 23, 2010; Online November 23, 2010

We present an interactive, computationally assisted, methodology for capturing and incorporating designer preferences into a numerical search for design concepts. An initial pool of manually created designs is parameterized and used in a computational search that recombines features to form new designs in a semi-automated way. Designs are evaluated quantitatively by performance calculations and evaluated qualitatively by human designers. Designer preference is captured when visual representations of designs are presented to the designer for subjective evaluation. The methodology searches for optimally performing designs, guided by quantitative performance models and designer preferences. The methodology couples the speed of computational searches with the ability of human designers to subjectively evaluate unmodeled objectives. The new methodology is demonstrated with a vehicle architecture example, which generates a set of designs that progressively improves in performance and more fully meets designer preference. The proposed method brings the ability to generate numerous, optimally performing solutions across a wide solution space, in an efficient and human-centered way, and does so in the early stages of design.

Copyright © 2010 by American Society of Mechanical Engineers
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Figure 1

The conceptual design context within which the preference-guided search methodology fits

Grahic Jump Location
Figure 2

Examples of (a) an estimated probability mass function for a discrete variable, (b) a standard kernel density estimate (29), and (c) an estimated probability density function for a continuous variable, used to model and predict a designer’s preference for the variables. Note that in all cases, the horizontal axis represents the gene values, and the vertical axis represents how preferable a gene value is.

Grahic Jump Location
Figure 3

Human-generated concept sketches of vehicles

Grahic Jump Location
Figure 4

A schematic of the vehicle architecture design example

Grahic Jump Location
Figure 5

The progression of the design objectives through the preference-guided search, changing from single objective optimization to multiobjective optimization after loop 3

Grahic Jump Location
Figure 6

Set of nonoptimized vehicle designs automatically formed and presented to human designers for subjective evaluation

Grahic Jump Location
Figure 7

The preference-based models for each design gene, formed from subjective evaluation of a designer. Compare with Fig. 2.

Grahic Jump Location
Figure 8

A set of vehicle designs that has converged using the preference-based models and physics-based models of performance

Grahic Jump Location
Figure 9

New, optimized vehicle designs automatically formed, consisting of new combinations of features and parameter values

Grahic Jump Location
Figure 10

Test data showing the preference model improvements during the learning period (top), and a higher quantity of preferred designs being found (bottom)




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In