Visualization Method for High-Dimensional Design Solution Spaces Using Constraint Relationships among Design Variables

Abstract

This study proposes a method to support design problem formulation in the early stages of the design process by visualizing high-dimensional design solution spaces using specific two-dimensional design variable planes. The proposed process consists of five steps: derivation of the design solution space, partitioning the design variable planes, evaluating the partitioned spaces, classifying constraint networks, and displaying constraint network-performance radar charts. Ward's method is used to partition the regions, and constraint networks are employed to evaluate the partitioned spaces. This paper applies the proposed method to a simulation model of a motor for an electric vehicle as a case study and visualizes the high-dimensional response space from the perspective of design characteristics using specific two-dimensional design variables. The proposed method comprehensively examines design features contained within the design solution space and classifies multiple characteristics for easy interpretation by designers. Visualizing the trends in constraint relationships between design variables is expected to reduce the need for rework in later stages of the design process during the early stages of design.