Multivariate Interactive Visualization of Data in Generative Design

Abstract

In this paper we describe our work on providing support for design decision making in generative design systems producing large quantities of simulation data. The work is motivated by the continuing challenge of making sense of large design and simulation result datasets. Our approach is to provide methods and tools for multivariate interactive data visualization of the generated designs and simulation results. These enable designers to focus not only on high-performing results but also to examine suboptimal designs’ attributes and outcomes so as to discover relationships giving greater insight to design performance and facilitating guidance of further design generation. We illustrate this with an example exploring building massing and envelope design (fenestration arrangement and external shading) with simulations of daylighting and heat gain. We conclude that the visualization techniques investigated are potentially useful in helping designers to better comprehend the inter-relationships between variable parameters, constraints and outcomes, with consequent benefits in finding good design outcomes as well as in verifying that simulation results are reliable and understanding characteristics of the fitness landscape.