Visual Analytics for Generative Design Exploration |

Visual Analytics for Generative Design Exploration

Jamal van Kastel (Delft University of Technology)

CHALLENGE - The goal of computational design workflow is to improve design performances by enabling exploration of a larger amount of design iterations in the early design phase. The computational design workflow uses generative design processes in conjunction with building performance simulations to generate a large data set of sports hall design alternatives. By analyzing this data set architects are able to make performance-driven design decisions in the early design process. Existing data analytics methods are not accessible for practitioners in the field of architecture, however, because interconnected analysis of quantitative and qualitative building performances is not facilitated.

SOLUTION - This research proposes a highly interactive, intuitive data environment that visualizes high-dimensional data alongside building geometries. The data environment integrates all buildings’ geometries alongside data analytics methods in a single viewport. Through multiple means of interaction users can analyze the design space to explore both qualitative and quantitative performances. modeFRONTIER is used to generate a Self-Organizing Map (SOM) from a data set of design alternatives. SOMs produce a two-dimensional representation of high-dimensional data using a sheet-like nodal network. The SOM nodes’ data produced by modeFRONTIER is used as the basis of a game-like, interactive data environment. 

BENEFITS - modeFRONTIER’s post-processing tools give insight in high-dimensional data sets in an efficient and meaningful manner, which was especially useful during the early development phase of the computational design flow. Multiple data analytics methods have been  to analyze and validate the data sets generated by the computational design workflow.