• Read
  • Publish
  • About

Read

Explore current and past TAD issues and related content.

Current Issue

Learn more about our current issue

Past Issues

Browse our compilation of past issues

Extras

Webinars, videos, articles and more

Publish

View submission guidelines, learn more about our review process and find helpful recommendations for publishing work in TAD Journal.

Call for Papers

Submit work for our next issue

Author Guide

Explore editorial tips and recommendations

About

TAD Journal is a peer-­reviewed international journal dedicated to the advancement of scholarship in the field of building technology and its translation, integration, and impact on architecture and design.

Our Mission

Learn more about our vision and values

Editorial Board

Meet the minds bringing our mission to life

Advisory Board

Meet the experts shaping TAD’s future

Issue 9.1

Urban Configurations: Mass Optimization with Real-World Constraints Using High-Performance Computing 

Generative design processes were used in the architecture, engineering, and construction (AEC) industry long before modern machine learning (ML) approaches emerged. For almost three decades, genetic algorithms (GA) have helped generate populations of solutions that comply with design aspirations and performance criteria. Although these systems have proven useful, they are not widely implemented because of the computational power required to generate and run analyses for thousands of design options in a timely fashion. This research focuses on the development of a bespoke optimization system that addresses these issues by using a custom graphics processing unit (GPU) based analytical engine (Cyclops) and a proprietary distributed computing system (Hydra). Together, Cyclops and Hydra enable optimizations to scale up to the urban development level and converge in a fraction of the time it takes commercially available software. In addition, this article will discuss the development of applications that extend this workflow through the visualization and postprocessing of results, the potential of ML methods being integrated into the process, and how deep generative models can benefit from the outputs of this workflow. 

Read Full Article (ACSA Member) Read Full Article (Non-member)