Generative Structural Design: An Algorithmic Approach to Synthesizing and Optimizing Steel Lateral Systems

Abstract

Mitigating the immense environmental impact of the built environment is an important objective for the architecture, engineering, and construction industries. As initial decisions around layout and configuration have significant effects on the structural efficiency of buildings and are difficult to revise later in the design process, it is essential to provide designers with accurate material quantity and embodied carbon estimates at early design stages. The diversity of architectural expression and complexity of structural calculation has made it challenging to develop a tool that is sufficiently accurate, adaptive, and automated to accomplish this goal.

This thesis presents a methodological and an analytical contribution. A novel generative structural design method is proposed, taking low-fidelity inputs, such as those that might be considered during early-stage design, and outputting a high-fidelity structural model that can be analyzed and iterated. The algorithm is tested on 233 structures drawn from wild and synthetic datasets, and a comparative analysis performed between five lateral system typologies. The findings correspond with the literature, verifying the premium for height proposed by Khan as well as Samyn’s slenderness premium.

The analysis demonstrates the utility of synthetic structural system design for individual building analysis and generates new knowledge about the relative efficiencies of different lateral system typologies at a range of heights. The method evaluates how computational tools, such as design space visualization and topology optimization, may be realistically integrated into generative algorithms. Finally, the rich data produced with generative structural design reveals new ways to visualize, analyze, and understand the ways in which designers’ choices affect the ultimate efficiency and environmental impact of built structures.