Massively Parallel Evolutionary Structural Optimization for High Resolution Architecture Design

Abstract

This paper shows how the proper use of massively parallel architectures can increase significantly the tractable resolution of topology optimization problems in the field of architecture and urban design. Evolutionary topology optimization techniques are redefining architectural practice providing structurally sound and aesthetically pleasing architectural designs, which commonly mimic nature’s own evolutionary optimization process. Though these techniques provide architects with a powerful tool to integrate function and form in a synergistic way, the resolution of the models to obtain proper designs may be challenging both in computation and memory consumption terms. This work aims to alleviate these constraints proposing a well-suited strategy for Graphics Processing Unit (GPU) computing. Such a proposal makes use of fine grained assembly-free methods along with multilevel parallelizable preconditioner, which notably increase the tractable resolution of the models. The stages of the evolutionary topology optimization pipeline using GPU are compared to the classically used CPU implementation achieving significant speedups. The proposal is evaluated in high resolution real-world architecture and urban designs.