This repository is the outcome of a research collaboration between the Chair of Digital Architectonics and the Chair of Structural Design at ETH Zurich. It was developed by Karla Saldana Ochoa, Patrick Ole Ohlbrock, Pierluigi D'Acunto, and Vahid Moosavi.
This work aims to navigate the enormous space of computer-generated forms, based on subjective (designer) and objective (machine) criteria. Hence, this approach surpasses a pure optimization in which only quantifiable measures are taken into account through a prototypical implementation of a computer-aided workflow in which the designer and the machine coexist and potentially yield in a process in which both can unfold their strengths. In particular, the machine is used to optimize a given task and guide and support designers in the quest for innovative and appropriate solutions for structural concepts and design tasks. In the present work, Machine intelligence in the form of three algorithms (Generation of Equilibrium Forms; Combinatorial Equilibrium Modelling (CEM), Clustering; Self Organizing Map (SOM), and Classification; Gradient Boosting Trees) are combined with the ability of humans to evaluate non-quantifiable aspects in a discursive manner. More specifically, a cyclic, fifth-stage workflow (generation, clustering, evaluation, selection, and regeneration)exemplifies the interaction between machine and human intelligence. Through this repeated interaction, the machine can learn the non-linear correlation of forms and their properties and generate candidates itself. A stadium roof is used as a case study to test the proposed approach. This test indicates that it is possible to combine the strengths of human and machine intelligence already in the conceptual generative design phase.
The repository makes use of built-in functions within the software Wolfram Mathematica (https://www.wolfram.com/mathematica/, accessed 04/2020).
If you use this repository, please refer to the official GitHub repository:
@Misc
{Beyond-typologies-beyond-optimization2020,
author = {Saldana Ochoa, Karla and Ohlbrock, Patrick Ole and D'Acunto, Pierluigi and Moosavi, Vahid},
title = {{Beyond typologies beyond optimization}},
year = {2020},
note = {Release 0.1},
url = { https://github.com/sakarla/Beyond-typologies-beyond-optimization },
}
Karla Saldana Ochoa, Ohlbrock,Patrick Ole Ohlbrock, Pierluigi D′Acunto, Vahid Moosavi: Beyond typologies, beyond optimization, International Journal of Architectural Computing (under review), 2020
Karla Saldana Ochoa, Ohlbrock,Patrick Ole Ohlbrock, Pierluigi D′Acunto, Vahid Moosavi: Beyond typologies, beyond optimization, IASS Form and Force, 2019.
Patrick Ole Ohlbrock, Pierluigi D′Acunto: A Computer-aided Approach to Equilibrium Design based on Graphic Statics and Combinatorial Variations, Computer-Aided Design, Volume 121, 102802, 2020
Lukas Fuhrimann, Vahid Moosavi, Patrick Ole Ohlbrock, Pierluigi D′Acunto: Data-Driven Design: Exploring new Structural Forms using Machine Learning and Graphic Statics, Proceedings of the IASS Symposium 2018 - Creativity in Structural Design, Boston, 2018
To start the code please download the sample data from this link.
https://drive.google.com/file/d/1Dfu_-6bL57p0wbuPLLcBM-Xbt8POtIo1/view?usp=sharing
Then download the files in this repository and save them all in one folder.
Start with the notebook call Part 1_QuantitativeEvaluation_HOS_SOM.nb and continue with Part2_QuilitativeSelection.nb the last file is Part3_RegenerationClassifier.nb
The instructions of use are included inside each notebook.