Optimization methods are increasingly popular in architecture, but they are not always used correctly. Formulate what the optimization problem is first and only then look at how to solve it, says Ding Yang. He developed a Multi-Objective and Multi-Disciplinary Optimization (MOMDO) method that highlights the reformulation of optimization problems.
Yang got the idea for his research through his experiences in designing indoor sports halls. These are relatively complex buildings that must meet various performance requirements simultaneously. For example, indoor sports halls usually have skylights for daylighting and have large column-free interior spaces for hosting sports events. This poses both climate and structural challenges for the designers. They must take into account the availability of daylight, the risk of overheating and glare and the efficient use of steel. ‘It's difficult to balance all these conflicting requirements,’ Yang says. ‘Therefore, it is smart to first collect and analyze the performance data and gain relevant information and knowledge. Based on that, it is possible to formulate a reliable optimization problem and hence arrive at optimal designs.’
Yang's research is a response to the increasing use in architecture of computer-aided design methods. These methods often result in a spectacular display of technical possibilities, but not necessarily in better designs, he observes. In fact, they may be abused, making things more complicated. According to him it's better to use computer-aided design optimization methods in a more reasonable way.
Multi-Objective and Multi-Disciplinary Optimization (MOMDO)
For the approach he envisions, he developed the Multi-Objective and Multi-Disciplinary Optimization (MOMDO) method that highlights reformulating optimization problems. MOMDO consists of two sub-methods which help to explore the design space both in a convergent and a divergent manner. He also involved the development of a plug-in for combining Grasshopper and modeFRONTIER, based on a collaboration between TU Delft Design Informatics chair and ESTECO SpA. This plug-in makes it possible to implement the MOMDO method. He applied the method to two practical cases.
The first case involved a large indoor stadium in Wuhan with a stepped roof and 6,000 seats. The big question was how the stepped shape, with glass sections, would affect the performances of the building. Computational techniques such as "Hierarchical Clustering" and "Self-Organizing Map" were used to analyze the performance data. ‘They serve as a tool for the architect in gathering a lot of information and knowledge,’ Yang explained. ‘In this case, it was mainly about a convergent approach: we were able to reduce the number of possible designs to a few optimized ones.’
In the second case, the MOMDO method worked in a divergent fashion. Here, the method helped expand the number of possible design concepts from three to nine. Such an approach allows the architect to choose a design concept that gives the best balance between the desired building performances.
The MOMDO method can help automating some of the design process. But fully automatic generation of promising design concepts is not yet possible. ‘And that is not the intention either,’ Yang says. ‘There will always be people needed for interpreting collected information in disciplinary contexts and deciding which aspects to focus on. From a practical point of view, human intervention in a design process is usually necessary and preferable.’
Published: January 2023
Ding Yang received his doctorate on 23 December 2022, with his dissertation, "Design as Exploration Multi-Objective and Multi-Disciplinary Optimization (MOMDO) of indoor sports halls. His dissertation can be read here.