11 Septembre – Thesis defense - Abdulaziz Afandi

10 h30 Amphi 3 - ENSAP Bordeaux (Talence)

Methodological and software approaches for decision-making and optimized parametric design of modular buildings.

Optimization is a profitable behavior for design in general and architectural design in specific. Many generative design optimization tools do exist. However, these tools are not widely used among architects. Design problems are ill-structured problems; designers’ creativity and interpretation are essential for solving these problems. In design, designers’ acceptability of the solutions is as important as the numerical optimality of their performance. The existing tools do not integrate designers’ preferences inside the optimization process; designers’ preferences are crucial for acceptability. The unbalance collaboration between the tools, and the designer is a major cause of the reluctance of architects from using these tools. The dissertation aims to define a set of recommendations that helps developers to introduce decision support systems that attract more architects by improving the collaboration between the designers and the tools.
To define the set of recommendations, the research started by exploring different design processes. Based on this exploration, a design framework based on four models Morphogenesis, Observation, Interpretation, Aggregation (MOIA) is defined. Next, the tool typologies the architects use are explored. Additionally, the popular generative design optimization workflows are investigated by using MOIA as a reference. Then, the research adopts an experimental approach based on designers’ acceptability. Five different experiments are performed. Two of the experiments compare different existing generative design optimization workflows by using designers’ acceptability as a reference. The other three experiments compare different aggregation functions by using designers’ judgment as a benchmark. These functions are Pareto’s function, Maximin, and Derringer & Suich’s.
The results of these experiments can be concluded in four points. First, visual programming is recommended for future generative optimization tools. Visual programming helps the architect describe sophisticated parametric models without coding; designers, in general, are not trained to code. Second, the graphical aspect of the tool can immensely influence the decision of the designer. The performance of the solutions must be graphically presented to the designers; the representation method must respond to the number of objectives. Third, using an interactive optimization algorithm that allows the designers to select the solution based on their subjective judgment of the form can increase the acceptability of the workflows. Fourth, the availability of information is the key to define the accessible aggregation function. We usually use the aggregation function that integrates more of the available information; this information includes designers’ preferences, which help to approach acceptability.
The existing generative design optimization tools need to attract more architects. The research adopts an experimental approach based on designers’ acceptability. The methodology helped the research define a set of recommendations that can help future tools attract more designers to optimize. The recommendation mainly focuses on enhancing the collaboration between the tools and the designers.

Event localization