Flexible, Data-Driven Parametric Design for Building Façades
Open Access
- Author:
- Hinkle, Laura
- Graduate Program:
- Architectural Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 12, 2024
- Committee Members:
- James Freihaut, Program Head/Chair
James Wang, Outside Unit & Field Member
Nathan Brown, Chair & Dissertation Advisor
Julian Wang, Co-Chair & Dissertation Advisor
Greg Pavlak, Major Field Member - Keywords:
- building facades
parametric design
surrogate modeling
dynamic facades - Abstract:
- Building façades are an important aspect of the building envelope, regulating the light, heat, and ventilation exchange from the outdoors to the indoor environment. A well-designed façade minimizes operational energy while also achieving ideal daylighting conditions and allowing for ventilation. Maintaining this balance is a challenging design task. With growing computational resources, it is now possible to simulate these behaviors and develop robust workflows to estimate the performance of building façades. However, these methods are not widespread due to the manual effort required to establish the building model, which is parametric in nature, and the challenge of analyzing data while exploring design decisions. Additionally, there is a new category of façade solutions that are dynamic, adding more complexity to the design task. This dissertation first proposes and demonstrates two new early design methods that improve upon the accessibility and flexibility of surrogate model-based workflows. The latter half of this dissertation focuses on optimizing dynamic façades, specifically in the glazing category, in order to begin to integrate dynamic façades with traditional early design decisions. The first study addresses the issue of accessibility by proposing a new tree-based surrogate model workflow that filters a large generalizable design space, making it reusable. Three early design spaces were constructed, and using this workflow, the large pre-computed dataset can be filtered to provide specific variable importance and performance estimates across early design changes and multiple projects. The second study demonstrates a second workflow that, instead of filtering down a generalizable design space, allows for the addition of new design variables to a custom parametric model and corresponding surrogate model with fewer simulations. It accomplishes this through the application of a tabular transfer learning approach paired with random walks sampling. Through a building façade case study, it is shown that applying this workflow can significantly reduce the number of samples required to achieve sufficient surrogate model performance compared to classical machine learning approaches. This approach reduces the time between applying early design changes, improving the flexibility of surrogate model-based workflows. However, considering dynamic façade elements in the early design process would improve flexibility further and introduce more creative, sustainable design solutions, which are the focus of the second half of this dissertation. The third study shifts to fundamental questions regarding dynamic façade performance, specifically in dynamic glazing. In this work, parametric energy simulations were conducted to determine the optimal dynamic glazing properties across multiple climates. This allowed for the determination of the ideal relationship between these properties, intended to guide future product development. It also identifies the ideal transition temperature for such technology and offers guidance on decoupling strategies for each climate zone. Finally, the fourth study begins to integrate traditional early design decisions considered in the first two studies with those from the third study. A series of constrained optimization runs were conducted to demonstrate the consequences of traditional sequential early design process that considers building geometry somewhat independently of façade materials. In the unique scenario of dynamic glazing applications, it is beneficial to consider dynamic glazing variables in the early stages since they are sensitive to orientation, self-shading, and radiant heat exchange with respect to building form. This work paves the way for a fully integrated design workflow that accounts for both static and dynamic design decisions, generating more innovative façade options necessary to meet current and future sustainability goals.