PBMDGO - Performance-based Building Massing Design Generation and Optimisation System/Toolkit
(by wk3849)
The system/toolkit is designed to enable architects carry out performance-based design optimisation or optimisation-based design exploration.
Support Email

NOTICE: Please be kindly aware that the current version is still a proof-of-concept one, and there are some components in the gha that are part of this system but I won't mention in the below description at this stage.

The system (PBMDGO system) is designed to provide a more integrated tool for helping architects carry out performance-based design optimisation or optimisation-based design exploration. The optimisation-based design exploration means that the usage of optimisation is not purely looking for a design solution but more of extracting information from the optimisation result for achieving a performance-aware/performance-informed design synthesis process. Therefore, the optimisation result is more like a "carrier" of design information revealing architectural implications related to building performance rather than a direct answer to a design problem. Thus, please keep in mind that the system is not aimed to produce eye-catching or "futuristic/parametricism" designs but to provide easy-to-understand design information through orthogonal building massing designs (so the generated designs may look simple and unsophisticated).

In order to do so, the system enables architects to perform computational optimisation for performance-based building massing design with two pre-defined generic building massing generative models, which also helps to avoid laborious parametric modelling. In addition, the system can also produce optimisation results with enhancing design information feedback to help architects understand the design problem. 

There are two parts of the components in this system including two building massing design generative models and an exploration-oriented evolutionary algorithm. Users can use these components and other three-party building performance simulation tools, such as DIVA and Honeybee, to establish a customised design optimisation workflow (namely the system) but without tedious parametric modelling. The workflow/system can be re-used in different building massing design tasks as the generative models can be adjusted to different types of building design.

The first part consists of two building massing generative models respectively based on the subtractive and additive form generation principles. The two models can generate design variants appearing as different types of building, such as high-rise towers or middle-rise slab type buildings, according to the user input. In addition, the two models can generate design variants with high design variability and differentiation, which help the optimisation process identify task-specific and site-specific solutions for solving the design problem.

The second part is a hybrid evolutionary algorithm based on an island model and the steady-state replacement strategy, name SSIEA. SSIEA can provide optimisation results with several high-fitness solutions with higher design differentiation than using standard EAs under single-objective optimisation. With a island approach, SSIEA also performs an implicit clustering of the design population, which allows the users to explore design alternatives in different directions. Another useful feature is that SSIEA can also create backup files. Thus, even the optimisation process is stopped due to computer crashes or power shutdown, you can resume the optimisation process by loading the backup file.

We envisage the architect can use this system in building massing design process before design ideation, and the optimisation result can provide information about the trade-offs and compromises and architectural implication related to building performance. This information can help architects to make better decisions in the data-poor environment of early design stages. In comparison, the conventional approach of using performance-based optimisation in architectural design requires the architects to translate their ideas or concepts into the parametric model, which actually locks the design space for the optimisation process into a specific type of building design proposed by the architects. In addition, parametric modelling is also time-consuming and laborious, which can significantly interrupt the architects' design process. Thus, using this system provides a more applicable way of performing performance-based design optimisation and optimisation-based design exploration.

This is the first version of this system. Therefore, there could be still many bugs when using the system. If there is any problem, feel free to contact me via the Support Email or LinkedIn/ResearchGate. To help everyone use this system, I provide an example project in the download area and will upload tutorial videos for using this system later.

Here are some papers about this system, and more papers are coming along in the near future. Hence, if you're interested in the theory or the details of this system, these papers or the Researchgate Project can provide more information about the generative models and SSIEA and help you understand how this system works:

Subtractive Building Massing for Performance-Based Architectural Design Exploration: A Case Study of Daylighting Optimization

DIVERSITY AND EFFICIENCY: A Hybrid Evolutionary Algorithm Combining an Island Model with a Steady-state Replacement Strategy

SSIEA: a hybrid evolutionary algorithm for supporting conceptual architectural design