Workflow for mitigating shape diversity in free-form surface panels through clustering and evaluation

This is the authors' abstract. We don't add key points for this paper.

With the advancement of digital technologies, free-form surfaces are increasingly utilized in architectural design. Constructing such surfaces typically involves dividing them into numerous panels, which are often uniquely shaped, thereby significantly increasing fabrication and construction complexity, as well as adding to the economic burden. This paper introduces a workflow for optimizing free-form surface facades by reducing panel shape diversity. The proposed workflow consists of four key components: similarity measures (Euclidean and Manhattan distances), clustering algorithms (hierarchical clustering, K-means, and HDBSCAN), evaluation metrics (SC, DB, and CH), and validation through three case studies. The study confirms the correlation between clustering algorithm performance and surface type, highlighting its dependence on input data characteristics. The workflow is applied to the panel optimization of the Hongdao Railway Station project, validating its practical applicability by focusing on the selection of optimal algorithms and the configuration of clustering parameters. As a modular approach, it establishes a foundation for further refinement and extension of methods in future applications.