Rationally and effectively balancing the design objectives of architecture and structure is challenging owing to the limitations of existing truss design methods. Inspired by visual design principles, this study proposes a non-black-box, parametric, and integrated solution based on principal stress lines (PSLs). The idea is to mine the potential data features of PSLs by the clustering algorithm. Clustering the endpoints and equipartition points of PSLs is the core operation of this method. The clustering results are then used to reflect the geometric features and distribution characteristics of PSLs. For this purpose, the framework of two clustering algorithms (DBSCAN and K-means) is adapted with additional functions for sorting, screening, and counting the sample size of specific clusters. Moreover, combined with the layout pattern of linear elements, the truss web number, size, and layout are defined based on the clustering results. In addition, to verify the feasibility of form-finding results, the method is applied to a cantilever truss and two practical projects: Taishan Bridge and the Wing Stairway. Finally, the application potential of this method is demonstrated by comparing its conventional trusses under the same load and support conditions (materials used, maximum displacement, and maximum axial stress). • The proposed truss generation concept is initiated from structural performance, and at the same time is highly compatible with the architect’s design demand for visual presentation. • Compared to the other conventional structural optimization methods based on principal stress lines, the workflow of truss form evolving methodology provided in the paper is non-black-box and achieves parametric visualization in the whole designing process. • There are multiple parameters to control truss webs’ number, size, and layout pattern, which makes diversified outcomes of form generation possible to meet the different requirements of the designers. • The tensile and compression members in trusses can be distinguished to effectively select their cross-sectional sizes, which are important in the accurate demonstration of material properties for both architects and engineers.
Read full abstract