.png){kind=icon}
.png){kind=icon}
Return
A robust force-finding framework for tensegrity structures using gradient-boosting decision trees and Latin hypercube sampling
.png){kind=icon}
PRE
AI
DOI：10.1016/j.istruc.2025.108863.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In a tensegrity structure with a specified configuration, mechanical performance is significantly influenced by its pre-stress values. Consequently, alongside form-finding, force-finding becomes a task of critical importance. To address the force-finding problem, this paper proposes a novel method integrating machine learning and statistical techniques as an alternative to traditional trial-and-error approaches, thereby eliminating reliance on element group theory and parameters adjustments. The method effectively overcomes challenges such as matrix singularity in finite element analysis and the excessive sample requirements of conventional artificial intelligence techniques. The proposed approach is demonstrated by solving the force-finding problem for both self-stressed and non-self-stressed systems, as well as for symmetric and asymmetric structures. This study also provides an in-depth investigation into three key aspects: the effect of varying element counts (both low and high), the impact of different element groupings (grouped and ungrouped), and the method's performance under various load cases encountered by tensegrity structures. Notably, the proposed method achieves the desired solutions with significantly fewer samples (fewer than 600) for the selected configurations.
Keywords:
Gradient boosting decision tree
Latin hypercube sampling
Tensegrity structure
Self-stressed
Non-self-stressed
External loads