Physics-informed radial basis networks for force finding of cable domes

The stiffness of cable dome structures is entirely derived from the prestress in their cables and struts, making force-finding a critical step in their design. However, traditional force-finding methods are often complex to implement and have limited applicability. To address these challenges, this paper establishes a general force-finding framework for cable domes based on physics-informed radial basis networks (PIRBN), utilizing neural network techniques to achieve an efficient and reliable force-finding process. Additionally, a loss function is derived that incorporates the physical characteristics of cable domes from the perspective of structural stiffness. Case studies on three types of cable domes were conducted, and the hyperparameter tuning of the network model was simplified using the Optuna hyperparameter optimization method. The results show that PIRBN is suitable for force-finding analysis in both single and multiple prestress mode cable domes, considering the effects of external loads, and provides high computational efficiency and broad applicability.