Implicit parametric modal expansion method for single-layer reticulated shells based on generative adversarial network

Abstract

Single-layer reticulated shells (SLRSs) are a prevalent form of roofing used in public buildings, including stadium and exhibition centers. Structural health monitoring (SHM) systems have been installed on SLRSs to obtain modal parameters for their design, assessment, and maintenance. However, modifying a complex finite element (FE) model with modal information acquired from a limited number of sensors can be challenging, especially for SLRSs. It is necessary to expand the measurement mode to obtain the complete mode before modifying the FE model. This paper proposes an implicit parametric modal expansion method for SLRSs based on generative adversarial network (GAN) to accomplish modal expansion without prior knowledge of the stiffness and mass of SLRSs. A GAN is trained to establish mapping relationship under different combinations of stiffness and mass of the SLRS. Then the expanded modals can be obtained through the trained GAN after obtaining measuring modals. The effectiveness of the proposed method is verified through numerical examples of spherical SLRSs. Using the method proposed in this paper, the accurate modal expansion results of numerical models can be obtained with a small number of measuring points. The results of the numerical examples indicate that the proposed method can achieve modal expansion through obtained measuring modals and trained GANs without prior knowledge stiffness and mass of SLRS.