Large composite structures, commonly used in aerospace and offshore wind energy, require effective joining techniques, with bolted connections playing a crucial role. However, developing high-fidelity Finite Element (FE) models for these joints is computationally expensive. The presented study proposes an alternative approach that accelerates the simulation process using submodeling with shell and beam elements.
The behavioral response of individual FE models is input into a Feed Forward Neural Network (FFN), which is then integrated into a two-noded user element. This hybrid model enables the accurate simulation of the nonlinear response of bolted joints while drastically reducing computational costs.
This method makes it possible to efficiently simulate large structures with multiple bolted joints, fully leveraging the potential of composite materials. It represents a significant advancement in numerical modeling for aerospace engineering and other structural applications.
🔗 Read more about the publication here: Source