Estimation of wave-induced full-field internal forces in submerged floating tunnels using LSTM network with finite impulse response filter and modal approach

초록

Submerged floating tunnels (SFTs) are an innovative solution for underwater transportation. Although they offer significant advantages for deep-water construction, the practical implementation of SFTs remains challenging. This study addresses the need for the accurate estimation of wave-induced internal forces in SFTs using acceleration under varying wave conditions. Building on previous studies that employed long short-term memory (LSTM) networks, we propose an enhanced method that integrates an LSTM-based network with a finite impulse response (FIR) filter and a modal approach. The FIR filter is used to reconstruct the displacement from acceleration data, providing a stronger correlation with internal forces. The LSTM-based network then estimates the full-field strain and internal forces (tether tension, vertical bending moment, and lateral bending moment) by incorporating the modal information. The proposed method was validated using simulated data under diverse wave conditions, demonstrating improved accuracy and generalizability compared with the conventional LSTM model. © 2025 Elsevier Ltd

키워드