Offshore structures and ships face significant nonlinear responses from large ocean waves. Generally, Computational Fluid Dynamics (CFD) models are used to solve these responses, but they are impractical for design phase evaluations as they are computationally expensive. Therefore, the industry relies on weakly nonlinear time domain methods to keep simulation time reasonable. Froude Krylov pressures on the instantaneous wetted surface area are estimated, accounting for the non-linear characteristics based on sinusoidal wave profiles. However, evaluating the wave profile accounting for its non-linear characteristics is critical for estimating the wetted surface area, forces, and moments. This study proposes a machine-learning approach to map wave profiles in space and time, accounting for non-linear characteristics, which can later be used to quickly compute water particle kinematics and hydrodynamic forces. The data for training the neural network are generated from IITM-FNPT2D (in-house code) to predict spatio-temporal wave elevations efficiently.
Ships and offshore structures exhibit nonlinear responses when subjected to large ocean waves. These responses have profound implications for offshore structural integrity (Venkataramana and Kawano, 1995; Yim, Solomon and Osborne, Alfred and Mohtat, and Ali, 2017) safety and navigation operations (Wang and Wan, 2020) and marine environmental impacts. Understanding the effect of these nonlinear responses of structures caused by ocean waves is essential for enhancing the safety of navigation and design and the reliability and sustainability of ships and offshore structures. Hence, the design strategies of these marine structures need to be analyzed using simulations to ensure their robustness while optimizing operational safety and efficiency in marine environments.
Various computational methods, such as advanced computational fluid dynamics (CFD) methods, potential flow theory, Morison equations, and experimental validations (Sarpkaya and Isaacson, 1981) are commonly used for accurate estimation of hydrodynamic forces and nonlinear responses of offshore structures and ships to waves. However, CFD methods are often considered to be computationally demanding and time-consuming for marine hydrodynamics. Current industries still rely on weakly non-linear time domain methods to maintain practical simulation times during the design phase (Somayajula and Falzarano, 2015). One crucial aspect involves evaluating Froude Krylov pressures on the instantaneous wetted surface area of the body under incident waves, where the wave elevation is often approximated as a sinusoidal profile. However, this approximation significantly impacts wetted surface area and the calculated forces and moments. To accurately calculate the wetted surface area, achieving a more precise estimation of the wave surface elevation is necessary. In the case of large amplitude waves where the shape of the water surface deviates significantly from a simple sinusoidal shape (Adcock, Gibbs, and Taylor, 2012; Adcock, Taylor, and Draper, 2015) Fully Nonlinear Potential Flow Theory (FNPT) (Wang, Wang, Yan, Ma, and Xia, 2019; Hu, Yan, Greaves, Mai, Raby, and Ma, 2020) can be used to capture it. However, traditional FNPT-based flow solvers must be simulated for the entire domain until the specified time to obtain the incident wave surface elevations at any given spatiotemporal coordinate. This makes the design process time-consuming for engineers who must analyze the structural responses.