Abstract:The use of neural operators in a digital twin model of an offshore floating structure can provide a paradigm shift in structural response prediction and health monitoring, providing valuable information for real-time control. In this work, the performance of three neural operators is evaluated, namely, deep operator network (DeepONet), Fourier neural operator (FNO), and Wavelet neural operator (WNO). We investigate the effectiveness of the operators to accurately capture the responses of a floating structure under six different sea state codes $(3-8)$ based on the wave characteristics described by the World Meteorological Organization (WMO). The results demonstrate that these high-precision neural operators can deliver structural responses more efficiently, up to two orders of magnitude faster than a dynamic analysis using conventional numerical solvers. Additionally, compared to gated recurrent units (GRUs), a commonly used recurrent neural network for time-series estimation, neural operators are both more accurate and efficient, especially in situations with limited data availability. To further enhance the accuracy, novel extensions, such as wavelet-DeepONet and self-adaptive WNO, are proposed. Taken together, our study shows that FNO outperforms all other operators for approximating the mapping of one input functional space to the output space as well as for responses that have small bandwidth of the frequency spectrum, whereas for learning the mapping of multiple functions in the input space to the output space as well as for capturing responses within a large frequency spectrum, DeepONet with historical states provides the highest accuracy.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: predicting the real - time responses of floating offshore structures under the action of irregular waves. Specifically, the research aims to improve the ability to predict the responses of floating structures in different sea states by using Deep Neural Operators, thereby achieving more effective real - time control and health monitoring. ### Research Background and Problem Description With the growth of deep - water exploration demands, especially the resource and energy development in harsh environments, safe and reliable floating offshore structures have become extremely important. Failures of these structures in deep water may lead to serious economic and environmental consequences. When floating structures are subjected to environmental forces (such as waves, wind, etc.), they can move in six degrees of freedom (DOFs), but their movements are restricted by the mooring system. Accurately predicting the responses of these structures under the action of waves is a major challenge because it involves complex nonlinear characteristics, wide - spectrum and multi - scale features. ### Solution To meet this challenge, researchers proposed and evaluated three main neural operators: Deep Operator Network (DeepONet), Fourier Neural Operator (FNO) and Wavelet Neural Operator (WNO). These neural operators can efficiently learn the mapping relationship from the input wave height to the structural response, and can provide prediction results two orders of magnitude faster than traditional numerical solvers. In addition, compared with the commonly used Gated Recurrent Unit (GRU), these neural operators show higher accuracy and efficiency in the case of limited data. ### Main Contributions 1. **Introducing New Extended Architectures**: To further improve the prediction accuracy, the paper proposes two new extended architectures - Wavelet - DeepONet and Self - adaptive WNO. 2. **Handling Initial Conditions**: Different response situations under zero - initial and non - zero - initial conditions are considered to ensure the applicability of the model in various practical application scenarios. 3. **Performance Comparison**: By comparing with traditional methods (such as GRU), the superiority of the proposed neural operators in predicting the responses of floating structures is verified. ### Formula Summary The key formulas involved in the paper include: - The combined operator equation for structural response: \[ (M + A)\ddot{x}(t)+\int_{0}^{t}K(t - \tau)\dot{x}(\tau)d\tau + Cx(t)+f_{\text{moor}}(x(t),\dot{x}(t))=f_{\text{wa}}(\eta(t)) \] where $M, A, C\in\mathbb{R}^{6\times6}$ represent the mass matrix, added mass matrix and linear hydrostatic restoring force coefficient matrix of the structure respectively; $K(t)$ is the hysteresis function; $\ddot{x}(t),\dot{x}(t),x(t)\in\mathbb{R}^{6\times1}$ represent the acceleration, velocity and displacement vectors respectively; $f_{\text{moor}}$ and $f_{\text{wa}}$ are the mooring force and wave force respectively. - The expression of the learning operator $G_{x}$: \[ x(t)=G_{x}(t;x(t_{0}),\dot{x}(t_{0}),\eta(t)) \] These formulas help define the dynamic behavior of floating structures under the action of waves and provide a theoretical basis for the learning of neural operators. In conclusion, this paper significantly improves the real - time response prediction ability of floating offshore structures in complex sea states by introducing advanced deep neural operator techniques, providing strong support for the safety and reliability of offshore engineering.

Deep neural operators can predict the real-time response of floating offshore structures under irregular waves

Evaluation of Deep Neural Operator Models toward Ocean Forecasting

Wavelet neural operator: a neural operator for parametric partial differential equations

A Physics-Informed Neural Operator for the Simulation of Surface Waves

A Data-Driven Methodology for Wave Time-Series Measurement on Floating Structures

Data-driven modelling of wave–structure interaction for a moored floating structure

Deep Learning-Based Prediction of Pitch Response for Floating Offshore Wind Turbines

Motion and Load Prediction of Floating Platform in South China Sea Using Deep Learning and Prototype Monitoring Information

Wavelet Neural Operator for solving parametric partial differential equations in computational mechanics problems

OceanNet: A principled neural operator-based digital twin for regional oceans

Forecasting Pitch Response of Floating Offshore Wind Turbines with a Deep Learning Model

Fourier Neural Operator Surrogate Model to Predict 3D Seismic Waves Propagation

Real-time prediction of hydrodynamic forces and dynamic responses of a generic jack-up platform using deep learning

Neural network survivability approach of a wave energy converter considering uncertainties in the prediction of future knowledge

Instantaneous prediction of irregular ocean surface wave based on deep learning

Phase-resolved wave prediction with varying buoy positions in the field using machine learning-based methods

Structural health monitoring on offshore jacket platforms using a novel ensemble deep learning model

On-line wave prediction

Applying Neural Networks to Predict Offshore Platform Dynamics

A Response Frequency Informed LSTM Model for Ultra-Short-Term Mooring Line Forces Prediction of Floating Wind Turbines

Extreme Motion Predictions of Semi-Submersible Platform Based on Deep Learning Method and Prototype Monitoring Data