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

TL;DR

This paper evaluates neural operators like DeepONet, FNO, and WNO for real-time prediction of offshore floating structure responses under irregular waves, demonstrating significant speed and accuracy improvements over traditional methods.

Contribution

It compares multiple neural operators for offshore structure response prediction, introduces novel extensions, and identifies the most effective models for different response characteristics.

Findings

Neural operators are up to 100 times faster than conventional numerical solvers.

FNO outperforms others for small bandwidth frequency responses.

DeepONet with historical states excels in large frequency spectrum responses.

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.