Fluid-structure interactions (FSI) play a fundamental role in marine engineering, influencing the stability, fatigue life, and performance of offshore risers, floating wind turbines, subsea pipelines, fish pens, propellers, and hydraulic energy devices. When a structure is deformed by fluid forces, the resulting motion changes the flow pattern, forming a tightly coupled nonlinear system. High-fidelity simulations using the Navier–Stokes solver require significant central processing unit (CPU) time, while experiments struggle with sparse sensing and moving object limitations. As marine renewable energy expands and system scale increases, faster, data-efficient, and generalizable modeling techniques are urgently needed. These challenges have necessitated further development to leverage machine learning (ML) to accelerate FSI modeling, prediction, and control.
A recent review conducted by the University of Stavanger (Norway) provides a comprehensive overview of how ML is accelerating progress in FSI research. Published at (DOI: 10.26599/OCEAN.2025.9470002) sea In 2025, this review focuses on advances in data-driven flow analysis, low-order modeling, and intelligent control strategies for offshore structures. This review introduces new avenues to improve predictive accuracy, reduce simulation costs, and support innovation in marine renewable energy engineering by integrating computational physics and ML.
This review divides the advances in ML-enabled FSI into three main directions: feature detection, dynamics prediction, and flow structure control. Unsupervised and supervised learning techniques such as suitable orthogonal decomposition (POD), dynamic mode decomposition (DMD), convolutional neural network (CNN) autoencoder, variational autoencoder, generative adversarial network (GAN), and sparse identification of nonlinear dynamics (SINDy) successfully extracted low-dimensional modes from the turbulent wake and revealed the underlying coherent vortex structure. Cylinders and hydrofoils. These models are able to reconstruct the flow field from limited inputs and provide high-resolution representations with negligible simulation costs. Recurrent architectures such as long short-term memory networks (LSTMs) and transformers have improved the temporal prediction of vortex shedding, and low-order models have enabled efficient load estimation and structural response prediction.
Beyond analysis, reinforcement learning is beginning to enable real-time control, allowing numerical tests to suppress vortex-induced vibrations or enhance wave energy collection. Physics-based neural networks integrate Navier-Stokes equations directly into training, achieving solutions without meshing and reducing dependence on extensive datasets.
The authors identify future challenges including nonlinear energy transfer in turbulent flows, sparse experimental sensing, realistic structural geometry, and high Reynolds number conditions. Future success will require hybrid modeling that leverages the efficiency of ML, especially in real ocean deployments, while respecting the laws of physics.
“Machine learning is not replacing classic FSI methods; it is expanding what we can solve,” the authors say. “The ability to decipher flow physics from data, predict future conditions, and adaptively operate controllers provides new avenues for renewable energy systems and marine infrastructure. As models become more physically informed, ML has the potential to transform ocean engineering in the same way it transformed vision and language processing.”
By integrating ML into FSI workflows, engineers can accelerate the design cycle of tidal turbines, extend the fatigue life of subsea pipelines, develop self-optimizing risers, and achieve real-time control of energy harvesting devices. ML-assisted reduced-order models reduce CPU demand from thousands of hours to seconds, enabling rapid risk assessment during storm and installation operations. In the long term, bridging mathematical modeling, big data ocean sensing, and physics-based algorithms could lead to digital twins of offshore structures and autonomous ocean systems. This review suggests that interdisciplinary innovation is key to moving FSI-ML research from laboratory cases to real-world marine industry deployment.
About sea
sea is a peer-reviewed, open-access, international journal that provides an interdisciplinary platform for cutting-edge research and practical applications in the fields of marine science, ocean technology, and ocean engineering. The journal publishes articles, reviews, and perspectives aimed at advancing theoretical, numerical, field-based, and experimental developments to promote global sustainability.
Disclaimer: AAAS and EurekAlert! We are not responsible for the accuracy of news releases posted on EurekAlert! Use of Information by Contributing Institutions or via the EurekAlert System.
