Riding the AI wave toward rapid, precise ocean simulations
New machine learning model cuts fluid simulation time from 45 minutes to 3
- Date:
- April 3, 2025
- Source:
- Osaka Metropolitan University
- Summary:
- Scientists have developed an AI-powered fluid simulation model that significantly reduces computation time while maintaining accuracy. Their approach could aid offshore power generation, ship design and ocean monitoring.
- Share:
AI has created a sea change in society; now, it is setting its sights on the sea itself.
Researchers at Osaka Metropolitan University have developed a machine learning-powered fluid simulation model that significantly reduces computation time without compromising accuracy. Their fast and precise technique opens up potential applications in offshore power generation, ship design and real-time ocean monitoring.
Accurately predicting fluid behavior is crucial for industries relying on wave and tidal energy, as well as for design of maritime structures and vessels. Whilst particle methods -- which allow particles to simulate the behavior of fluid flow -- are a common approach, they require extensive computational resources, including processing power and time. By simplifying and accelerating fluid simulations, AI-powered surrogate models are making waves in fluid dynamics research.
However, AI is not without its flaws.
"AI can deliver exceptional results for specific problems but often struggles when applied to different conditions," said Takefumi Higaki, an assistant professor at Osaka Metropolitan University's Graduate School of Engineering and lead author of the study.
Aiming to create a tool that is consistently fast and accurate, the team developed a new surrogate model using a deep learning technology called graph neural networks. The researchers first compared different training conditions to determine what factors were essential for high-precision fluid calculations. They then systematically evaluated how well their model adapted to different simulation speeds, known as time step sizes, and various types of fluid movements.
The results demonstrated strong generalization capabilities across different fluid behaviors.
"Our model maintains the same level of accuracy as traditional particle-based simulations, throughout various fluid scenarios, while reducing computation time from approximately 45 minutes to just three minutes," Higaki said.
This research marks a step forward in high-performance fluid simulation, offering a scalable and generalizable solution that balances accuracy with efficiency. Such improvements extend beyond the lab.
"Faster and more precise fluid simulations can mean a significant acceleration in the design process for ships and offshore energy systems," Higaki said. "They also enable real-time fluid behavior analysis, which could maximize the efficiency of ocean energy systems."
The study was published in Applied Ocean Research.
Story Source:
Materials provided by Osaka Metropolitan University. Note: Content may be edited for style and length.
Journal Reference:
- Takefumi Higaki, Yuki Tanabe, Hirotada Hashimoto, Takahito Iida. Step-by-step enhancement of a graph neural network-based surrogate model for Lagrangian fluid simulations with flexible time step sizes. Applied Ocean Research, 2025; 154: 104424 DOI: 10.1016/j.apor.2025.104424
Cite This Page: