Floating-Body Hydrodynamic Neural Networks

2025-09-17

AI Summary

Addressing the challenges of modeling dissipative dynamics, ensuring long-term prediction stability, and improving interpretability in floating-body fluid–structure interaction systems, this paper proposes a physics-structured neural network framework. The method couples rigid-body dynamical equations with data-driven hydrodynamic parameter estimation via differentiable modeling to jointly learn orientation-dependent added mass, nonlinear drag coefficients, and background flow fields parameterized by stream functions. Crucially, hydrodynamic priors are embedded directly into the network architecture, thereby constraining the hypothesis space—enhancing both physical interpretability and long-term predictive stability. Evaluated on a synthetic vortex dataset, the approach reduces prediction error by an order of magnitude compared to Neural ODEs, recovers physically consistent flow fields, and outperforms Hamiltonian and Lagrangian neural networks in dissipative regimes.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Intelligent Robots: State Estimation
• Machine Learning: Structured Learning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Fluid-structure interaction is common in engineering and natural systems, where floating-body motion is governed by added mass, drag, and background flows. Modeling these dissipative dynamics is difficult: black-box neural models regress state derivatives with limited interpretability and unstable long-horizon predictions. We propose Floating-Body Hydrodynamic Neural Networks (FHNN), a physics-structured framework that predicts interpretable hydrodynamic parameters such as directional added masses, drag coefficients, and a streamfunction-based flow, and couples them with analytic equations of motion. This design constrains the hypothesis space, enhances interpretability, and stabilizes integration. On synthetic vortex datasets, FHNN achieves up to an order-of-magnitude lower error than Neural ODEs, recovers physically consistent flow fields. Compared with Hamiltonian and Lagrangian neural networks, FHNN more effectively handles dissipative dynamics while preserving interpretability, which bridges the gap between black-box learning and transparent system identification.

Problem

• Modeling dissipative fluid-structure interaction dynamics
• Improving interpretability and stability in hydrodynamic predictions
• Bridging black-box learning with transparent system identification

Innovation

• Physics-structured neural network framework
• Predicts interpretable hydrodynamic parameters
• Couples parameters with analytic motion equations