Modeling Dynamic Gas-Liquid Interfaces in Underwater Explosions Using Interval-Constrained Physics-Informed Neural Networks

TL;DR

This paper introduces a dual-network physics-informed neural network framework with interval constraints for accurately modeling dynamic gas-liquid interfaces in underwater explosions, improving efficiency over traditional CFD methods.

Contribution

The paper presents a novel dual-network PINN architecture with interval-constraint training and physics-preserving mappings to effectively model complex underwater explosion phenomena.

Findings

Accurately reconstructs spatiotemporal fields from coarse data

Achieves higher computational efficiency than traditional CFD methods

Extensible to higher-dimensional problems

Abstract

Underwater explosion modeling faces a critical challenge of simultaneously resolving shock waves and gas-liquid interfaces, as traditional methods struggle to balance accuracy and computational efficiency. To address this, we develop a physics-informed neural network (PINN) framework featuring a dual-network architecture, that one network learns flow-field variables (pressure, density, velocity) from simulation data, while another network tracks the gas-liquid interface despite lacking direct numerical solutions. Crucially, we introduce an interval-constraint training strategy that penalizes interface deviations beyond grid spacing limits, paired with a physics-preserving linear mapping of 1D spherical Euler equations to ensure consistency. Our results show that this approach accurately reconstructs spatiotemporal fields from coarse-grid data, achieving superior computational efficiency over conventional CFD-enabling rapid, mesh-free blast-load analysis for near/far-field scenarios and extensibility to higher-dimensional problems.