AI and the World's Hard Problems · Physics · arXiv · 2026-10-26
Sensitivity-Guided Parameter Estimation of Bubbly Liquid via a Nonlinear Wave Equation and Physics-InformedNeuralNetworks
Predicting how physical parameters govern nonlinear wave propagation in bubbly liquids is difficult. The authors combined a Korteweg-de Vries-Burgers (KdVB) equation with a physics-informed neural network (PINN), ran a sensitivity analysis over eight physical parameters, and then performed single-parameter inverse estimation for seven of them. Liquid viscosity, bubble initial radius, liquid density, sound speed and liquid pressure were estimated with errors of 0.8–2.1%, the polytropic exponent converged only partially, and the surface tension coefficient could not be estimated, consistent with the sensitivity analysis.