arXiv subjects

Soheil Radfar

Publications and source records attributed to Soheil Radfar.

2 recordsLinked to original sources

arXiv2026

How Sparse and How Noisy? Systematic Benchmarking of Inverse Physics-Informed Neural Networks for Manning Friction Estimation in Shallow Water Equations

Physics-informed neural networks (PINNs) offer a promising framework for inverse hydrodynamic modeling by combining sparse observations with governing physical constraints. However, their reliability for estimating hydraulic parameters under data limitations remains insufficiently characterized. This study benchmarks inverse PINN recovery of the Manning friction coefficient in the shallow water equations under controlled variations in observation sparsity, noise, and observed variable type. Two cases are considered: a one-dimensional MacDonald subcritical channel with an analytical steady reference solution, and a two-dimensional sloped channel with a parabolic transverse bed generated using a balanced finite-volume solver. The Manning coefficient is treated as a trainable positive scalar and recovered jointly with the flow field using a two-phase strategy that first fits observations and then incorporates the physics residual. Results show that the two-dimensional case achieves robust friction recovery, with errors below 5% when at least 10 depth and velocity observations are available and noise is at or below 10% of the field standard deviation. Recovery remains stable up to 20% noise with 50 observations, but becomes unreliable with only five observations. In contrast, the one-dimensional case shows a persistent positive bias of about 15% that is largely insensitive to observation count and noise, indicating a structural identifiability limitation rather than a data-density limitation. Observation-type ablation shows that recovery degrades substantially when only depth or velocity is observed, demonstrating that joint depth-velocity information is essential for reliable inverse identification. Overall, the results provide a reproducible benchmark for assessing when inverse PINNs can and cannot reliably estimate Manning friction from sparse and noisy shallow-water observations.

physics.flu-dyn↗

arXiv2025

Integrating Newton's Laws with deep learning for enhanced physics-informed compound flood modelling

Coastal communities increasingly face compound floods, where multiple drivers like storm surge, high tide, heavy rainfall, and river discharge occur together or in sequence to produce impacts far greater than any single driver alone. Traditional hydrodynamic models can provide accurate physics-based simulations but require substantial computational resources for real-time applications or risk assessments, while machine learning alternatives often sacrifice physical consistency for speed, producing unrealistic predictions during extreme events. This study addresses these challenges by developing ALPINE (All-in-one Physics Informed Neural Emulator), a physics-informed neural network (PINN) framework to enforce complete shallow water dynamics in compound flood modeling. Unlike previous approaches that implement partial constraints, our framework simultaneously enforces mass conservation and both momentum equations, ensuring full adherence to Newton's laws throughout the prediction process. The model integrates a convolutional encoder-decoder architecture with ConvLSTM temporal processing, trained using a composite loss function that balances data fidelity with physics-based residuals. Using six historical storm events (four for training, one for validation, and one held-out for unseen testing), we observe substantial improvements over baseline neural networks. ALPINE reduces domain-averaged prediction errors and improves model skill metrics for water surface elevation and velocity components. Physics-informed constraints prove most valuable during peak storm intensity, when multiple flood drivers interact and reliable predictions matter most. This approach yields a physically consistent emulator capable of supporting compound-flood forecasting and large-scale risk analyses while preserving physical realism essential for coastal emergency management.

physics.geo-ph↗