PHYSICS-INFORMED NEURAL NETWORKS FOR TRANSIENT GAS DISPERSION MODELLING

This work investigates the application of Physics-Informed Neural Networks (PINNs) to model transient gas dispersion. The model combines data and governing equations to predict concentration, velocity and pressure fields without explicitly imposing boundary or initial conditions. The results show that the PINN can capture the general behaviour of the gas plume and its propagation and interaction with obstacles.