Physics‐Informed Neural Networks for Modeling Near‐Wall Steady Laminar Flows at High Reynolds Numbers

In this work, a multiscale, zonal‐coupled physics‐informed neural network (PINN) framework, termed BL‐NS‐PINNs, is proposed for accurately predicting steady laminar near‐wall flow fields at high Reynolds numbers without requiring near‐wall data. The framework ensures physical continuity across flow regions by solving the Blasius and Navier–Stokes equations in the inner and outer zones, respectively, while enforcing velocity consistency within the matching region. Numerical experiments demonstrate that BL‐NS‐PINNs significantly outperform conventional PINNs trained with data supervision, accurately capturing near‐wall flow features and sharp gradients in high‐Reynolds‐number laminar boundary layers. The method is further extended to wedge flows with self‐similar solutions, offering both a theoretical foundation and a practical pathway for integrating wall functions into data‐driven boundary‐layer modeling. Overall, BL‐NS‐PINNs effectively overcome the limitations of conventional PINNs in resolving sharp near‐wall velocity gradients and show strong potential for efficient, high‐fidelity modeling of high‐Reynolds‐number flows.