PINN framework for stellarator heat conductivity inference

A physics-informed neural network (PINN) framework infers perpendicular heat conductivity in stellarator scrape-off layers by combining radial measurements with 1D transport equation residuals. The method uses three simultaneous networks to reconstruct density, temperature, and effective conductivity.