Author(s): Rijurekha Dasgupta; Subhasish Das; Gourab Banerjee; Asis Mazumdar
Linked Author(s):
Keywords: Physics-informed neural network; Stage-discharge rating curve; Symbolic regression
Abstract: Static stage-discharge rating curves often produce errors as they cannot represent hysteresis caused by unsteady conditions. Several studies have attempted to incorporate dynamic behavior into stage-discharge models. Here, Jones formula based Physics-Informed Neural Network is developed and explored by Symbolic Regression to derive new robust equation. The result shows the superiority of this approach with a significant improvement in R2 and other measures. The study concludes the importance of incorporating physical laws into neural networks for better appropriation of the stage-discharge relationship under dynamic conditions.
Year: 2026