Paper Key : IRJ************841
Author: Munesh Meghwar,Ahmer Waleed,Salman Ali,Mehshan Mehboob
Date Published: 02 Apr 2025
Abstract
This study investigates the application of Physics-Informed Neural Networks (PINNs) to solve the Euler-Bernoulli beam equation for a simply supported beam subjected to a uniformly distributed load. The performance of the PINN model is evaluated against the exact analytical solution and a Finite Element-based Weighted Residual Method (WRM), using normalized parameters to ensure consistency across all methods. Results show that the PINN model accurately captures the beams deflection behavior, with a predicted mid-span displacement of 12.9333 mm compared to 13.0192 mm analytically. While the WRM demonstrates higher numerical precision with a lower mean absolute error (0.0340 mm vs. 0.0865 mm for PINN), the PINN offers a flexible, mesh-free alternative capable of learning from governing physics alone. The study highlights key advantages and limitations of PINNs in handling high-order boundary conditions and suggests future extensions to full-scale structures and hybrid frameworks incorporating experimental data.
DOI Requested