Eig-PIELM: an efficient mesh-free method for eigenvalue problems using physics-informed extreme learning machines

R. K. Mishra

doi:10.1016/j.cma.2025.118674

Eig-PIELM: an efficient mesh-free method for eigenvalue problems using physics-informed extreme learning machines

2026-01-03

0

PRE

AI

Ganapathy Krishnamurthi

B

Balaji Vasan Srinivasan

S

Sundararajan Natarajan

DOI：10.1016/j.cma.2025.118674

Original

Original request for help

Save

Abstract

En 中文

• Extends the PIELM framework to linear eigenvalue problems by reformulating them into a single direct linear system. • Computes entire eigenspectrum and corresponding mode shapes in one linear solve without iterative optimization. • Enforces boundary conditions exactly at the boundary collocation points via algebraic projection, eliminating penalty tuning. • Mesh-free framework using direct linear algebra only, avoiding backpropagation and gradient-based training. • Achieves high levels of accuracy while maintaining very low computational cost across diverse benchmark problems.

Eig-PIELM: an efficient mesh-free method for eigenvalue problems using physics-informed extreme learning machines

Abstract

Journal

Organization