Physics Informed Functional Link Neural Network for Solving Bending Analysis of Perforated Structures with Sinusoidal Load

Abstract

This work introduces a framework called the Physics Informed Functional Link Neural Network (PI FLNN) for solving the bending analysis of perforated structures with sinusoidal load, in which the hidden layer is replaced by a functional expansion block for enhancement of the input patterns using orthogonal polynomials to estimates solution of differential equations by integrating neural networks with constrained expression (CE) that contain a free function, and the CE satisfies boundary conditions. The objective is to understand how PI FLNN analyzed the bending behaviour of the squared holes beam structures with external force. In this approach, the free function is represented b y a functional link neural network, which learns to solve the resulting unconstrained optimization problem. The influence of filling ratio and the number of rows of holes due to the perforation on the bending behaviour of this perforated structures has been systematically investigated by using this PI FLNN method and compared with the numerical results. The outcomes show that the PI FLNN method enables quicker training, delivers more precise results, and is more efficient.