Solving Heat Transfer Equation by Using Feed Forward Neural Networks

Abstract

The aim of this paper is to design feed forward neural network (FFNN) to solvethe heat equation. Using a multi-layer with 7 hidden units (neurons) and one linearoutput unit, the sigmoid activation function of each unit in hidden layer is tansigfunction, where the Levenberg – Marquardt training algorithm is used to train thenetwork .The existence of the proposed solution was proved. The suggested networkshave been studied intensively for a few decades and have provided an option formodeling complex systems. Therefore this option was utilized to reduce thecomputation of solution, and finally the method is demonstrated through illustrativeexamples.