Integration Wavelet Neural Network for Steady Convection Dominated Diffusion Problem

In this paper, we first propose an Integration Wavelet Neural Network (I_WNN) based on spline wavelet functions for solving differential equations. Then the proposed method is verified successfully by solving two steady convection dominated diffusion problems and the numerical perturbation doesn´t happen when the ratio of convective coefficient to diffusive coefficient is as high as 100. Moreover, once the I_WNN is trained and the parameters are stored, it allows instantaneous evaluation of solution at any desired point spending negligible computing time and memory and the maximum relative error is 1.13%.