Hossein Hosseinzadeh

The subject of this thesis is to propose a numerical method based on an artificial neural networks and optimization for solving one dimensional advection-dispersion equations arising in contaminant transport through porous media. The approximate solutions are considered in such a way that satisfy boundary conditions. For updating network weights, the residual error is used. By starting from a random initial solution and using gradient descent algorithm the weights are updated. Two test problems are considered and our results are compared with the results of some existing methods. Our obtained results are of higher accuracy and the method is more accurate.