REDES NEURONALES ARTIFICIALES EN CONDUCCIÓN DE CALOR MULTIDIMENSIONAL TRANSITORIO

Neural networks to prediet the multidimensional unsteady-state temperature projile in a solidhave been used; convective and Dirichlet boundmy conditions for the mathematical model wereapplied to salve the model. For computer simulations several neural networks following theMultilayer Perceptron architecture were trained using the Levenberg-Marquardt algorithm. Resultsshowed an excellent agreement between numerical solutions afthe mathematical model and theneural network predictions.