Algorithm development based on artificial intelligence for function manipulation and comparation with numerical techniques used in engineering teaching

This paper aims to develop and verify the efficiency of algorithm aplications and compare them with the numerical techniques used in engineering areas, stem from structures of neural network for function approximation. For that, function modeling were used, based on complex formulas, making approximation mandatory. The statement investigated is that neural networks are capable to solve with great accuracy problems of this kind. As a test model, a measurement approximation of thermodynamics tables (saturated water) was carried and it was compared with the results of traditional methods (linear and quadratic functions), confirming the efficiency and possibility of its use.