An artificial intelligence method is presented for on-line microstructural characterization of decarburized steels. Detection of microstructural changes is a great importance matter in production lines of steel parts. A new method for microstructural characterization based on the theory of magnetic Barkhausen noise nondestructive testing method is introduced using artificial neural network (ANN). In order to obtain the accurate depth of decarburized layer of carbon steels and to eliminate the frequency effect on the magnetic Barkhausen noise outputs, the magnetic responses were fed into the ANN structure in terms of position, height and width of the Barkhausen profiles. The obtained results showed that the ANN is able to detect and characterize microstructural changes, accurately, despite serious effect of the frequency on the outputs. In other words, implementing multiple outputs simultaneously enables the ANN modeling to approach to the accurate results using only height, position and width of the magnetic Barkhausen noise peaks without knowing the amount of the used frequency.

Keywords