Neural computation to predict magnetic properties of mechanically alloyed Fe–10%Ni and Fe–20%Ni nanocrystalline

Abstract

An artificial neural network (ANN) methodology was applied to relate the powder-milling process parameters to the magnetic properties of Fe–10%Ni and Fe–20%Ni alloy materials. An optimization procedure based on ANN training and testing steps has been developed to predict magnetic properties over a large range of process parameters. A good agreement was found between experimental and predicted results with an average standard deviation between experimental and predicted values less than 6%. The following features have been anticipated: (i) the higher disc rotation speed and the lower the vial rotation speed, the higher the coercivity; (ii) the lower the disc rotation speed and the higher the vial rotation speed, the lower the coercivity.