Artificial neural network modelling of crystallization temperatures of the Ni–P based amorphous alloys

A computer model based on backpropagation multilayer feedforward artificial neural networks (ANN) was designed for the simulation and prediction of crystallization temperatures of the Ni–P based amorphous alloys, as functions of alloy composition and heating rate. The model was trained using data from the published literature as well as own experimental data. The input parameters of the neural network (NN) model are alloy composition, heating rate of the heating processes, and processing parameters of the alloys. The output parameters are crystallization temperatures (onset, peak and end temperatures) of major crystallization reaction of the amorphous alloys. In training and testing, the neural network for the simulation and prediction of crystallization peak temperature shows a comparatively good result. However, the neural networks for the crystallization onset and end temperatures did not show satisfactory performance, due to some unjustified collected data. Some comparisons between NN predictions and own experimental data are given. For easy use of the model a graphical user interface (GUI) was created.