Prediction of giant magneto impedance on As-cast and post production treated Fe4.3Co68.2Si12.5B15 amorphous wires using neural network

Abstract

A giant magneto impedance effect was experimentally measured on as-cast and post production treated amorphous wires although it takes some time due to varying measuring condition such as sample, static magnetic field and frequency. Measured data from different as-cast and post production treated samples was used for training of the network. A 3-node input layer, 1-node output layer neural network model with 3 hidden layers and full connectivity between nodes were developed. A total of 1600 input vectors obtained from varied samples were available in the training set. The network was formed by hybrid transfer functions and 21 numbers of nodes in the hidden layers, after the performance of many models were tried. A set of test data, different from the training data set was used to investigate the network performance. The average correlation and prediction error of giant magneto impedance effect were found to be 99% and 1% for tested Fe4.3Co68.2 Si12.5B15 amorphous wires.