Prediction of induction parameters on toroidal wound cores using neural network

This paper presents a new approach based on neural network to predict the induction parameters of the toroidal wound cores. The input parameters were the geometrical dimensions of the toroidal core, frequency and magnetic flux density. A total of 3176 input vector from previously measured 52 varied dimensions and built 27M4 material toroidal samples were available in the training set to a back-propagation feed forward neural network. The sigmoid and hyperbolic tangent transfer functions and full connectivity were used in the hidden layers. The correlation coefficients for the total harmonic distortion and form factor were found to be 0.99 and 0.98, respectively after the network was trained