Hybrid Method for Inverse Electromagnetic Coil Optimization Using Multi-transition and Hopfield Neural Networks

In this paper, a hybrid method for inverse optimization of electromagnetic coils utilizing the multi-transition neural network and the Hopfield neural network is proposed. Due to the discrete character of the neural network, an optimization problem is transformed into a discrete problem through the division of the entire coil area into elemental coils with constant current density. The minimization of the objective function is performed by the multi-transition neural network and the Hopfield neural network in turns. Subdivision of the elemental coils is performed in order to achieved better accuracy of the results which are verified using 2-D finite element analysis. The application of the proposed method for inverse optimization of MRI device is also presented