3D ELF magnetic field strength modeling through fully connected cascade networks

The present work documents the research towards the development of an efficient, fast and reliable black-box model to assess the magnetic field at extremely low frequency in a given volume. The approach is based on the implementation of an array of neural networks (aggregated/bootstrapped) trained on suitably conditioned experimental measurements. To enhance the computational performance of the method, a newly developed architecture was used for the neural networks: the fully connected cascade. To validate the approach, the same implementation using classic Feed Forward networks is compared in terms of both computational costs and precision.