Neural mesh iteration for the FE computation of charged-particle trajectories in vacuum electron devices

In this paper, a neural mesh-adaptation iterative procedure is presented for the finite-element solution of coupled steady-state electromechanical problems describing the motion of charged particles inside vacuum electron devices. In the procedure at each step a neural algorithm performs a mesh refinement in which the mesh "density" is related to the current estimate of the three-dimensional (3-D) spatial charge density distribution. Convergence to an accurate solution is obtained with a reduced computational effort in comparison with conventional fixed mesh iterative techniques. An example of application to the 3-D calculation of electron trajectories inside the collector of a travelling wave tube is shown in order to illustrate the advantages of the iterative scheme.