A Novel Design Method of Shapes of Ferromagnetic Materials for the Superconducting MRI Magnets

We present a novel non-stochastic method for optimizing the shapes of ferromagnetic materials to generate a target static magnetic field. A magnetizing current corresponding to an error field is calculated as an inverse problem and then the current is replaced with an equivalent ferromagnetic material. Iteration of this process leads to an error field of almost zero. It should be noted that the initial condition of the material shape is critical for convergence. Consequently, we propose a combined method: the initial shape of the material is generated with a stochastic optimization algorithm, and then the shape is updated using an explicit method. We show that this method works well with a test 2D axisymmetric magnet.