Speaker
Ruben Specogna
(DPIA)
Description
We compare three methods for the solution of eddy current problems arising in fusion technology. We first consider the Finite Element Method formulation based on the reduced magnetic vector potential [1]. This formulation provides a very sparse system matrix and is able to solve problems on meshes composed of tens of millions elements. Yet, it requires to produce the mesh for both conducting and insulating regions, something which is very time consuming. That is why as a second method we consider the volume integral formulation in term of the electric vector potential [2], which requires to mesh only passive conductive structures but having the drawback of dealing with a full system matrix. The third method is based on iteratively solving a Poisson problem in the conductive region and computing a correction magnetic field with the Biot-Savart law [3]. The Poisson problem, solved with a formulation providing a solenoidal current density [4], gives an estimate of the current density by enforcing the Faraday's law. Then, an updated and solenoidal magnetic induction field is obtained form the computed current density with the Biot-Savart law. This magnetic induction field will be in turn used as a source in the Poisson problem. The iterations stop when the computed update is negligible, which means that a self-consistent solution of Faraday's and magnetic Gauss' laws if found. Pros and cons of the proposed methods are assessed on a benchmark problem, i.e. eddy currents induced in ITER-like 3D conducting structures.
[1] O. Biro et al, IEEE Trans Magn 25 (1989) 3145-3159[2] R. Albanese et al, IEE Proc. A 135 (1988) 457-462[3] T. Takagi et al, IEEE Trans Magn 24 (1988) 2682-2684[4] P. Bettini et al, JCP 273 (2014) 100-117
Co-authors
Paolo Bettini
(Consorzio RFX, Padua, Italy)
Ruben Specogna
(DPIA, University of Udine, Udine, Italy)
