Thomas Henneron

The Finite Element method is widely used to study and design electromagnetic devices. This method is very accurate but often requires huge computational time even during the post processing step where quantities of interest are calculated. In this article, we propose to apply and to compare clustering methods, i.e. the K-Means and Gaussian Mixture methods, in order to reduce the computational time of the iron loss distribution in a postprocessing stage. Indeed, the iron loss distribution calculation relies only on the calculation on only a few elements of mesh and not on all elements like in the case of the standard method. The proposed methods are tested on a nonlinear magnetic example based on an axial flux machine showing a good agreement between the iron losses distribution obtained by the standard and the proposed methods.

In order to survey possible local demagnetization of permanent magnet drives during transient operations and/or faulty functioning in case of multi‐phase drives, co‐simulation including converter, vector control, and finite element model could be investigated. Nevertheless, expensive simulation time limits practically this solution. To solve this problem, the paper proposes a surrogate model of an electrical machine taking into account locally the irreversible demagnetization effect. In case of a multi‐phase machine working in transient operations and irreversible demagnetization, the proposed surrogate model gives large benefit in computational time in comparison with an finite element approach.

In order to design efficient power converter operating at high switching frequencies, the electromagnetic behavior of the printed circuit board (PCB) must be well known. Then, numerical approaches can be used to study the voltage and current waveforms as well as the current density distribution associated with a PCB. Considering a half-bridge converter connected to a passive electrical load, we propose a co-simulation based on the Cauer ladder network approach providing an equivalent circuit of the PCB combined with models of GaN transistors. The results from the co-simulation are compared with measurements at 1 MHz with an output power of 200 W.

To study a low frequency electromagnetic device represented by a Finite Element (FE) model coupled with a complex electrical circuit, an approach based on the co-simulation principle can be investigated. Then, each component of a system is solved by a dedicated software. To reduce the computational time of a co-simulation, an approach based on the Proper Generalized Decomposition (PGD) method which is a model order reduction approach is investigated. The proposed approach is evaluated using three examples: a single-phase inductance coupled with a simple circuit equation, a 3-phase transformer connected to a rectifier supplying a variable electrical load and a system composed of four FE models coupled with electrical equations.

In an industrial context of electrical device design or expertise, it is common for engineers to use models based on the finite element (FE) method. In the case of nonlinear magneto-quasistatic problems, it can lead to prohibitive computational times. Then, model order reduction (MOR) approaches based on the proper orthogonal decomposition (POD) combined with a hyper-reduction (HR) method can be employed to reduce the computational time but often leads to numerical instabilities. In this article, we aim to compare two HR methods that were developed to improve the stability of the nonlinear reduced model: the Gaussian Newton augmented tensors (GNAT) method and the energy-conserving mesh sampling and weighting (ECSW) method. The reduced models are evaluated in terms of accuracy on global and local quantities of interest and in terms of speedup, with application to a squirrel cage induction machine. Improvement of the ECSW method has been proposed such that it appears that this method is more competitive than the GNAT method and enables to obtain very good results both in terms of accuracy and time computation, which was reduced by a factor of 40 compared to the FE model.