Numerical Analysis of Electromagnetic Fields by Proper Generalized Decomposition in Electrical Machines
The calculation of electromagnetic fields uses numerical tools such as the finite element method to accurately represent the real behavior. To obtain a technically relevant accuracy of the solution, many degrees of freedom are necessary. By using model order reduction techniques, the computational effort involved can be reduced. This DFG project investigated Proper Generalized Decomposition and discussed FE simulations of various low frequency electromagnetic problem classes.
In the scope of the basic research project (HA 4395/18-2), funded by the German Research Foundation, the Proper Generalized Decomposition was investigated as an a priori reduction technique. Subject of the research was the suitability and accuracy of the method with respect to diverse problem classes. The problem was decomposed into its individual parameters and then modes were enriched to approximate the influence of the individual parameters. Starting from transient problems, in which the electromagnetic fields are decomposed into temporal and spatial components, the method was extended to nonlinear and parametric problems. In the design process of electrical machines, many parameters are varied, such as the remanence of permanent magnets, materials or current combinations. The computational effort of the simulations can be significantly reduced with the parametric extension of the Proper Generalized Decomposition, while a technically relevant accuracy is given and characteristic quantities, such as the torque or the eddy current losses, are precisely determined.
To demonstrate the general applicability of the method, different potential formulations in two- and three-dimensional form were used. While the use of a magnetic vector potential corresponds to the state of the art, the combination with the T-Ω formulation represents an advantage especially in the investigation of dynamic three-dimensional fields under consideration of eddy currents.