Electromagnetic FEA Applications
The design of electromagnetic devices has been a core part of our consultancy work for many years, and electromagnetic finite element analysis is an essential tool. The range of applications we encounter is large, and so it is important that the tools we use are general, and not limited to particular application areas. An example of where this can be problematic is in the use of model boundary conditions which can be used to exploit symmetry, reduce problem size, and therefore run times. Some electromagnetic FEA packages are aimed at certain industries and application areas, and even though they attempt to maintain a good level of generality, the prescribed types of boundary condition limit the kinds of symmetry that can be modelled. This is more commonly an issue in 3d modelling, simply because there is a bigger incentive here to reduce problem size through the use of symmetry. But if our tasks involve solving very large numbers of non-linear electromagnetic finite element models, then 2d problems also need to be optimised. There are always good reasons to minimise run-times!
Electromagnetic FEA Examples
Common symmetry types for electromagnetic FEA problems include flux-normal or flux-parallel boundaries, or a form of periodicity, typically that found in multi-pole rotating machines or actuators. A particularly unusual set of symmetry boundary conditions arose in the finite element modelling of a novel non-linear fault current limiter. I’ll put some further detail on the site soon, but for now it’s worth noting that at least one of the commercial codes out there couldn’t exploit the problem symmetry to its fullest extent.
A topic that surfaces from time to time is that of adaptive meshing, where software can automatically grade and refine a finite element mesh in order to better resolve local field behaviour. Some packages take this technique to the extent that the use of the finite element method, and the mesh design itself is almost completely hidden from the user. This is a justifiable approach for some classes of problem, but it has limitations. Some problems need regular, well designed meshes in order to deliver accurate results. Using electromagnetic FEA for the accurate prediction of cogging torques in PM motors is particularly challenging. I have yet to see this accomplished with adaptive meshing alone in all regions of geometry including air gap, though it is useful for generating an efficient mesh in less critical regions. This has been my experience, but if there’s anyone using adaptive meshing that can prove me wrong, I’d be interested to hear from you..
