Capabilities of finite element analysis and magnetic equivalent circuits for electrical machine analysis and design

This paper reviews the literature concerned with capabilities and limitations of finite element analysis (FEA) and magnetic equivalent circuit (MEC) analysis for electrical machine design. The most common known models are based on equivalent circuits and related analytical models, or on FEA. Analytical models use highly simplified magnetics, and have difficulty extending into saturation. FEA typically uses magnetic vector potential representations that model additional effects such as eddy currents, but requires detailed nonlinear models for saturation and hysteresis. MEC methods represent a third possibility for electrical machine analysis, based on permeance network models comprising reluctances and mmf sources. Advantages of the MEC method include reduced model complexity compared to FEA, enhanced accuracy compared to analytical approaches, ease of parameterization, methods for extension to 3-D capability, and fast computation time. One of the most significant concerns related to literature in this area is that very few papers report thorough comparisons between experimental measurements and simulation tools for electromechanical devices. Among those that do, even fewer compare electromechanical forces and torques. With less than 15 exceptions, the few papers providing such comparisons report ldquogood agreement, rdquo but in fact show errors of 10% or more between tests and simulations. Many papers with experimental results do not compare torque or force results or present error analysis. This is most unfortunate, as many authors deen to consider FEA results ldquodefinitiverdquo and use them as a basis for model comparisons. Saturation and iron losses appear to be the likely culprits. In a few papers, the reported analysis method takes full nonlinear magnetic effects into account. When magnetic saturation, eddy currents, hysteresis losses, and similar effects are modeled with care and in detail, differences between simulations and experiments typically are on the order of 5%.