Thermal design of air-cooled axial flux permanent magnet machines

Abstract

Accurate thermal analysis of axial flux permanent magnet (AFPM) machines is crucial in predicting maximum power output, and a number of heat transfer paths exist making it difficult to undertake a general analysis. Stator convective heat transfer is one of the most important and least investigated heat transfer mechanisms and therefore is the focus of the present work. Experimental measurements were undertaken using a thin-film electrical heating method based on a printed circuit board heater array, providing radially resolved steady state heat transfer data from an experimental rotor-stator system designed as a geometric mockup of a through-flow ventilated AFPM machine. Using a flat rotor, local Nusselt numbers Nu(r) = hR/k were measured across 0.6<r/R< 1, as a function of non-dimensional gap ratio 0.0106 < G < 0.0467 and rotational Reynolds number 3.7e4 < Re [Theta]1e6 where G = g/R and Re [Theta] = [omega]R2/[Nu]. Averaged results Nu were correlated with a power law and it was found that Nu [is approximately equal to] ARe0.7 [Theta] in the fully turbulent regime (Re [Theta] > 3e5), with A being a function of G. In the laminar regime, stator Nu was found to be similar to that of the free rotor. Transition at the stator occurred at Re [Theta] = 3e5 for all G and is particularly marked at G < 0.02. Increased Nusselt numbers at the periphery were always observed because of the ingress of ambient air along the stator due to the rotor pumping effect. A slotted rotor was also tested, and was found to improve stator heat transfer compared with a flat rotor. The measurements were compared with computational fluid dynamics simulations. These were found to give a conservative estimate of heat transfer, with inaccuracies near the edge (r/R > 0.85) and in the transitional flow regime. Predicted stator heat transfer was found to be relatively insensitive to the choice of turbulence model and the two-equation SST model was used for most of the simulations.