Experimental calibration in thermal analysis of PM electrical machines

Thermal design of electric machines frequently involves tests on a fully constructed prototype to calibrate various build factors associated with the manufacture, assembly and materials used in the hardware construction. The prototype machine is usually instrumented with multiple temperature sensors providing a detailed insight into the temperature distribution. The resolution of the experimentally gathered data is usually limited by the number of temperature sensors, and therefore the quality of model calibration is highly affected by the input data. This paper investigates the issue of thermal model calibration in the context of available machine hardware and measured data resolution. Also, the research evaluates the most suitable thermocouple location with reference to the model complexity, from reduced-order lumped-parameters circuit to high-fidelity finite element method (FEM). The investigation is focused on the stator-winding assembly, which is frequently associated with the main source of power loss within a PM machine body. A prototype of a PM generator has been selected to illustrate the effects associated with the model calibration. Tests on a representative stator-winding sub-assembly (motorette) have been used in the analysis. The results suggest that the measured data from alternative sensor locations for a given machine region has a significant impact on the quality of the model calibration and consequently temperature predictions.