Methodical considerations for setting up space-resolved lumped-parameter thermal models for electrical machines

Abstract

Lumped Parameter Thermal Networks (LPTNs) are widely used to predict the thermal behavior of electrical machines. The networks usually have a simple structure and low number of nodes to limit the build-up effort, which reduces the spatial resolution of the thermal modeling. This paper presents a methodology that allows automated generation of high-order space-resolved LPTNs for electrical machines within minutes. The resistor network is automatically determined according to the structure of the machine even for complex geometries. In contrast to conventional LPTNs, the proposed thermal model is capable of predicting hot-spot temperature as well as the temperature at any user-specified location within the machine. The simulation results match well with the results of the precise but time-consuming finite element simulation. Meanwhile the calculation effort does not become prohibitive. These detailed thermal investigations can thus be routinely included in the standard machine-design process without considerably delaying it.