Sensitivity analysis using Sobol indices for the thermal modelling of an electrical machine for sizing by optimization

Abstract

Purpose The thermal modelling of an electrical machine is difficult because the thermal behavior depends on its geometry, the used materials and its manufacturing process. In the paper, such a thermal model is used during the sizing process by optimization of a hybrid electric vehicle (HEV). This paper aims to deal with the sensitivities of thermal parameters on temperatures inside the electrical machine to allow the assessment of the influence of thermal parameters that are hard to assess. Design/methodology/approach A sensitivity analysis by Sobol indices is used to assess the sensitivities of the thermal parameters on electrical machine temperatures. As the optimization process needs fast computations, a lumped parameter thermal network (LPTN) is proposed for the thermal modelling of the machine, because of its fastness. This is also useful for the Sobol method that needs too many calls to this thermal model. This model is also used in a global model of a hybrid vehicle. Findings The difficulty is the thermal modelling of the machine on the validity domain of the sizing problem. The Sobol indices allow to find where a modelling effort has to be carried out. Research limitations/implications The Sobol indices have a significant value according to the number of calls of the model and their type (first-order, total, etc.). Therefore, the quality of the thermal sensitivity analysis is a compromise between computation times and modelling accuracy. Practical implications Thermal modelling of an electrical machine in a sizing process by optimization. Originality/value The use of Sobol indices for the sensitivity analysis of the thermal parameters of an electrical machine.