Sensitivity Analysis of Inverse Thermal Modeling to Determine Power Losses in Electrical Machines

Abstract

Inverse analysis is a known mathematical approach, which has been used to solve physical problems of a particular nature. Nevertheless, it has seldom been applied directly for loss reconstruction of electrical machines. This paper aims to verify the accuracy of an inverse methodology used in mapping power loss distribution in an induction motor. Conjugate gradient method is used to iteratively find the unique inverse solution when simulated temperature measurement data are available. Realistic measurement situations are considered and the measurement errors corresponding to thermographic measurements and temperature sensor measurements are used to generate simulated numerical measurement data. An accurate 2-D finite-element thermal model of a 37 kW cage induction motor serves as the forward solution. The inverse model’s objective is to map the power loss density in the motor accurately from noisy temperature measurements made on the motor housing’s outer surface. Furthermore, the sensitivity of the adopted inverse methodology to variations in the number of available measurements is also considered. Filtering the applied noise to acceptable ranges is shown to improve the inverse mapping results.