Uncertainty propagation of iron loss from characterization measurements to computation of electrical machines

The aim of the research is to find out how uncertainties in the characterization of magnetic materials propagate through identification and numerical simulation to the computation of iron losses in electrical machines. Design/methodology/approach The probabilistic uncertainties in the iron losses are modelled with the spectral approach using chaos polynomials. The Sobol indices are used for the global sensitivity analysis. The machine is modelled with a 2D finite element method and the iron losses are computed with a previously developed accurate method. Findings The uncertainties propagate in different ways to the different components of losses, i.e. eddy current, hysteresis, and excess losses. The propagation is also different depending on the investigated region of the machine, i.e. Stator or rotor teeth, yokes, tooth tips. Research limitations/implications The method does not account for uncertainties related to the manufacturing process, which might result in even larger variability. Practical implications A major implication of the findings is that the identification of iron loss parameters at low frequencies does not affect the loss variability. The identification with high frequency measurement is very important for the rotor tooth tips. The variability in the excess loss parameters is of low impact. Originality/value The presented results are of importance for the magnetic material manufacturers and the electrical machine designers. The manufacturers can plan the measurement and identification procedures as to minimize the output variability of the parameters. The designers of the machine can use the result and the presented procedures to estimate the variability of their desig