Variance-Based Sensitivity Analysis of Significant Design Parameters of an Induction Machine

A method to perform a variance-based global sensitivity analysis is presented which allows the investigation of significant design parameters of electrical machines. As matrices with random numbers are required for the calculation of the sensitivity indices, the most frequently used sampling methods which are the random, the latin hypercube and the Sobol' sequences sampling are discussed in particular. Furthermore, it is investigated which of the sampling methods provides the best coverage of the sampling space. The design parameters are generated randomly from the samples, which can result in unfeasible motor designs. For this purpose a description of the induction machine with squirrel cage rotor using geometric ratios is presented. Finally, a variance-based global sensitivity analysis is performed for the power losses to show its applicability to squirrel-cage induction motors.