A Comparative Study on Probabilistic Optimization Methods for Electromagnetic Design

A reliability-based robust design optimization (RBRDO) is developed to ensure the product quality as well as the confidence in product reliability of electromagnetic devices. In the method, the first two statistical moments, mean and variance, of a quality loss function are estimated by the univariate dimension reduction method, while desired probabilistic constraint conditions are assessed by one of the first-order reliability analysis methods, called performance measure approach. For a better understanding between probabilistic design optimization methods, three different formulations of reliability-based design optimization, robust design optimization (RDO), and RBRDO are compared with one another. A simple mathematical model and a brushless direct current (BLDC) motor design problem are tested to demonstrate the features of the three methods and to examine their numerical efficiency.