Robust optimization using a methodology based on cross entropy methods

To address the performance degradations of optimal designs arising from uncertainties using traditional concepts and methodologies, a robust oriented cross entropy method is proposed. To efficiently compute the solutions with robust performances, the normal distribution function is used as the probability density function, and a methodology for evaluating and assigning robust performance to promising solutions is proposed. A statistical model is introduced for the constraint functions to enhance the quality of the final design. To find the global and robust optimal solutions simultaneously in a single run, the original objective rather than the robust performance parameters is used for selecting the elite solutions. Two examples are reported to validate the proposed algorithm.