A New Methodology for Robust Optimizations of Optimal Design Problems Under Interval Uncertainty

To consider the interval uncertainty in a practical optimal design problem, a new methodology for efficient robust optimizations is proposed. The proposed methodology uses a constrained formulation for robust performances not only in alleviating the inefficiency of the existing approaches in modeling interval uncertainties but also in avoiding the deficiency in the biasing force selection. The gradient information is used as both a trigger to activate the uncertain quantification procedure and the steepest increment direction to develop a fast searching phase. The stochastic approximation method is employed to minimize the computational burdens in computing the gradients. The numerical results on a case study are reported to validate the proposed methodology.