Efficient robust design via Monte Carlo sample reweighting

A novel probabilistic method for the optimization of robust design problems is presented. The approach is based on an efficient variation of the Monte Carlo simulation method. By shifting most of the computational burden to outside of the optimization loop, optimum designs can be achieved efficiently and accurately. Furthermore by reweighting an initial set of samples the objective function and constraints become smooth functions of changes in the probability distribution of the parameters, rather than the stochastic functions obtained using a standard Monte Carlo method. The approach is demonstrated on a beam truss example, and the optimum designs are verified with regular Monte Carlo simulation. Copyright (c) 2006 John Wiley & Sons, Ltd.