Stochastic design using cross-validated nonparametric meta-modelling techniques

This study adopts the quantile performance measures that address stochastic uncertainties in cae modelling and simulations to support design decision-making. Two nonparametric meta-modelling techniques are introduced in conjunction with a uniform sampling scheme to approximate nonlinear cae model input/output relationships. An integrated stochastic design framework that facilitates practical applications of time-consuming cae simulations is described. A vehicle restraint system example is used to demonstrate the effectiveness of the proposed framework and the enabling techniques.