ABSTRACT
A new dimension-reduction method based on most probable point (MPP) is presented for predicting reliability of mechanical systems subject to random loads, material properties, and geometry. The method involves univariate function representation at MPP, approximate response surface generation using dimension reduction, and Monte-Carlo simulation. The method yields higher-order approximation of a performance function without requiring any gradients. Results of two numerical examples involving an elementary mathematical function and a structural truss problem indicate that the proposed method provides accurate and computationally efficient estimates of the probability of failure.