Sensitivity analysis of laterally loaded pile involving correlated non-normal variables

Abstract Probabilistic sensitivity analysis plays an important and practical role in many applications of reliability-based analysis, design and optimization. A new and efficient method to estimate probabilistic sensitivities is proposed and illustrated for the case of a laterally loaded pile involving correlated non-normal basic random variables and implicit limit state functions. Stochastic sensitivity analyses are carried out on pile-head deflection failure and bending moment failure. The soil medium is characterized by a one-dimensional random field. The proposed sensitivity factors are based on the Hasofer-Lind second-moment reliability index and the first-order reliability method (FORM), and can be extended to compute reliability sensitivities with respect to changes in the mean value or standard deviation. The method described herein is efficient because the sensitivity factors can be evaluated immediately after the design point and the direction cosines are obtained using FORM, without the need for further evaluations of the limit state function. Comparisons are made between the proposed approach and numerical differentiation based on reliability methods, as well as that based on Monte Carlo simulations with the use of importance sampling.