RELIABILITY ANALYSIS OF DESIGN PROBLEM IN PRESENCE OF UNCERTAINTY USING STOCHASTIC RESPONSE SURFACE WITH CORRELATED PARAMETERS

摘要

The reliability analysis is important in structural engineering. Stochastic Response Surface gives an efficient platform to evaluate the probability of failure of model. Few sample runs are enough to characterize the system as it takes the value from high probability region. Hermite polynomial as the basis polynomial expansion was chosen for Stochastic Response Surface. Model input uncertain parameters were correlated and handled using Cholesky decomposition for evaluating the probability of failure when parameters were normal in nature. Non-normal correlated parameters were further addressed by Nataf transformation. An example of transport model and heat transfer problem for probability of failure is presented to demonstrate the capability of Stochastic Response Surface. It shows that it can handle correlated parameters accurately and efficiently.