Practical random field discretization in stochastic finite element analysis

Stochastic finite element-based reliability analysis is applied to structures with distributed parameters that can be modeled as random fields. In this method, reliability is estimated through analytical computation of the sensitivity of stochastic response to the basic random variables. The random fields are discretized into sets of correlated random variables using two methods of discretization. The sensitivity measures are further used to selectively consider only a few of the distributed parameters as random fields, to ensure computational efficiency. The issue of choosing the appropriate mesh for the discretization of the random field is addressed through mesh refinement studies. With the help of three numerical examples, the paper examines the effects of the correlation characteristics of the random field on discretization and reliability analysis, and develops guidelines for efficient application of stochastic finite element analysis to structures with distributed parameters.