Many engineering systems have attributes which exhibit complex patterns of variation in space. Such spatially random properties can be modeled as random fields. The effects of the random field models in some structures are strong and can adversely affect structure reliability. One of the most important features of structures involving random field models is that the response of the structure often does not have a closed-form solution. Stochastic finite element methods (SFEM) are emerging as powerful tools to analyze structural response and perform reliability analysis. A methodology for SFEM-based structure analysis and structural reliability assessment is developed in this study. This methodology improves the efficiency of SFEM by using a new approach to represent a random field as continuous function, and incorporates the expanded random field into the finite element formulation directly. A reduction procedure, similar to that employed in modal analysis in structural dynamics, is used to reduce the number of basic random variables. The method is used to investigate the effects of random field models on structure stability and reliability of linear and nonlinear structures, and to identify situations in which the use of a random field rather than a random variable model has a positive or negative impact on second moment-response or reliability. Errors in alternate discontinuous and continuous representations of random fields are examined in a second-moment sense. The effects of random spatial fluctuation either in material properties or the distributed loads are significant when the scales of the correlation are less than the lengths of the member. The random fluctuation in bending rigidity causes the mean and variability of the buckling load to decrease. Random fluctuations in material properties or the distributed loads cause the reliability of a structure to increase or decrease depending on the particular problem.
展开▼
