This paper presents an adaptive stochastic isogeometric method to incorporate material uncertainties in the nonlinear bending analysis of thin functionally graded material (FGM) shells. The gradient index is modeled as a second-order random field to describe the spatial randomness of material properties. An adaptive, nested, and non-intrusive Chebyshev interpolation process based on Leja sequences is formulated to approximate the response surface concerning random inputs, in which the discrete responses are parallelly obtained and incrementally augmented to save computational effort. The response surface is taken as a surrogate model to promote the efficiency of the stochastic analysis. In particular, a new probabilistic post-processing technique is proposed to accelerate the evaluation of the probability characteristics of the response. Isogeometric analysis is applied to solve the discrete responses of FGM shells, eliminating geometric errors and simplifying the implementation. Through three examples with different computational costs, the effectiveness of the present method is demonstrated by comparing it with two forms of Monte Carlo simulation methods. The influence of the statistical characteristics of the gradient index field on the response surface analysis and the random response is investigated.
Read full abstract