Stochastic Finite Element Formulation of a Three-Node Quadratic Bar Element with Non-Uniform Cross-Section Based on the Perturbation Method for Simultaneously Non-Deterministic Elastic Modulus and Applied Load

Abstract

<div class="section abstract"><div class="htmlview paragraph">The finite element method (FEM) is one of the most robust tools in structural analysis. Typically, the input parameters in a finite element model are assumed to be deterministic. However, in practice, almost all material and geometrical properties, including the load, possess randomness. The consideration of the probabilistic nature of these quantities is essential to effectively designing a system that is robust against the uncertainties arising due to the variation in the input parameters, the significance of which has been emphasized by space agencies like NASA. Among the various techniques applicable for stochastic analysis, the perturbation method, which is based on a sound mathematical foundation derived from Taylor’s series expansion, is widely acknowledged for its much higher efficiency compared to the well-known Monte-Carlo method. With this motivation, this work presents a stochastic finite element formulation of a bar element, commonly used in aerospace structural simulations. To the best of the authors’ knowledge, this is the first time a formulation will be put forth for the case of a three-node quadratic bar element with a non-uniform cross-section based on the second-order perturbation method for a simultaneously non-deterministic elastic modulus and distributed axial load with the discretization of the stochastic problem based on the averaged mean-square error for the approximation of the random fields. Subsequently, the finite element implementation shall be elucidated, followed by a validation via the Monte Carlo method, which includes a results-based discussion on the validation and investigation of the behaviour of the moments of the computed structural response in a sample problem for varying randomness of the inputs, complementarily endorsing the high efficiency of the method compared to the Monte Carlo method.</div></div>