A sample-based spectral method approach for solving high-dimensional SPDEs

TL;DR

This paper introduces a novel spectral method based on sampling to efficiently solve high-dimensional stochastic partial differential equations, overcoming the curse of dimensionality in uncertainty quantification.

Contribution

The paper develops a universal stochastic solution framework and a general algorithm that transforms high-dimensional stochastic PDEs into deterministic problems, significantly reducing computational costs.

Findings

Effective in high-dimensional stochastic problems

Outperforms traditional SFEM in computational efficiency

Applicable to linear and nonlinear SPDEs

Abstract

Uncertainty quantification appears today as a crucial point in numerous branches of science and engineering. In the past two decades, a growing interest has been devoted to stochastic finite element method (SFEM) for the propagation of uncertainties through physical models governed by stochastic partial differential equations (SPDEs). Despite its success and applications, the SFEM is mainly limited to small-scale and low-dimensional stochastic problems due to the extreme computational cost. In this article, by developing an universal construct of stochastic solution and a general algorithm for linear/nonlinear SFE equation, we explore a new strategy for the solution of high-dimensional stochastic problems, where stochastic problems are transformed into deterministic problems and stochastic algebraic equations. Since computational cost is almost proportional to the stochastic dimensionality of the problem, our method beats the so-called Curse of Dimensionality with great success. Numerical examples, including linear, nonlinear and high-dimensional stochastic problems, are used to demonstrate the method. Results show that our algorithm provides a highly efficient and unified framework for problems involving uncertainties, and is particularly appropriate for high stochastic dimensional problems of practical interest.