Dimension-adaptive algorithm-based PCE for models with many model parameters

Abstract

Purpose

To present the models with many model parameters by polynomial chaos expansion (PCE), and improve the accuracy, this paper aims to present dimension-adaptive algorithm-based PCE technique and verify the feasibility of the proposed method through taking solid rocket motor ignition under low temperature as an example.

Design/methodology/approach

The main approaches of this work are as follows: presenting a two-step dimension-adaptive algorithm; through computing the PCE coefficients using dimension-adaptive algorithm, improving the accuracy of PCE surrogate model obtained; and applying the proposed method to uncertainty quantification (UQ) of solid rocket motor ignition under low temperature to verify the feasibility of the proposed method.

Findings

The result indicates that by means of comparing with some conventional non-invasive method, the proposed method is able to raise the computational accuracy significantly on condition of meeting the efficiency requirement.

Originality/value

This paper proposes an approach in which the optimal non-uniform grid that can avoid the issue of overfitting or underfitting is obtained.

Citation

Li, Y., Li, H. and Wei, G. (2019), "Dimension-adaptive algorithm-based PCE for models with many model parameters", Engineering Computations, Vol. 37 No. 2, pp. 522-545. https://doi.org/10.1108/EC-12-2018-0595