Does a reduced model reproduce the uncertainty of the original full-size model?

摘要

The present study addresses the title question by taking a detailed mechanism of a concrete combustion system, syngas ignition, deriving from it three reduced models with different levels of accuracy, and comparing propagation of uncertainty by the reduced and detailed models. Two reduced models were developed by the method of Detained Reduction and the third one was adopted from recent literature. The uncertainty quantification was carried out through the deterministic framework of Bound-to-Bound Data Collaboration (B2BDC). The numerical results demonstrate that assessment of the quality of a reduced model without considering parameter uncertainty may be misleading. By including parameter uncertainty, several numerical measures can be developed to quantify the reduced model performance and those tested in the present study showed mutually consistent and qualitatively similar outcomes. One of such measures, built on the B2BDC methodology, offers a numerically-efficient approach to quantifying the propagation of uncertainty and its sensitivity measures for models having time-demanding evaluations.(c) 2020 The Combustion Institute. Published by Elsevier Inc. All rights reserved.