Uncertainty quantification in DIC with Kriging regression



Wang, Dezhi, Diaz De La O, Francisco Alejandro, Wang, Weizhuo Victor, Lin, Xiaoshan, Patterson, Eann ORCID: 0000-0003-4397-2160 and Mottershead, John ORCID: 0000-0003-1279-2562
(2016) Uncertainty quantification in DIC with Kriging regression. Optics and Lasers in Engineering, 78. 182 - 195.

[img] Text
UQ_DIC_5August2015_Revision -Submission.pdf - Accepted Version

Download (1MB)

Abstract

A Kriging regression model is developed as a post-processing technique for the treatment of measurement uncertainty in classical subset-based Digital Image Correlation (DIC). Regression is achieved by regularising the sample-point correlation matrix using a local, subset-based, assessment of the measurement error with assumed statistical normality and based on the Sum of Squared Differences (SSD) criterion. This leads to a Kriging-regression model in the form of a Gaussian process representing uncertainty on the Kriging estimate of the measured displacement field. The method is demonstrated using numerical and experimental examples. Kriging estimates of displacement fields are shown to be in excellent agreement with ‘true’ values for the numerical cases and in the experimental example uncertainty quantification is carried out using the Gaussian random process that forms part of the Kriging model. The root mean square error (RMSE) on the estimated displacements is produced and standard deviations on local strain estimates are determined.

Item Type: Article
Uncontrolled Keywords: Digital Image Correlation, measurement error, Kriging regression, uncertainty quantification
Depositing User: Symplectic Admin
Date Deposited: 17 Aug 2016 14:48
Last Modified: 09 Jan 2021 14:10
DOI: 10.1016/j.optlaseng.2015.09.006
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3001955