Lye Tee Siang, Adolphus 
ORCID: 0000-0002-1803-8344, Cicirello, Alice and Patelli, Edoardo
(2023)
An investigation towards the Uncertainty Model calibration approaches for NASA-Langley UQ Challenge 2019.
Journal of Physics: Conference Series.
|
PDF
MPSVA_Conference_2022 [Accepted].pdf - Author Accepted Manuscript Download (2MB) | Preview |
Abstract
The paper presents a series of analysis based on the recent NASA-Langley Uncertainty Quantification Challenge 2019 aimed towards calibrating an Uncertainty Model through a black-box model. In this research, 4 alternative Uncertainty Models are being proposed to investigate the following factors which contribute to the lowest degree of uncertainty over the aleatory and epistemic uncertain model parameters: 1) the choice of distribution of the aleatory model parameters; 2) the choice of the stochastic distance metric within the likelihood function to model the data variability; and 3) the choice of data type used within the likelihood function for the Bayesian model updating approach. To model the distribution of the aleatory model parameters, 2 distribution functions are considered: Beta vs Staircase Density Function. To quantify the variability of the input data used for model calibration, 2 types of stochastic distances are considered: Wasserstein’s distance vs Bhattacharyya’s distance. For the input data used for model calibration, 2 data types are considered: Time-domain vs Frequency domain. Based on the results, it was found that the Uncertainty Model incorporating the Beta distribution to model the aleatory model parameters, the Bhattacharyya’s distance as the stochastic metric, and the time-based data as the input data, yielded P-box estimates of the aleatory distribution and probabilistic estimates of the epistemic model parameters with the lowest degree of uncertainty.
| Item Type: | Article |
|---|---|
| Divisions: | Faculty of Science and Engineering > School of Engineering |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 17 Mar 2023 08:20 |
| Last Modified: | 17 Mar 2023 08:20 |
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3169135 |
Share
Share
