[2D13-15] Uncertainty Analysis Method

[2D15] Inverse Uncertainty Quantification with the use of Deep Neural Networks

*Miki Saito¹, Atsushi Ui¹ (1. CRIEPI)

Keywords:Inverse Uncertainty Quantification, Statistical Safety Evaluation, Critical Flow, Deep Learning

In international projects related to statistical safety assessment, reducing uncertainty on the input side has been recognized as crucial for mitigating the variability and uncertainty of accident progression analysis results. Method development and implementation standards have been advanced by the OECD/NEA. Through projects like the OECD/NEA ATRIUM project, our team has worked on quantifying and refining input uncertainties through comparisons with measured values. This aims to reduce uncertainty in results and enhance predictability through the specific development of Inverse Uncertainty Quantification (IUQ) methods. This paper reports on the development of a deep neural network-based IUQ method and its applicability confirmation.