Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
Nuclear: Model and Data Uncertainty
Download citation file:
- Ris (Zotero)
- Reference Manager
This paper discusses a general thermal-hydraulics uncertainty analysis methodology with application to the Marviken test facility blowdown transient scenario. The methodology, developed as part of a broader research by the authors, is for applications to best estimate (BE) analyses by complex thermal hydraulics (TH) codes. Developing a comprehensive method to make the TH systems codes capable of supporting the uncertainty assessment with the ability to handle important accident transients is the goal of this research. It considers the TH code as a “white box” explicitly treating internal sub-model uncertainties, and propagation of such model uncertainties through the code structure. Also explicitly treated are the various input parameter uncertainties. Bayesian techniques are used to incorporate available data (experimental and expert opinion) to estimate the parameter uncertainties. The code output is further corrected through the use of a Bayesian method with available experimental data at the integrated test facilities. There are various sources of epistemic uncertainty in TH code predictions. These include model uncertainties covering conceptualization and use of surrogate models and their implementation. In assessing corresponding uncertainties, possible sources of information include results of code validation experiments and expert judgments are used. For this purpose a Bayesian framework for model uncertainty is developed. Integration of parameter uncertainty can be done by accounting for code input and internal parameter variabilities. Mixing aleatory and epistemic uncertainties, and representation of uncertainty bound with physical meaning (history trace) ather than a statistical representation are among the issues tackled in this research. The proposed methodology uses an efficient Monte Carlo sampling technique for the propagation of uncertainty. Because of the computational intensity of the Monte Carlo simulations, especially when applied to complex TH calculations, Wilks' sampling criteria of tolerance limits is used to significantly reduce the number of Monte Carlo type iterations required, depending on the accuracy required. This paper describes the key elements of the uncertainty methodology developed, and summarizes its application to the Marviken test facility chocked flow transient.