Quantifying uncertainties in computer models An overview - Merlin KELLER - EDF R&D

Value of the Information & Strength of the Argument

The talk provides a high-level but comprehensive overview of UQ methodology, emphasizing practical applicability in industrial settings. The argumentation is clear and logical, building from motivation to methodology to advanced topics. The speaker effectively illustrates concepts with real-world examples from nuclear safety and hydrology, enhancing credibility. The value lies in its synthesis of established methods and its emphasis on a unified framework that can be applied across various problems. The argumentation is solid, though it does not delve into deep technical details, which is appropriate for an overview talk.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates scientific rigor by referencing established guidelines (e.g., NRC guides, metrology guides) and mentioning collaborative efforts with academic and industrial partners. The speaker mentions specific tools (OpenTURNS, Persalys, Uranie) and publications, though no direct URLs are provided in the description. The title accurately reflects the content, which is a broad overview of UQ. The talk is well-structured and credible, though it lacks explicit citations to specific papers, which is common for such overview presentations.

The title accurately reflects the content: a comprehensive overview of uncertainty quantification in computer models.

8/10

Talk by an expert from EDF R&D, presenting established UQ methodology and referencing institutional guidelines and tools. No formal peer review, but high practical authority.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and motivation for UQ in numerical simulations
• 05:00Challenges in UQ for nuclear safety and asset management
• 10:00Overview of the unified UQ methodology and its steps
• 15:00Step 1: Defining the quantity of interest (QoI)
• 20:00Step 2: Probabilistic modeling of uncertain inputs, including copulas and extreme value theory
• 25:00Step 3: Uncertainty propagation methods (Taylor, Monte Carlo, rare event simulation)
• 30:00Step 4: Sensitivity analysis - screening and global sensitivity measures
• 35:00Advanced methods: meta-models, robust optimization, and open questions
• 40:00Conclusion and summary of key takeaways

• OpenTURNS — Mentioned as an open-source UQ toolbox developed by EDF and others.
• Persalys — Graphical interface for OpenTURNS, mentioned in the talk.
• Uranie — Complementary software developed by CEA, mentioned in the talk.

• OpenTURNS — The talk mentions OpenTURNS as a tool for UQ; the official documentation confirms its capabilities.