Uncertainty in System Codes - Current Status and Future Work in STARS

The starting point of the paper was a reminder that uncertainty has a statistical nature. It is therefore quantified by statistical measures: confidence intervals, variances, tolerance intervals, probabilistic density or cumulative probability functions. An uncertainty measure quantifies how reliable a code results are. Complex codes used in many engineering and scientific fields require a measure of the uncertainty of their results. Dr Macian described uncertainty research at PSI (based on RETRAN-3D) and uncertainty application in STARS, using the GRS approach (SUSA program developed by GRS for the statistical part of the methodology). He also described briefly the new research efforts under way at the Institute. The conclusion of the work was that the methodology under investigation should provide more accurate uncertainty information of the code results than currently available, and a solid basis for the generation of objective uncertainty information for physical models of complex codes. It was applicable to Best Estimate codes and to other complex codes in other scientific and engineering fields. A common disadvantage had been discovered in the above-mentioned uncertainty research: it was necessary to introduce subjective expert opinion. The proposed research plan for the development of the methodology was therefore to obtain objective, statistically sound uncertainty information for code physical models