Severe accident analyses: A historical review from the very early days to the near-term future

Severe accident codes can be seen as the capitalization of severe accidents knowledge. Their high complexity comes from the intrinsic nature of these events. Thermal, fluid, mechanical, chemical and physical phenomena take place at the same time and strongly interact with each other in a scenario in which boundary conditions vary over a broad range and often become extreme. Besides, these events extend for days and even weeks, like in the Fukushima accident, which turns out to be an outstanding challenge for codes that have to integrate over long periods in which very fast and slow phenomena may be occurring at the same time. Even further, severe accidents are not restricted to a limited region of the plant, they usually encompass the entire plant and all the physical scales in which phenomena develop, from the micro to the macro one; in other words, the only way reasonable accuracy can be achieved in the mathematical integration is by meshing the entire spatial domain. Lastly, safety systems can drastically affect the accident progression, so that their performance should be also accounted for in codes under all possible conditions (human action included). As a consequence, major hypotheses, assumptions and approximations are in the essence of severe accident codes, either in their approach and/or in the specific models. This paper describes the evolution of methods used for severe accident analyses from the early days to present. A review of historical milestones in severe accident simulation is followed by an illustration of their current predictive capability by referring to major international benchmarking in reactor-like scenarios. Finally, uncertainty quantification is discussed as the next necessary step in severe accident analysis. The scope of the paper is focused on reactor scenario applications of integral tools, with a strong flavour of USA developments when describing the historical records of severe accident analyses.