How does deterministic safety analysis differ from prohabilistic safety analysis?
The classic (deterministic) safety analysis is entirely based on well-defined and rather extensive rules and guidelines, such as those of the US-NRC (Nuclear Regulatory Commission, USA) or of the IAEA (International Atomic Energy Agency, Vienna). It is documented in a safety report that describes the installation design and its operating conditions. In practice, the procedure is as follows:
First, all conceivable accidents that should be considered in the design basis of the installation are identified and grouped. For each accident category, “enveloping” scenarios are identified, so that the further analysis can be limited to those “enveloping” design basis accidents, taking into account conservative (and therefore penalizing) hypotheses for the initial conditions and the further course of the accident. The potential radiological consequences are calculated, also taking into account conservative hypotheses. Finally, the results of the analysis are compared to previously defined acceptance criteria, which allows to verify the adequacy of the safety system's design .
As such, the deterministic safety analysis takes into account numerous safety margins, but does not include an assessment of the so-called residual risk beyond the design basis. It is still used as a basis for the licencing of nuclear installations.
The probabilistic safety analysis (PSA, PRA) assumes that even rare events (or combinations of them) can occur. Single failures, as well as multiple failures, common cause failures (CCF) and human errors, can be considered, with their respective probability. In order to obtain usable results, these probabilities should be assessed as realistically as possible (best estimate). For nuclear power plants, the procedure is as follows:
For each operating condition of a nuclear power plant, a list of initiating events that could lead to an accident is established and their occurrence frequencies are determined. Event trees describe the development of different scenarios that either lead to a successful mission or to core damage (core melt). A reliability analysis (fault trees) is made for all systems involved. The probability of potential human errors important for these accident scenarios is considered. With the use of a computer, all these elements allow to estimate the core melt frequency per reactor year, and also to analyze the risk structure (per initiating event, per condition of the NPP, per failure category, and so on). The above represents a level 1 PSA. More elaborated studies continue to assess the level of release into the environment (level 2) or of the risk for the population (level 3).
The probabilistic safety analysis gives a good idea of the extent and the structure of the risk (also beyond the design basis). In order to define the accident scenarios, input from deterministic thermohydraulic studies with realistic hypotheses is required. However, the probability of some failures (such as human errors) cannot be estimated with a large accuracy, which, in combination with the applied work hypotheses, leads to a rather important uncertainty on the numerical end result. The probabilistic safety analysis is increasingly used in addition to (and not in replacement of) the deterministic safety analysis and allows for corrections in various fields (risk-informed practices).
also on this website: Deterministic and probabilistic safety analysis: to which extent are they complementary?