12.231. To assess risks, we need to relate failure probabilities with con­sequences. For this purpose, the event tree (§12.220) links an initiating event, with a probability determined by fault tree analysis, through a chain of events with evaluated probabilities, leading to radioactive release to the environment and transport to populated areas. Finally, the dosage received and known health effects provides a measure of the consequences of the initiating event. An example of such a result is shown in Fig. 12.18, which is a risk curve from the 1975 Reactor Safety Study. Here, the ordinate is the frequency (or probability) that fatalities of magnitude X or greater will be produced.

12.232. Modeling of the various steps leading to risk assessment results in a considerable challenge. We need to consider the source term and severe accident progression steps, atmospheric transport, and finally, health effects. Therefore, results such as those shown in Fig. 12.18 have substantial uncertainty bands, the magnitude of which have been debated. Typically, such bands cover about two orders of magnitude. However, since the failure probabilities are so small, the reactor risks are still orders of magnitude less than natural or other industrial risks.

12.233. As mentioned, the pioneering reactor risk assessment study was the 1975 Reactor Safety Study. Subsequent critiques identified various weaknesses in the methodology and questioned the confidence levels es­timated and some of the results. However, follow-up studies tended to confirm the general order of magnitude levels of the RSS results. Consid­ering that the RSS prediction of a meltdown probability of once in 20,000 reactor-years of operation, it is logical to conclude that the systems are acceptably safe, even if there is an order-of-magnitude uncertainty in the probability prediction.

12.234. A comprehensive risk assessment study with emphasis on severe accidents was completed in 1990 [31]. This effort, commonly referred to as the NUREG-1150 study was the most ambitious since the 1975 Reactor Safety Study (RSS). It utilized many methodology improvements devel­oped since the earlier study, particularly in the areas of core damage eval­uation, source terms, containment event tree development, and conse­quence modeling. New codes such as CRAC2 for consequence modeling,

including atmospheric dispersion, were utilized [32]. An important new feature was the use of an expert system approach as part of the uncertainty analysis.

12.235. The results, developed for five representative LWRs, predict risk levels somewhat lower than those estimated in the RSS. However, the overall uncertainty range envelopes the results of both studies. A great deal of detail was provided regarding the behavior of each of the plants studied and possible safety-related backfits to reduce some of the uncer­tainties identified.

12.236. Although general confirmation of earlier study meltdown prob­ability predictions was important, the NUREG-1150 study’s primary sig­nificance was to provide a much improved picture of plant component performance under severe accident conditions that provides guidance for design and regulation. Possible equipment backfits for each plant were identified and their cost-effectiveness determined. Information on the harsh environment during an accident provided a new basis for equipment qual­ification as part of the regulatory process. In fact, the study provides a basis for the reexamination of many safety-related equipment licensing requirements.