In the PRA-guided design or safety-driven design, PRA is instituted from the very inception of the concept development, and is used continually and iteratively to evaluate and inform the design and identify safety-beneficial design changes. Obviously, any reactor development can benefit from the PRA-guided design, but there are specific technical and practical considerations that make SMRs more likely to benefit from this approach than evolutionary advanced large LWRs. Essentially, SMR designs tend to start with fewer well-defined details and preconceived solutions and are therefore more open when considering even significant modifications. At the same time, they are more conducive to novel design solutions (such as the integral reactor vessel) needed to address the identified safety weak spots.

The PRA-guided design as implemented for the IRIS reactor is illustrated in Figure 8.4 (Kling et al., 2005; Petrovic et al., 2012). The process is conceptually straightforward, but in practice involved tens of redesign iterations.

By consistently applying PRA to enhance and extend its safety-by-design approach, IRIS has lowered the predicted CDF to below 10-7 events/reactor-year and LERF to below 10-9 events/reactor-year. PRA analyses and design modifications are indicated in Figure 8.5.

The CDF for the initial design, after reviewing dominant cut-sets, was estimated to be ~2 X 10-6, a respectable number, but far from the IRIS target. In the next phase, marked in Fig. 8.5 as Step 1, sensitivity cases on individual significant factors (test intervals, diversity, reassessment) were performed and the design was modified, reducing the CDF to ~5 X 10-7. That was a limit achievable by optimizing single parameters. In the next phase (Step 2) more complex design changes were evaluated to understand and improve coupled processes by simultaneous optimization of several parameters, enabling reduction of CDF to ~1.2 X 10-8. The next Step 3 accounted for higher level of design details, which increased the CDF to ~2 X 10-8. Step 4 evaluated IRIS specific auxiliary systems, anticipated transients without scram (ATWS), human

reliability analysis, and further design details. Some weaknesses were identified that temporarily increased the estimated CDF, and then the design was improved, restoring the low CDF value of ~2 X 10-8. Step 5 indicates initial evaluation of external events. Every point on the graph represents an iteration including a PRA and a design modification. After implementing the PRA-suggested modifications to the reactor system layout, the preliminary PRA level 1 analysis estimated the CDF due to internal events (including ATWS) to be about 2 X 10-8, more than one order of magnitude lower than in current advanced LWRs. Such improvement would not have been possible by mere ‘engineering judgment’, without the benefit of systematic PRA-guided improvements, and without the design characteristics of iPWR SMRs.

Clearly, PRA limited to only internal events would be misleading, and in fact, the external events frequently become the limiting factors in iPWR SMRs, but the approach may be extended to external events as well, as for example in Alzbutas et al. (2005).