Using seismic isolation to reduce risk and capital cost of safety-related nuclear structures

Highlights: • Seismic risk reduced significantly by the implementation of base isolation. • Isolation reduces capital cost at sites of moderate and high seismic hazard. • Innovative seismic risk assessment procedure can be implemented to infuse risk calculations into the design process. - Abstract: The implementation of seismic base isolation can substantially reduce horizontal seismic demands on structures, systems, and components (SSCs) in a nuclear facility, potentially providing significant benefits in terms of increased safety (smaller seismic risk) and reduced capital construction cost. Although increased safety of SSCs has been demonstrated previously, the possible reduction in their capital cost has not been explored. To quantify the reduction in risk enabled by isolation, nonlinear response-history analysis of a conventionally founded and a base-isolated model of a generic nuclear facility (GNF) is performed at the sites of the Idaho National Laboratory and the Los Alamos National Laboratory: sites of moderate and high seismic hazard, respectively. Seismic probabilistic risk assessment is performed to compute the mean annual frequency of unacceptable performance. The seismic risk is reduced by 7 to 8 orders of magnitude by the implementation of isolation. The costs of addressing seismic loadings are estimated for the GNF in both the conventionally founded and base-isolated GNF. The possible reductions in the required seismic ruggedness and in the cost of SSCs in the isolated GNF are quantified at both sites. A reduction in cost enabled by isolation is possible at nearly all sites of nuclear facilities in the United States, with the greatest benefit at sites of high seismic hazard, such as LANL. Two risk-calculation procedures are used in the assessment: a simplified method based on Boolean mathematics and a rigorous method based on Monte Carlo analysis. The simplified procedure, which is suitable for implementation with preliminary design calculations, produces accurate estimates of risk unless the mean annual frequencies of unacceptable performance are very small, measured here as smaller than 10⁻¹⁰. The sensitivity of the calculated risk in the conventionally founded GNF, to the choice of anchor period for the seismic hazard curve, is investigated and found to be insignificant over the range considered: 0 to 0.10 s.