For seismic probabilistic safety assessment of nuclear power plants (NPPs), test-based seismic fragilities for equipment can be developed using the results of dynamic testing. Electric Power Research Institute (EPRI) TR-103959  formulates test-based equipment fragilities by comparing broadband test response spectra (TRS) to broadband demand spectra. Alternatively, broadband capacity spectra based on earthquake experience can be compared to broadband demand spectra to develop seismic fragilities for equipment that meet the screening criteria of EPRI NP-6041-SL .
In-structure demand spectra can exhibit narrow spectral peaks. A clipping factor recommended by EPRI TR-103959  is typically used to clip these peaks to obtain median broadband demand spectra for comparison to broadband TRS. Similarly, a clipping factor recommended by EPRI NP-6041-SL  is typically used to obtain conservative deterministic broadband demand spectra for seismic margin assessment.
Four methods are investigated to assess their applicability for estimating clipping factor variability and structure response variability from probabilistic in-structure response spectra (ISRS) with narrow spectral peaks. Probabilistic ISRS at a given structural degree of freedom (DOF) are obtained from thirty simulations by the Latin Hypercube Sampling method. The most rigorous method involves independently clipping each of the thirty probabilistic ISRS. Structure response and clipping factor variabilities are calculated based on statistical analyses of the thirty clipped ISRS.
The other three methods are simplified alternatives to the more rigorous method and require less quantification effort. They involve calculating the median and 84th percentile non-exceedance probability (NEP) ISRS from the thirty unclipped ISRS. The clipping factors and associated clipping factor and structural response variabilities are calculated following the guidance of EPRI TR-103959  and EPRI NP-6041-SL .
The results of the four methods are compared, and recommendations for determining the total variability due to structure response and peak clipping are developed. Two approaches that take the lesser combined variability for structure response and clipping from two methods are found to reasonably approximate the more rigorous method.