Statistical analysis of the variation of floor vibrations in nuclear power plants subject to seismic loads

Floor vibration of a reactor building subjected to seismic loads was investigated, with the aim of quantifying the variability of vibrations on each floor. A detailed 3D building model founded on the bedrock was excited simultaneously in three directions by artificial accelerograms compatible with Finnish ground response spectra. Dynamic simulation for 21 s was carried out using explicit time integration. The extracted results of the simulation were acceleration in several floor locations, transformed to pseudoacceleration (PSA) spectra in the next stage. At first, the monitored locations on the floors were estimated by engineering judgement in order to arrive at a feasible number of floor nodes for post processing of the data. It became apparent that engineering judgment was insufficient to depict the key locations with high floor vibrations, which resulted in un-conservative vibration estimates. For this reason, a more systematic approach was later considered, in which nodes of the floors were selected with a more refined grid of 2 m. With this method, in addition to the highest PSA peaks in all directions, the full vibration distribution in each floor can be determined. A statistical evaluation of the floor responses was also carried out in order to define floor accelerations and PSAs with high confidence of non-exceedance. The conclusion was that in-floor variability can be as high as 50-60% and models with sufficiently dense node grids should be used in order to achieve a realistic estimate of floor vibration under seismic action. The effects of the shape of the input spectra, damping, and flexibility of the floors on floor vibration of nuclear power plant buildings were also investigated.