In case of some nuclear power plants constructed at the soft soil sites, liquefaction should be analysed as beyond design basis hazard. The aim of the analysis is to define the postevent condition of the plant, definition of plant vulnerabilities, and identification of the necessary measures for accident management. In the paper, the methodology of the analysis of liquefaction effects for nuclear power plants is outlined. The procedure includes identification of the scope of the safety analysis and the acceptable limit cases for plant structures having different role from accident management point of view. Considerations are made for identification of dominating effects of liquefaction. The possibility of the decoupling of the analysis of liquefaction effects from the analysis of vibratory ground motion is discussed. It is shown in the paper that the practicable empirical methods for definition of liquefaction susceptibility provide rather controversial results. Selection of method for assessment of soil behaviour that affects the integrity of structures requires specific considerations. The case of nuclear power plant at Paks, Hungary, is used as an example for demonstration of practical importance of the presented results and considerations.

The seismicity of Hungary can be considered moderately active, nevertheless contemporary reports from the past approx. 350 years documented surface manifestations of liquefaction occurrences. The last such earthquake was the 1956 Dunaharaszti ground motion, for which the location of two liquefied sites could be identified approx. 60 years after the event. This provided an excellent opportunity to analyze possibly the only accessible liquefied sites in Hungary. Analysis of the two sites included field and laboratory tests allowing the back-calculation of maximum horizontal ground acceleration of the earthquake. This parameter was previously unknown because the closest seismometer saturated during the event. The performed back-analysis using the principles of paleoliquefaction studies was the first of such analyses in the country. In areas with low to moderate seismicity, geotechnical engineers often neglect and overlook liquefaction hazard, however, when it is addressed, the hazard is often overestimated due to improper characterization of the seismic loading and site characterization. To explore this observation more deeply, probabilistic seismic and liquefaction hazard assessment were carried out at the two liquefied sites and it was found that this conclusion is also valid for Hungary, but the degree of conservatism of the pseudo-probabilistic procedures decreases with increasing earthquake return period (lower annual probability of occurrence).

Numerous methods exist to determine the liquefaction potential of a site due to earthquake, from which the stress-based empirical methods are the most commonly used in practice. Despite of their widespread use, their shortcomings have given way to the evolution of strain and energy-based methods. Their benefits make them a very promising alternative candidate for liquefaction potential evaluation. To reveal differences and uncertainties involved in the different methods, a comparative analysis was performed for the site of the Paks Nuclear Power Plant with the aim of contributing to the safety assessment of the plant with respect to liquefaction effects.

Empirical liquefaction potential assessment is generally based on the results of CPT, SPT or shear wave velocity (VS) measurement. In more complex or high-risk projects CPT and VS measurement are often performed at the same location commonly in the form of seismic CPT. However, combined use of both in-situ indices in one single empirical method has been limited. After the compilation of a case history database, the authors have developed a combined probabilistic method where the results of CPT and VS measurement can be used in parallel. The goal of this paper was to evaluate the prediction capability of the developed equation on an independent dataset of the 2010-2011 Canterbury Earthquake Sequence and to compare it with commonly used empirical procedures. It was found that the error index defined to quantify the false predictions is the largest for the recommended method but regarding the number of false predictions, it outperforms the other methods used for comparison.