Abstract: Field data from liquefaction case histories are important basis for the development, calibration, and validation of the liquefaction evaluation methods, and major standard for the validation of current liquefaction theories. Collecting liquefaction data from ChiChi, Bachu and Songyuan earthquakes, the Chinese standard penetration test (SPT)-based liquefaction database significantly increases from 121 to 465 in number. The dataset is used to validate the reliability of four liquefaction evaluation methods based on standard penetration test (SPT), i.e., the Chinese seismic design of building code method (code method), two hyperbolic models, and the cyclic stress ratio (CSR) simplified method. The results indicate that the two hyperbolic models can satisfactorily distinguish the liquefaction data from the non-liquefaction data, with success rates higher than 85% for both liquefaction data and non-liquefaction data. The code method and the simplified CSR method exhibit disadvantages for liquefaction evaluation. The predicted results are not satisfactory for all four methods for seismic intensity of 7, as the liquefaction data are mixed with the non-liquefaction data for this intensity. The overall success rates of the four methods are high for the data from seismic intensities of 8 and 9. A new probabilistic liquefaction evaluation formula based on the CSR method is proposed by regression analysis of the new liquefaction dataset. The predicted critical liquefaction lines are in reasonable agreement with reported probabilistic formula, even though the datasets ever adopted are different. The Chinese code method has obvious limitation with conservative results for soil layers at depth greater than 10 m. The analytical results provide reference for improving the liquefaction evaluation method in Chinese code.

Key words: sand liquefaction, database, standard penetration test, liquefaction evaluation, probability level