Abstract: The liquefaction behaviors in the great Wenchuan 8.0 Earthquake in 2008 are quite notable, and the liquefaction of gravel soil is significant. Considering the wide distribution of gravel soil in some places in China, the liquefaction prediction methods should be developed. The existing methods for evaluating the liquefaction of sand soil result from the sand liquefaction cases, however, SPT technique can not be conducted in the gravel soil layers, and as a result, the existing code is not suitable for the liquefaction assessment of gravel soil. After the investigation for the liquefaction-induced damages in the great Wenchuan Earthquake and in-situ tests for the liquefied and non-liquefied sites, the liquefaction prediction method of gravel soil based on DPT, i.e., the dynamic penetration tests, is presented, and the corresponding model and formula are obtained. The analytical results indicate the liquefaction discrimination of the gravel soil can be divided into two steps, the initial discrimination and the second discrimination. In the initial step, the impossible liquefaction cases are selected, and in the second, the calculation model is adopted by using N120 as the basic index from DPT. The geological ages, buried condition of the gravel soil layer and gravel contents of gravel soils are considered in the initial discrimination. In the second discrimination, five parameters including the reference value of N120, gravel content of gravel soils, depth of gravel soils, water table and seismic intensity are concerned. Considering the wide range of liquefied soil depths and its water levels, the reference value of N120 is deduced by the normalization method and the influence coefficients of the gravel soil depths and the water levels are obtained by the optimal method. The effect of various factors on liquefaction possibility of gravel soil is considered in the present method and the advantages of the model and formula for the second discrimination are noticed by clear expression, high success ratio of regression discrimination, good connection with the past work and convenience in engineering application.