LM algorithm of BP neural network model for estimation of sand seismic liquefaction potential is built in the paper. Sorting the data with LVQ model sorting net and selecting 320 groups of field liquefaction data of 76 sites during Tangshan earthquake in China

data of each train and test neural network model are organized in 160 groups. It shows that the estimation veracity of the method is 96.8%. Based on the drill data of typical sites in Huai′an city

the sand liquefaction potential estimation method of the BP neural network model suggested in this paper

the method of code for seismic design of building in China （Code method）

the method suggested by the National Center for Earthquake Engineering Research of American （NCEER method） and the method suggested by Xie Junfei-Chen Guoxing （Xie-Chen method） are applied to the seismic liquefaction estimation of 120 standard penetration test data of 22 drills. According to the liquefaction estimation results

it can be shown that: ① the results obtained by the above four evaluation methods of sand seismic liquefaction potential are almost uniform; ② the result obtained by NCEER method is more conservative than that obtained by Xie-Chen method; ③ the result obtained by Xie-Chen method is close to that obtained by code method

and the correctness rates of the two methods are almost coincident; ④ the Code method is more conservative than the method of BP neural network model; ⑤ the results obtained by the method of BP neural network model is more identical with the field liquefaction rule than those obtained by other three methods. Considering the different estimation results of four methods

the comprehensive evaluation of sand seismic liquefaction potential in Huai′an city are obtained

enclosing the sand seismic liquefaction potential evaluation charts of some typical geology sections.