Abstract: Based on simply analyzing the back propagation algorithm, the principles of the BP neural network are applied to predicting sand liquefaction with eight factors listed as follows: seismic intensity, epicenter distance, mean diameter, coefficient of non-uniformity, underground water depth, sand depth, blow number of standard penetration test, ratio of shearing stress. Through computing practical examples and assessing the model, the model is manifested to be scientific and effective with much more accurate results than the Norm method and seed simplified method. The results show that the method is an efficient one in solving nonlinear problems.

Key words: artificial neural networks, BP algorithm, prediction, liquefaction of sand soil, assessment of model