Abstract:

In order to investigate various factors to affect the liquefaction of sandy soil.different BP neural network models for liquefaction evaluation were established based on different combinationS 0f the input neurons.A BP neural network-based program,namely SLV,was then presented for the liquefaction evaluation of saturated sandv soil.Some comparison analyses were finally carried out according to the results of field observation.It is shown that the earthquake action is a direct cause of the liquefaction ,that the saturation state of sandy soil is the precondition of the liquefaction,and that the standard penetration blow.count,the non.uniformity coefficient and the shearing stress ratio are the main influencing factors of the liquefaction.Moreover , it is indicated that the proposed BP neural net.work model can efectively evaluate the factors to affect the liquefaction of sandy soil due to its strong classifying and distinguishing ability.

Key words: sand, liquefaction, influencing factor, evaluation, neural network, Vogl fast algorithm