The Application Of BP Neural Network Based On MATLAB In Evaluation Of Loess Seismic Liquefaction

The main distribution area of loess is also the area of macroseism,which, in the history, the big earthquake has happened and initiated the massive loess liquefication and landslide, created seriously earthquake disaster .In recent years, the investigation and the laboratory test have indicated that the liquefication and the landslide could occurred under the high water content and the certain seismic action.This thesis introduced the mechanism of loess liquefaction, the methods of liquefaction judgment, the liquefaction condition as well as the influential factors. It is difficult to propose a accurate formula of liquefaction because the factors of the loess liquefaction are complex, multiple and non-linear. The artificial neural networks are composed of the massive neurons which orgnize non-linear auto-adapted dynamic system , can promulgate the non-linear relations which the data sample contains. Those system are provided with the strive ability of study, association, fault-tolerant and antijam, may facilitate modeling to the multi- origins and complex ignorance factor in varying degree and level.The specialists who grasp the computer programming language can establish BP network model ,which does not favor the promotion of the nerve network technology in a way. The MATLAB software has provided an available nerve network toolbox (NNT). This article utilizing the nerve network toolbox (NNT), considering the characteristic of loess liquefication, has established two BP network modeles which distinct the potential and damage degree of the loess liquefication, and elaborated the feasibility and the use value of BP network in the loess liquefication referring to the Tangshan earthquake and the overseas liquefied materia.By this thesis, we may draw the conclusion that it is easy and effective to establish BP network model by utilizing the nerve network toolbox based on MATLAB to evaluate the liquefication of the loess, which needn't carry on the tedious programming. Especially, this methold has the very good versatility and the use value.