砂土液化预测的Fisher判别模型及应用

Fisher discriminant analysis model of sand liquefaction and its application

基于Fisher判别理论建立了砂土液化可能性的Fisher判别分析(FDA)模型。在分析砂土液化影响因素的基础上,选取烈度、震中距、地下水位、砂层埋深、标贯击数、平均粒径、不均匀系数、剪应力比等8个实测特征指标作为FDA模型的预测指标。利用砂土液化的实测数据作为训练样本进行训练,建立FDA模型对砂土液化进行预测,并用其他未参加训练的实测数据进行了验证。研究结果表明,FDA模型简便可行、预测精度高,是解决砂土液化预测问题的有效方法之一。

Based on Fisher discriminant theory,the Fisher discriminant analysis（FDA） model was established for forecasting the possibility of sand liquefaction.Eight factors were selected such as seismic intensity,epicenteral distance,groundwater level,sand depth,blow number of standard penetration test,mean granular diameter,coefficient of nonuniformity,ratio of shearing stress to effective overburden stress as the discriminant factors of the FDA model.A series of experimental data of sand liquefaction were taken as the training and testing samples,then some other practical data were used to verify this model.The results show that the FDA model is a simple,feasible and high accurate prediction method;and it is one of the efficient methods for solving prediction of sand liquefaction.