Artificial Neural Network of Liquefaction Evaluation for Soils with High Fines Content

370 sets of standard penetration test (SPT) data are collected and synthesized, including the liquefaction and non-liquefaction cases generated during the Chi-Chi (M_w=7.6) earthquake of September 21,1999 and the earthquake (M_w=6.9) of June 11, 2000 in Taiwan. Other data are adopted from previous studies and the database. An approach based on Artificial Neural Network (ANN) technology was adopted and used to obtain the limit state curves for liquefaction evaluation. Probability of prediction and correlation coefficient are used to examine the correctness of the back-propagation neural network model. Limit state curves for fines content (FC) les 5%, FC=20%, and FC=35% are fitted from the critical state points from ANN modeling. A new limit state curve for a high fines content of 50% is also proposed and established. Each limit state yields a cyclic resistance ratio for a given set of soil resistance parameters. Examinations of all cases in the database show that the obtained critical limit state curves are able to predict the actual performance satisfactorily.