The limit state for liquefaction triggering\u00a0is obtained using a neural network-based searching technique developed\u00a0by Juang et al. \u00012000b\u0002. The technique involves the training\u00a0of supervised feed-forward neural networks with the \u201cfull\u201d\u00a0database of case histories and its subsets or samples. The successfully\u00a0trained neural network that generates the most accurate\u00a0input\u2013output relationship is adopted in the subsequent step for\u00a0searching \u201cdata points\u201d on the unknown boundary surface.\u00a0Regression analyses of the searched data points, with some engineering\u00a0judgment, yields the following empirical equation for\u00a0liquefaction resistance:<\/p>\n
CRR = exp\u0003(-2.8781 + 0.000309\u0001(qc1N,m\u0002)1.8\u0004 \u00011<\/sup>\u0002) \u00a0 \u00a0(1)<\/strong><\/p>\n where:<\/p>\n qc1N,m=stress-normalized cone tip resistance qc1N adjusted\u00a0for the effect of \u201cfines\u201d on liquefaction \u0001thus, qc1N,m=K\u00b7qc1N\u0002.<\/p>\n The \u00a0stress-normalized cone tip resistance qc1N used herein follows the\u00a0definition by Idriss and Boulanger \u00012004\u0002, although the difference\u00a0between this definition and that by Robertson and Wride \u00011998\u0002 is\u00a0rather small for the cases examined.<\/p>\n K is part of the regression model and expressed as:<\/p>\n