## Abstract

This paper presents an update of the Bayesian mapping function that relates nominal factor of safety (F̃_{S}) to probability of liquefaction (P_{L}) published by Juang and his colleagues in 2002. As part of this update, the existing simplified models for liquefaction probability are also assessed. An updated model, in the form of P_{L} = f(F̃_{S}), where f is an empirical function derived from the adopted database, is proposed. Insights gained in the assessment of the existing probability-based models are presented and discussed. The key findings include (1) the proposed model for liquefaction probability, developed using a simple and yet innovative approach, is shown to be able to yield results that agree well with those obtained using the more sophisticated maximum likelihood method and (2) if a data-driven model is developed based on true parameter values, one should use the true parameter values for prediction of a future case using the developed model; on the other hand, if the model is developed based on the nominal parameter values, one should use the nominal parameter values for future prediction. As nominal parameters are much easier to determine than true parameter values, models based on nominal parameter values are also easier to apply and less likely to be misused.

Original language | English |
---|---|

Pages (from-to) | 137-150 |

Number of pages | 14 |

Journal | Georisk |

Volume | 7 |

Issue number | 3 |

DOIs | |

State | Published - Sep 2013 |

## Keywords

- liquefaction
- maximum likelihood
- probability
- standard penetration test
- statistical methods