LLDV-a Comprehensive Framework for Assessing the Effects of Liquefaction Land Damage Potential

摘要

The unprecedented liquefaction related land damage during earthquake has highlighted the need to develop a model that better interpret the liquefaction land damage vulnerability (LLDV) when determining whether liquefaction is probably to cause damage at the ground surface. This paper presents the development of a new comprehensive framework explored from select case history records using Bayesian belief network (BBN) methodology, to assess liquefaction land damage vulnerability. The BBN based liquefaction land damage vulnerability model has been developed by integrating multi-related factors of soil liquefaction and its induced hazards into one model on the basis of cone penetration test (CPT) case history records using K2 machine learning (ML) algorithm and domain knowledge (DK). The paper highlights that the proposed BBN is a substitutive novel LLDV framework and showed relatively better performance when compared with C4.5 decision tree (DT) - J48 model, naive Bayesian (NB) classifier, and BBN – K2 ML prediction methods in terms of overall accuracy (OA) and the Cohen’s kappa coefficient. The proposed BBN–K2 and DK model is simple to perform in practice and the results are likely to assist decisions on seismic risk mitigation measures for sustainable development and provide a step towards a more sophisticated liquefaction risk assessment modeling.