Vs-based Assessment of Soil Liquefaction Potential Using Ensembling of GWO–KLEM and Bayesian Theorem: a Full Probabilistic Design Perspective

Shear wave velocity (Vs) offers engineers a promising alternative means to evaluate liquefaction resistance of sandy soils. However, the Vs-based probabilistic evaluation of soil liquefaction using machine learning method or Bayesian network method is limited. To this end, an attractive probability framework incorporating grey wolf optimization optimized kernel extreme learning machine (GWO–KLEM), and Bayesian theorem was developed for Vs-based probabilistic liquefaction model. The framework mainly includes two components: (1) an GWO–KLEM-based metamodel is developed to approximate the limit state function and identify the uncertain location of the boundary between the liquefaction and the non-liquefaction points; (2) the Bayesian theorem along with total probability theorem is used to integrate the prior knowledge with the site-specific Vs data for the updated distributions of input parameters. Then, a large number of equivalent samples from the above integrated knowledge can be generated based on the Markov chain Monte Carlo simulation; thus the probability of liquefaction can be calculated. Based on searching technique, perspective under simplified framework is also developed. The results show that limit state function searched by GWO–KELM is reliable and the proposed Vs-based probabilistic liquefaction model has a very good performance. Details on the development of the proposed system are presented, along with comparisons of the results obtained by this system with existing methods. Overall, the proposed method is advantageous and reliable for probabilistic assessment of liquefaction potential.