Soil Liquefaction Assessment by CPT and V_S Data and Incomplete-Fuzzy C-Means Clustering

Assessing soil liquefaction potential is a crucial consideration in the seismic design of structures and their seismic stability. The complex nonlinear behavior of the liquefiable soils and the non-deterministic nature of earthquakes make the liquefaction analysis vague. Accordingly, researchers have progressively focused on employing machine learning and mathematical algorithms to address the complexities and uncertainties of evaluating soil liquefaction potential. This paper investigates the performance of fuzzy c-means clustering of incomplete data for assessing liquefaction potential based on cone penetration test (CPT) and shear wave velocity (V-s) field data. The research was conducted using two approaches: (1) whole data strategy; (2) partial distance strategy. The used database contains 786 CPT and 846 V-s records, with specified liquefaction conditions in past earthquake events. We compared the effectiveness and success of this method with traditional deterministic and probabilistic liquefaction evaluation approaches. It was found that the fuzzy c-means clustering model had a comparable predictive ability with other methods and would be reliable when assessing the liquefaction possibility.