Author(s)

ALI SHAFIEE, AMIR HOSSEIN SHAFIEE, AMIN FALAMAKI

Abstract

Liquefaction, a critical geotechnical hazard during seismic events, necessitates accurate assessment methods to mitigate associated risks effectively. Traditional approaches often exhibit limitations in capturing the intricate relationships between earthquake characteristics, soil properties, and liquefaction susceptibility. In this study, we present a novel methodology employing the adaptive neuro-fuzzy inference system (ANFIS) to comprehensively assess liquefaction potential, integrating earthquake parameters (magnitude and maximum acceleration) and soil properties (e.g., fines content, standard penetration test blow count). The ANFIS framework combines the adaptive learning capabilities of neural networks with the interpretability of fuzzy logic, enabling the modelling of complex and nonlinear relationships. By training the ANFIS model on a diverse dataset comprising seismic and geotechnical data, we developed a robust tool for predicting liquefaction susceptibility across varying geological and seismic conditions. Our analysis demonstrates the effectiveness of the ANFIS approach in accurately assessing liquefaction potential, surpassing traditional methods by considering multiple influencing factors simultaneously. Through iterative optimization, the ANFIS model yields precise predictions while accommodating uncertainties inherent in input data. Key findings highlight the significant influence of earthquake characteristics and soil properties on liquefaction susceptibility. This research contributes to advancing the understanding and prediction of liquefaction hazards, offering a valuable tool for engineers, planners, and policymakers involved in disaster risk management. The proposed methodology presents a promising avenue for further refinement and application in diverse geotechnical and seismological contexts, ultimately enhancing resilience against seismic events.

Keywords

liquefaction, adaptive neuro-fuzzy inference system (ANFIS), standard penetration test