Ultrasonic Characterization of Compacted Salty Kaolin–Sand Mixtures Under Nearly Zero Vertical Stress Using Experimental Study and Machine Learning
Abolfazl Baghbani,Hossam Abuel-Naga,Roohollah Shirani Faradonbeh,Susanga Costa,Rayed Almasoudi
DOI: https://doi.org/10.1007/s10706-023-02441-5
2023-04-14
Geotechnical and Geological Engineering
Abstract:Understanding the dynamic behavior of soil is crucial for developing effective mitigation strategies for natural hazards such as earthquakes, landslides, and soil liquefaction, which can cause significant damage and loss of life. The ultrasonic wave testing method provides a non-invasive and reliable way of measuring the shear modulus, damping ratio and density of soils, which are fundamental parameters for understanding soil's dynamic characteristics. The aim of this study was to investigate the effects of environmental factors, such as water salinity, soil liquid limit, plasticity index, dry density, and water content, on ultrasonic wave velocities (specifically shear and primary waves) in kaolin–sand mixtures subjected to near-zero vertical stress, as well as to predict these effects utilizing two unique artificial intelligence methods, including Classification and Regression Random Forests (CRRF) and Artificial Neural Networks (ANN), which, to our knowledge, have not been utilized in previous literature. The CRRF and ANN models were developed using two well-known algorithms and five different architectures using a database of 128 datasets. Water salinity, dry density, water content, liquid limit and plasticity index were predictor variables. The results showed that both CRRF and ANN were highly accurate. The coefficient of determination (R 2 ) and mean absolute error (MAE) of the best CRRF were 0.963 and 9.191, respectively to predict V s , and 0.974 and 7.809 to predict V p , respectively. Furthermore, in ANN, R 2 and MAE were respectively 0.994 and 0.016 to predict both V s and V p . According to importance analysis, liquid limit, molality, and dry density are the most critical parameters, while water content is the least critical.