Understanding soil shear stiffness (G) and damping ratio (D) is essential for evaluating geotechnical stability under dynamic loads like earthquakes and machine vibrations. These parameters are critical for assessing site response, designing foundations, and ensuring overall structural stability. Traditional methods, such as cyclic simple shear, cyclic triaxial, and resonant column tests, often face limitations due to inadequate site sampling, potential sampling errors, and difficulties in replicating field conditions in the laboratory. These challenges necessitate numerous tests, which in turn escalate costs related to equipment and labor. In response to these issues, Artificial Intelligence (AI) and Machine Learning (ML) have emerged as promising alternatives. This paper reviews several ML techniques applied in the literature for predicting soil dynamic properties, including Artificial Neural Networks (ANN), Support Vector Machines (SVM), tree-based models, Evolutionary Algorithms (EA), and Fuzzy Logic-based (FL) models. The analysis demonstrates that ML models not only provide enhanced accuracy (coefficient of determination, R 2 , up to 0.994) compared to traditional empirical equations (R 2 typically below 0.85) but also effectively capture the complex behavior of soils under dynamic conditions. The study also examines the advantages and limitations of current ML-based models, identifying key challenges and future research directions in predicting G and D. The findings demonstrate that ML techniques enhance prediction accuracy, reduce testing costs, and improve efficiency, making them valuable tools for geotechnical engineering applications. This review provides guidance for engineers and researchers in selecting suitable ML models and optimization strategies to improve the prediction of soil dynamic properties. • The review explores Machine learning (ML) techniques to predict soil G and D. • Comprehensive review assesses ML accuracy and scalability for geotechnical use. • The study highlights ML limits and future needs, like big data and tuning. • Results show ML surpasses empirical formulas in nonlinear soil response. • Advances in ML enhance geotechnical design through better soil prediction.

Shear wave velocity (V s ) of soils is a crucial property in geotechnical engineering practice, affecting seismic site response analysis, seismic hazard assessment , and dynamic soil-structure interaction. The precise determination of V s is crucial in assessing the dynamic behavior of soils during seismic events, as it markedly influences the amplification and attenuation of ground motions. While several empirical equations have been proposed thus far for estimating V s in earthen materials, the majority of them lack the required accuracy and predictive capability. As a result, there has been a growing tendency among practicing engineers towards utilizing Artificial Intelligence (AI) for V s prediction. This paper presents an extensive overview of the developments in deploying AI and its subsets, including Machine Learning (ML) and Deep Learning (DL) techniques, for the precise estimation of V s in soil deposits. Notably, despite the importance of shear wave velocity as a key geotechnical parameter, no prior review study has exclusively focused on evaluating it using AI-based models. This review systematically examines various AI-based methodologies employed by researchers to enhance the reliability and precision of V s predictions using soil properties and in-situ test data. The advantages of ML techniques over conventional empirical correlations are thoroughly analyzed and critically compared. Additionally, the paper discusses the relative performance of different AI-based approaches, outlining their strengths and limitations in V s estimation. The review also provides a general qualitative assessment of V s measurement methods, offering guidance on selecting the most appropriate approach based on project-specific requirements and constraints. Finally, through a critical evaluation of existing literature, key knowledge gaps are identified, and potential directions for future research in this domain are proposed.

Blast-induced overbreak, characterized by the excessive removal of rock mass beyond the planned tunnel profile, poses significant safety risks, increases costs, and causes project delays during tunneling. Traditional machine/deep learning models have been developed to predict overbreak. However, these models are often inadequate because they are static and lack the flexibility to adapt to new, real-world data continuously. This study addresses this limitation by introducing a novel data-driven approach based on deep continual learning. The primary objective is to develop an adaptable predictive model with the ability of continual learning, which is particularly advantageous in dynamic environments like tunnel blasting. To achieve this, a self-attention multi-layer perceptron (MLP) model for overbreak prediction, integrated with two continual learning strategies (elastic weight consolidation (EWC) and memory replay (MR)), is developed. This step enables the overbreak prediction model to possess the ability to continuously learn real-world scenarios and adapt to the dynamic environment of tunnel blasting. The findings show that the continuous MLP model, empowered by EWC and MR, demonstrates superior adaptability and accuracy in predicting overbreak. Compared with the standard MLP model, which achieves a predictive accuracy of 0.831, the continuous MLP model achieves a predictive accuracy of 0.845 on unseen data. The integration of EWC and MR strategies proves to be a pivotal factor in developing deep learning models for the dynamic task of predicting overbreak. The continual learning strategies ensure that the models remain adaptable and accurate over time, which is essential for practical applications in dynamic environments of tunnel blasting operations.