Seismic Geotechnical Engineering: Soil Liquefaction and Soil–Structure Interaction Between Traditional Approaches and Modern Techniques

  Share :          
  74

Prepared by: Asst. Lect. Lina Hussein Ali Rapporteur, Department of Civil Engineering, College of Engineering, Al-Mustaqbal University Introduction The performance of structures under earthquake loading depends not only on the strength of their structural elements, since the behavior of the soil that supports them plays a decisive role. Past earthquakes have shown that the response of soil to dynamic loads has a significant effect on structural damage. This article addresses two central phenomena in this field: soil liquefaction and soil–structure interaction. It presents the traditional and modern approaches used to assess them, and then discusses the challenges that remain. First: Soil Liquefaction and Methods of Assessing It Liquefaction refers to the loss of soil strength caused by a rise in pore water pressure during shaking, and it is among the most dangerous seismic phenomena for the safety of structures. The Cone Penetration Test (CPT) is one of the most widely adopted field tests for assessing liquefaction susceptibility, owing to its consistent and repeatable measurements over the full depth of the investigated layers. The Standard Penetration Test (SPT) is also used alongside it, and numerous empirical methods for assessing liquefaction have been built on both tests. Machine Learning in Liquefaction Assessment With the accumulation of field-test data and records from past earthquakes, machine learning approaches have emerged as a strong alternative or complement to empirical methods. Their main advantage lies in their ability to capture nonlinear relationships. Several studies have treated the problem as a binary classification (liquefaction or no liquefaction), comparing three classifiers, namely Support Vector Machines, Decision Trees, and Quadratic Discriminant Analysis, using CPT data. In another comparative study, two machine learning methods reached accuracies of about 97.6% and 97.5%, with a high degree of consistency in the importance ranking of the influencing variables. This has been helped by the availability of extensive databases, such as the Next Generation Liquefaction (NGL) database and the New Zealand Geotechnical Database, which have provided unprecedented access to data. Challenges Despite these results, a review of seventy-five publications concluded that artificial intelligence models for liquefaction have remained largely overlooked by practitioners and researchers alike. A comprehensive review covering the period 1994–2021 likewise confirms the need to update liquefaction databases and modify existing algorithms so that they become more computationally efficient and more reliable. High accuracy on test data does not necessarily mean engineering acceptance, since the designer needs a model whose results can be interpreted and defended. Second: Soil–Structure Interaction When a structure is subjected to an earthquake, the structure and the soil influence each other. In design practice, the seismic response of buildings is usually calculated by assuming that the structure is fixed at its base, which is an oversimplification that is valid only when the structure rests on very stiff soil. The interaction takes two main forms: kinematic interaction, resulting from the difference in stiffness between the foundation and the soil, and inertial interaction, which arises from the inertia of the vibrating structure, increasing the base shear and overturning moment and causing additional deformation in the soil. This interaction has traditionally been regarded as beneficial to the seismic response. However, recent studies and evidence from past earthquakes have shown detrimental effects under some conditions. This is attributed mainly to the nonlinearity of soil behavior. The issue remains contested: the role of interaction in the response of structures built on soft soil is still a matter of debate among researchers and in the provisions of design codes. As for modeling, a recent review on industrial machine foundations highlighted that the response is affected by factors including embedment depth, soil stiffness, and foundation properties, and that the finite element method is particularly effective in analyzing settlement under reciprocating loads. Conclusion and Recommendations The reviewed studies show that the assessment of seismic geotechnical hazards is shifting from reliance on simple empirical relationships to data-driven models and numerical simulation, and that each approach has its limitations. The main recommendations can be summarized as follows: Updating databases: Expand and document field liquefaction databases to include a wider range of soil types and earthquakes. Interpretability: Prefer machine learning models whose outputs can be interpreted from an engineering standpoint, so that they earn designers’ trust. Not neglecting interaction: Include the effect of soil–structure interaction in the analysis of structures built on medium or soft soil, instead of settling for the fixed-base assumption. Integration between the two disciplines: Strengthen cooperation between the geotechnical engineer and the structural engineer from the preliminary design stage. References Bapir, B., Abrahamczyk, L., Wichtmann, T., & Prada-Sarmiento, L. F. (2023). Soil-structure interaction: A state-of-the-art review of modeling techniques and studies on seismic response of building structures. Frontiers in Built Environment, 9. https://doi.org/10.3389/fbuil.2023.1120351 Noman, B. J., & Albusoda, B. S. (2024). Effect of soil-structure interaction on the response of machine foundation subjected to seismic loading: A review study. Journal of Engineering, 30(4). https://doi.org/10.31026/j.eng.2024.04.10 Hanandeh, S. M., Al-Bodour, W. A., & Hajij, M. M. (2022). A comparative study of soil liquefaction assessment using machine learning models. Geotechnical and Geological Engineering, 40, 4721–4734. https://doi.org/10.1007/s10706-022-02180-z Jas, K., & Dodagoudar, G. R. (2023). Liquefaction potential assessment of soils using machine learning techniques: A state-of-the-art review from 1994–2021. Vol. 23. Li, C., Cui, Y., Liu, Z., Yuan, X., Xu, L., & Wei, Q. (2025). Comparison of CPT liquefaction discrimination methods and analysis of liquefaction influencing factors based on machine learning. https://doi.org/10.13197/j.eeed.2025.0309 Cheng, K., & Ziotopoulou, K. Machine learning applications in geotechnical earthquake engineering: Progress, gaps, and opportunities. Review of AI liquefaction models in practice (75 publications). https://doi.org/10.1177/87552930231173711