ABSTRACT
Knowing a soil’s strength and deformation characteristics is crucial for load-bearing applications, and they are often determined through costly and onerous laboratory procedures. However, these characteristics can vary based on a multitude of factors, meaning new tests are needed each time the material properties or loading environment changes. Empirical correlations can be used to remedy this, though extrapolation inaccuracies may arise due to their simplistic modelling frameworks. This paper presents a short critical review of selected studies using machine learning (ML) techniques to predict the shear strength characteristics and resilient modulus of granular materials. Given the popularity of ML, these studies are scrutinized from a geotechnical perspective, rather than focusing on specific modelling intricacies. From this review, it is evident that there is a clear divide between studies that use ML as a tool to either fit a given set of data, or to enhance well-established geotechnical relationships and concepts.
