Soil Properties: Physics Inspired, Data Driven

Research and engineering projects during the last century have advanced the understanding of soil behavior and contributed extensive datasets. Nevertheless, the granular nature of soils challenges the accurate prediction of soil properties. In this context, a physics-inspired and data-driven approach helps us anticipate the soil response. The granular nature of soils defines their inherent properties (e.g., non-linear, non-elastic, porous, pervious) and their effective stress-dependent stiffness, frictional strength and dilation upon shear. The revised soil classification builds on the physical understanding of soils (e.g., packing characteristics and the effect of pore fluid chemistry on fines) and the extensive data accumulated in the field. Asymptotically correct compression models adequately fit experimental data and avoid numerical difficulties. Constant volume friction reflects particle shape and it is strongly dependent on stress path. Repetitive loading leads to characteristic asymptotic conditions (terminal density, and either ratcheting or shakedown). Data and physical analyses suggest a power relationship between void ratio and hydraulic conductivity. The pore-scale origin of suction is interfacial tension and contact angle. P-wave velocity is a good indicator of loss of saturation and S-wave velocity measures the skeletal shear stiffness. Permittivity, electrical conductivity and thermal conductivity are sensitive to water content. Finally, ubiquitous sensors, information technology and cellular communication support the development of effective laboratory characterization techniques and allow us to access large databases. These are transformative changes in geotechnical engineering.