IoT-Enabled Data-Driven ML for Predictive Modeling in Unsaturated Soil Mechanics

The integration of advanced machine learning (ML) algorithms with unsaturated soil mechanics provides a transformative methodology for accurately forecasting soil behavior across a broad spectrum of environmental conditions. This research specifically addresses the complex interrelations among electrical resistivity, volumetric moisture content, wet–dry cycling, and pore water pressure, each of which crucially determines soil stability. An IoT-based data acquisition system employing the Wenner array configuration is used to continuously gather large-scale, high-resolution measurements on soils of varying bulk density and water retention properties. Such comprehensive datasets enable the development of sophisticated ML models—such as neural networks and ensemble methods—that efficiently capture intricate, nonlinear coupling among hydraulic conductivity, matric suction, and microstructural changes during cyclic wetting and drying. By systematically incorporating these parameters, the models deliver robust predictions of soil-water characteristic curves (SWCC), shear strength variability, and infiltration dynamics. The integration of sensor-driven, real-time data with ML further empowers geotechnical engineers to refine slope stability analyses and foundation design, thus minimizing risk under transient loading and environmental fluctuations. Ultimately, these innovative approaches herald a new era in geotechnical engineering, providing data-driven insights that enhance the reliability of infrastructure design, operation, and maintenance, even under the most dynamic site conditions.