Abstract
Expansive soils rich in montmorillonite minerals undergo significant volume changes with moisture fluctuations, generating swelling pressures that can adversely affect civil engineering structures. This study developed nonlinear regression models to predict the swelling pressure of expansive soils in Woldia, Ethiopia, using readily measurable soil index properties. Twenty‐six disturbed and undisturbed soil samples were collected from 16 test pits (TPs) at depths of 1.5 and 3.0 m, and laboratory tests were conducted in accordance with American Society for Testing and Materials (ASTM) standards. The study’s novelty lies in the development and evaluation of one‐, two‐, and three‐parameter nonlinear regression models using systematic data transformation and nonlinear subset‐selection procedures in NCSS‐2025 software. More than 214 parameter transformations were assessed to establish optimal predictive relationships between swelling pressure and soil properties. The measured swelling pressure ranged from 160 to 445 kPa and was primarily influenced by natural moisture content (NMC; 32.47%–42.85%), dry density (1.134–1.382 g/cm
3
), and liquidity index (−0.043 to 0.122). Among the 12 developed models, the three‐parameter nonlinear model incorporating these variables provided the best predictive performance, with
R
2
= 0.9583, mean square error (MSE) = 0.00042, and SRMSE = 0.02058. Independent validation confirmed its reliability, yielding a prediction error of only 3.2%. The proposed model offers a rapid, reliable, and cost‐effective alternative for estimating swelling pressure from routine laboratory tests, thereby reducing the need for extensive swell‐consolidation testing and supporting geotechnical site investigation, foundation design, pavement engineering, and expansive‐soil risk assessment in regions with similar geological and climatic conditions.