Predicting Swelling Pressure in Expansive Soils Using Nonlinear Regression Models of Index Properties

Due to the presence of considerable amount of Montmorillonite clay mineral, expansive soils have a peculiar nature of expanding and shrinking characteristics when exposed to moisture variation. As a result of swelling, swelling pressure is exerted by the soil on the structures built on it and causes an increase in volume of the soil mass which also lifts the structure. The magnitude of swelling pressure is influenced by environmental conditions and soil index properties, making accurate prediction essential for safe design. Previous studies have mostly relied on linear regression models and often ignored the normality of laboratory data. This study investigates the correlation between index properties and swelling pressure (Pa) of expansive soils in the Woldia area using one-, two-, and three-parameter nonlinear regression predictive models. Twenty-six disturbed and undisturbed soil samples were collected from 16 test pits at depths of 1.5 m and 3.0 m. Index property tests including Atterberg limits, free swell, dry density, specific gravity, and natural moisture content and swell-consolidation tests were conducted in accordance with ASTM standards. According to Unified soil classification System (USCS), the soil of the study area is classified as High plastic clay (CH) with a potential of expansion. American Association of State Highway and Transportation Officials Classification System (AASHTO) also show that the soil is Plastic Clay (A-7-5) with high volume change capacity and Swelling pressure ranged from 160 to 445 kPa. The developed nonlinear regression models indicated that swelling pressure is highly sensitive to natural moisture content, dry density and liquidity index. The three-parameter nonlinear predictive model provided the highest predictive accuracy, outperforming linear models. The proposed model, with a coefficient of determination (R²) of 0.9583, mean square error (MSE) of 0.00042, and square root of mean square error (SRMSE) of 0.02058, provides a reliable and cost-effective method for estimating swelling pressure from easily measurable soil properties, thereby reducing the need for extensive laboratory testing and supporting safer geotechnical design in expansive soil areas.