Predictive model for California Bearing Ratio (CBR) in expansive coastal subgrades: a rapid geotechnical assessment for degraded and marginal lowland areas

Abstrak

In geotechnical engineering, professional actions and expert judgment are often essential in soil investigation methods. In lowland and coastal areas, expansive, fine-grained soils and sandy sedimentation lead to reduced bearing capacity, posing significant challenges for sustainable infrastructure development on marginal and degraded lands. Such conditions are prevalent in post-mining or naturally poor coastal environments, necessitating cost-effective and rapid assessment tools. This study modeled a clayey sand mixture using bentonite and kaolin as fine fractions, which exhibit expansive behavior and poor gradation, simulating worst-case geotechnically degraded subgrades. The mechanical behavior of the soil was evaluated through modified compaction, using the CBR test and CPT test as bearing capacity parameters. Soil mixtures were simulated with sand fractions?65% and bentonite-kaolin compositions with?50% bentonite. Compaction was modeled using variations in energy compaction and water content under conditions below the maximum dry density. CBR prediction was conducted using Qc as the primary predictor and dry density as a supporting predictor. A hybrid stepwise regression analysis in the Z-score scale identified positively correlated predictors: +3.00 (Qc), +0.55 (?dry), and +1.28 (Qc?dry interaction). The regression model showed strong statistical performance with R² = 0.84 and high significance with the lowest p-values. The resulting regression equation offers an applicable approach to rapidly evaluate the bearing capacity of subgrade soils in degraded coastal or marginal conditions, thereby facilitating geotechnical engineering design and initial site assessment crucial for land management and rehabilitation actions.

Kata kunci

• bentonite-kaolin mixture
• CBR prediction
• degraded land management
• empirical correlation
• lowland area
• stepwise regression