Geophysical, geotechnical and mineralogical characterization of subgrades: implication for pavement stability and machine learning prediction: a case study
摘要
The foundation soils (subgrades) of highways play critical roles in the stability of the overlying pavement. Studies indicate that geological considerations are often neglected during highway design and construction. This study therefore probed into the cause of persistent failure of a major highway in North-central Nigeria from a geological standpoint with particular focus on the subgrades. The study aimed at characterizing the subgrade soils underneath the pavement and determining optimum conditions for improving it if required and to establish empirical relationships between evaluated properties. Petrographic study was conducted on the parent rocks underlying the pavement to determine the major rock forming minerals in the rocks, while geophysical surveys involving multichannel analysis of surface waves (MASW) was carried out to determine the subgrades’ stiffness. The geotechnical tests involved grain size distribution, Atterberg limits, compaction, California bearing ratio (CBR) and unconfined compression test at varying lime percent addition and curing time. The clay and non-clay minerals in the subgrades were determined through XRD test. Modal composition of rocks revealed that high amounts of feldspar (40–50%), biotite (30–35%) and low quartz (10–15%) dominate the mica-schist that underlies the road’s failed section causing the preponderance of clays in subgrade soil. This is in contrast with the moderate feldspar (25–40%), low mica (10–15%) and high quartz (45–55%) that dominate the granite-gneiss underlying the stable section. The foundation soils of the unstable and stable segments are respectively characterized by low (183–258 m/s) and high shear wave velocity (393–440 m/s) which correspond to loose and stiff subgrade soils. The XRD results showed the presence of clays that have swelling potentials upon contact with water which are notably absent in the subgrades of the stable section. Subgrade soils consistency in terms of Unconfined Compressive Strength (UCS) of the failed section improved significantly from soft (< 50 kN/m2) to stiff (> 100 kN/m2) upon 5% lime and 7 days of curing time. Metrics for model evaluation revealed good models with high R2-Score (0.77–0.8) and low Mean Absolute Error (MAE) of 1.34–1.84. A good coefficient of correlation (r) between modelled CBR and laboratory CBR (r = 0.71–0.89) was also obtained. The study demonstrated that geological studies are paramount not just to provide insights into highway failures but to aid in optimizing the design and construction of healthy and long-lasting pavements as well as preventing continuous future occurrences of failure.