Unconfined Compressive Strength Modelling of Soil Stabilized with Millet Husk Ash, Quarry Dust and Bush Gravel

Abstract

Clay soils rich in aluminosilicate minerals are often characterized by
low shear strength, high plasticity, and inadequate load-bearing
capacity, rendering them unsuitable for pavement and structural
applications in their natural state. This research explores the
stabilization of weak clay soil through the combined use of millet
husk ash (MHA), quarry dust (QD), and bush gravel (BG), employing
laboratory geotechnical testing, SEM EDX microstructural
characterization, and predictive modelling of unconfined
compressive strength (UCS). Experimental results indicate that an
optimal mixture comprising 5% MHA, 15% QD, and 25% BG
increased the UCS from 289 kPa in untreated soil to 1459 kPa,
representing a 405% improvement. Correspondingly, the California
Bearing Ratio (CBR) rose from 4.2% to 13.2%, and maximum dry
density (MDD) improved from 1.62 g/cm³ to 1.74 g/cm³. SEM EDX
analysis revealed the development of calcium silicate hydrate (C
S H) gels and enhanced particle interlock as the primary mechanisms
responsible for strength enhancement. Predictive modelling using
Linear Regression, Gaussian Process Regression (GPR), Support
Vector Regression (SVR), and Bayesian Ridge Regression
demonstrated limited generalization due to the small dataset;
however, GPR achieved the lowest prediction errors (MAE = 209.24
kPa, RMSE = 295.76 kPa) while allowing uncertainty quantification.
The study highlights the synergistic effect of chemical and physical
stabilization, emphasizes the critical importance of optimized
additive ratios, and demonstrates the potential of predictive models
for preliminary design in sustainable soil stabilization.