Machine Learning-Enhanced Soil Stabilization Using Multiwalled Carbon Nanotubes and Fly Ash for Road Construction
摘要
This study investigates the effects of multi-walled carbon nanotubes (MWCNTs) on the engineering properties of soil-fly ash mixes, with a focus on enhancing California Bearing Ratio (CBR) for potential use as subgrade material in pavement construction. Soil-fly ash mixes were treated with various concentrations of MWCNT, SHMP (Sodium Hexametaphosphate), and cement. An optimal model and an alternative model were developed using CART regression analysis, with R2 values of 0.95 and 0.84 for training and testing of the optimal model, respectively. A 13-node CART model was selected over a 26-node model to balance predictive accuracy and interpretability. The maximum CBR value observed was 50.78% for a mix of 0.01% MWCNT, 2% SHMP, and 3% cement (20% soil replacement with fly ash), compared to a minimum CBR of 4.39% for untreated natural soil. The findings suggest that adding MWCNT and cement to soil-fly ash mixes significantly enhances CBR, supporting the use of these stabilized materials for developing resilient subgrade layers.