Machine learning approaches for resilient modulus modeling of cement-stabilized magnetite and hematite iron ore tailings

Resilient modulus (Mr) is key for understanding the stiffness and stress‒strain properties of road materials for flexible pavement design. Measuring Mr in a laboratory requires conducting dynamic triaxial loading tests with varying confining and deviatoric stresses, which can be costly and time-consuming process. This study evaluates various machine learning (ML) models to predict the Mr of cement-stabilized magnetite and hematite iron ore tailings based on multiple variables such as cement content, curing time, bulk stress, and deviatoric stress, which are considered as input parameters. For developing ML models, a set of data from experimental studies was collected. After comparison, Gaussian Process Regression outperformed other methods in predicting Mr of both MIOT and HIOT. For the MIOT and HIOT materials, the R² values were 0.9936 and 0.9876, and 0.9893 and 0.9825 for the train and test datasets, respectively. The sensitivity analysis revealed that the curing time was the least important input variable, whereas the Portland cement percentage was the most effective for predicting the Mr of cement-stabilized iron ore tailings. Additionally, a parametric study was undertaken to investigate the impact of each input variable on Mr.