Hybrid neuro-fuzzy models for assessing the optimum moisture content of lime cement-treated soil

This study explores the application of machine learning (ML) techniques to predict the optimum moisture content (OMC) of soil-stabilizer combinations. OMC represents the moisture level where soil achieves peak compaction and strength in conjunction with a stabilizer, playing a vital role in attaining desired engineering properties in soil stabilization endeavors. Employing the adaptive neuro-fuzzy inference system (ANFIS) as a robust ML tool, this research endeavors to formulate intricate and accurate models. These models forge connections between OMC and many intrinsic soil properties, including particle-size linear shrinkage, plasticity, distribution, and the nature and quantity of stabilizing additives. A diverse dataset is curated to ascertain the responsiveness of OMC to variations in influential factors, encompassing distinct soil types and previously documented results from stabilization tests. In an endeavor to enhance model precision, this study integrates two meta-heuristic algorithms: the Cheetah optimization algorithm (CO) and the equilibrium slime mould algorithm (ESM). By synergistically leveraging these algorithms, the accuracy of the models is fortified. Rigorous validation ensues through an analysis of \(OMC\) $\mathit{OMC}$ samples drawn from diverse soil types obtained from historical stabilization test outcomes. The study unveils three notable models: ANCO (ANFIS + CO), ANES (ANFIS + ESM), and an independent ANFIS model. Each of these models furnishes invaluable insights that substantiate the meticulous projection of OMC for soil-stabilizer blends. Noteworthy among them is the ANCO model, exhibiting exceptional performance metrics. The R² (correlation coefficient) value of 0.996 and an impressively low RMSE of 0.436 indicate its precision and reliability. These findings not only underscore the accuracy of the ANCO model but also underscore its efficacy in prognosticating soil stabilization outcomes. This methodology introduces a promising avenue for accurately predicting \(OMC\) $\mathit{OMC}$ across a spectrum of engineering applications connected to soil-stabilizer amalgamations.