State-of-the-art review on predicting stabilized soil properties using artificial neural networks

Purpose This study aims to assist researchers in achieving accurate predictions of stabilized soil properties using artificial neural networks (ANNs). It also highlights and documents the limitations of existing studies, while presenting suggestions and potential directions for future advancements. Design/methodology/approach This study explores the extensive application of ANNs in predicting soil characteristics, with a particular focus on stabilized soils. A structured framework (PRISMA) was used for the systematic selection and refinement of the literature. Over a hundred published articles related to ANNs in the field of soil stabilization were reviewed, and the key insights were identified. The processes involved in the ANN analysis, such as the collection of data required, the network architecture, the performance of the model and the sensitivity analysis, will be briefly discussed. This approach is expected to serve as a valuable reference for researchers pursuing work in this area. Findings The findings from the study indicate that ANN is a reliable tool for accurately predicting the mechanical properties of soil. One notable shortcoming in the reviewed studies is the lack of specification regarding clay mineralogy. Since the properties of the soil depend on clay mineralogy, the developed ANN models without the specifications of clay mineralogy and soil type cannot be regarded as impartial predictors. The stabilized soil properties like permeability coefficient, consolidation characteristics, resilient modulus and free swelling potential found limited in research and require further exploration. Future research should also address the influence of climatic variations on stabilized soil behavior and explore the integration of other machine learning approaches to enhance predictive capabilities. Originality/value This study can serve as a guide for researchers interested in predicting the properties of stabilized soils using ANNs. It also identifies existing research gaps and provides possible suggestions for addressing them.