Stabilization of Expansive Clays with Basalt Fibers and Prediction of Strength by Machine Learning

Expansive clays with high plasticity need to be stabilized to prevent hazards that may arise due to the extreme volume changes experienced with moisture fluctuations. Utilizing a kind of natural and eco-friendly sustainable fiber named as basalt fiber into the soils has become a new issue that needs to be expanded in scope. In this paper, a high plastic soil was stabilized by these natural basalt fibers to reduce possible soil-induced disasters. Basalt fibers in different lengths were mixed into the clay at varied amounts. Due to the soil's sensitivity to water, the samples were prepared at distinct water contents, 2 on the dry and 3 on the wet side of the optimum. To question whether the strength loss due to the moisture change can be regained with basalt fibers or not, the strength tests were performed on both natural and stabilized samples. Through the tests, it was revealed that the strength of the expansive clays can be enhanced up to 280% at a fiber content of 2%. The highest strength was obtained at approximately 880 kPa by mixing 24 mm fibers with 15% water at 1 and 2% ratios. As anticipated, the long fibers (24 mm) supplied a real reinforcement even at high water contents. In addition, the obtained data set was used to train machine learning algorithms (linear, ridge, lasso, support vector, decision tree) that have just started to be applied in geotechnical engineering. Results have proved that, the decision tree regression outperformed the stress and strain with 0.85 R-squared (R2) in stress and 0.91 R2 in strain estimation. Additionally, it was revealed from the feature importance analyses that water content has an importance of approximately 85% on stress and up to 97% on strain.