Because the experimental trials in civil engineering field are difficult and time-consuming, the application of artificial intelligence (AI) techniques is attracting considerable attention, with their use enabling successful results to be more easily obtained. In this study, we investigated the effect of fiber size, fiber amount, water content, and cell pressure on maximum deviator stress (MDS) and failure deformation (FD) of basalt fiber (BF)-reinforced, unsaturated silty soils using three AI techniques: the artificial neural network (ANN), support vector machine (SVM), and fuzzy logic (FL). The numerical analyses and experiments were conducted using varying amounts (1, 1.5, and 2%) and lengths (6, 12, and 24 mm) of BF, and a total of 180 samples were prepared for the detailed investigation. In order to compare model performances, R2 and MAPE goodness-of-fit metrics were used. The experimental results revealed that the addition of BF generally increased the MDS of the soils, which corresponds to the shearing resistance. According to AI models result, FL outperformed the SVM and ANN, with a R2 value of 0.938, especially in FD prediction. The sensitivity analysis was performed to ascertain the effect of the inputs on the MDS and FD response variables. Results revealed that fiber length and cell pressure have substantial influence in MDS estimations.
Read full abstract