Estimation of Compressive Strength of Glass Fiber Reinforced Expansive Soil in Alkaline Activated Binder by an Artificial Neural Network Based Model

Expansive soil possesses inferior geomechanical behavior with large heaving cracking propagation due to seasonal variation, resulting in early strength reduction. This paper adopted the artificial intelligence techniques for predicting the compressive shear strength of geopolymer stabilized expansive soil reinforced with glass fiber (GF). The effectiveness of GF composite material mixed with an alkaline activated binder (AAB) against conventional lime mixed in soft soil was also evaluated in the study. AAB was produced by combining aluminosilicate precursors (Class-F fly ash and slag) in an alkali solution comprising sodium silicate and sodium hydroxide with 0.4 water to solid ratio (w/s). An artificial neural network (ANN) model was proposed based on unconfined compressive shear strength (UCS) test results as a performance indicator. ANN study revealed substantial correlations (R² = 0.90–0.95) between the dosage of fiber, slag, fly ash, fiber length, and UCS of geopolymerized soil. Moreover, microstructural and morphological studies were carried out for unreinforced geopolymerized and GF-soils. It was observed that GF-AAB-soil achieved a higher frictional bonding with strong interfacial density and low linear shrinkage and tensile cracking compared to GF-lime stabilized soils. The correlation equations derived from ANN analysis were found to be satisfactory with the test findings. It was suggested that the proposed correlations might be ideal for a preliminary design of a project with a financial and schedule constraints.