Estimation of Compressive Strength of Basalt Fiber-Reinforced Kaolin Clay Mixture Using Extreme Learning Machine

Abstract

Background: In this study, the unconfined compressive strength (qu) of a mixture consisting of clay reinforced with 24 mm-long basalt fiber was estimated using extreme learning machine (ELM). The aim of this study is to estimate the results closest to the data obtained through experimental studies without the need for experimental studies. The literature review reveals that the ELM technique has not been applied to predict the compressive strength of basalt fiber-reinforced clay, and this study aims to provide a novel contribution in this area. Methods: The experimental studies included data derived from a series of mixtures where water contents of 20%, 25%, 30%, and 35% were combined with kaolin clay reinforced with 24 mm-long basalt fiber at reinforcement rates of 0%, 1%, 2%, and 3%. Based on the experimental results obtained for these mixtures, an ELM model was developed to predict the qu. Results: ELM, recognized for its computational efficiency and high predictive accuracy, demonstrated exceptional performance in this application, achieving an R value of 0.9976 and an RMSE of 0.0001. Furthermore, this study includes a figure representation illustrating that the ELM-based predictions align closely with the experimental results, underscoring its reliability. Conclusions: To further validate its performance, ELM was compared with other artificial intelligence models through a 5-fold cross-validation approach. The analysis revealed that ELM outperformed its counterparts, achieving a remarkable RMSE value of 0.000174, thereby solidifying its capability to accurately estimate the compressive strength of the soil under varying reinforcement and water content conditions. Thus, it is aimed to save labor, material, and time.