Prediction of Unconfined Compressive Strength of Stabilized Sand Using Machine Learning Methods: Fid-Bau Portal

Prediction of Unconfined Compressive Strength of Stabilized Sand Using Machine Learning Methods

Zhao, Qinggang / Shi, Yan

This article introduces a novel method using the K-nearest neighbors (KNN) model to predict the unconfined compressive strength (UCS) of soil-stabilizer combinations. To ensure accurate UCS estimation, customized KNN prediction models are created. The research combines two meta-heuristic algorithms, the giant trevally optimizer (GTO) and the tunicate swarm algorithm (TSA), for improved accuracy. These algorithms examine UCS samples made from various soil types and validate the models using the results of previous stabilization tests. The study’s findings point to three distinct models: KNGT, KNTS, and an independent KNN model. Each of these models offers insightful information that helps with the precise prediction of UCS. As a result, the KNGT model performs best by displaying suitable statistics like a high R² value of 0.995 and a low RMSE value of 84.5 (kN/m²). In addition, KNTS obtained the second suitable performance with R² and RMSE equal to 0.989 and 121.2 (kN/m²), alternatively. These results highlight the accuracy, dependability, and prognostication capabilities of the KNGT model. This study describes a unique machine-learning approach for the exact prediction of critical soil parameters, with an emphasis on UCS in civil engineering. In order to overcome data difficulties, it leverages the KNN model and presents a novel dual-algorithm technique that combines the GTO and TSA. This integration improves KNN’s accuracy and efficiency while also easing UCS structure planning in civil engineering. The work not only improves UCS prediction accuracy but also opens up a viable route for future research and practical applications in civil engineering undertakings, giving significant insights and a fresh method for UCS performance prediction.