Predicting the unconfined compressive strength of stabilized soil using random forest coupled with meta-heuristic algorithms

Unconfined Compressive Strength (UCS) is a crucial mechanical parameter of rocks, which is pivotal in developing accurate geomechanical models. Traditionally, UCS estimation involves expensive and time-consuming methods, such as lab testing of retrieved core samples or well-log data analysis. This research presents a novel approach for real-time estimation of UCS, crucial in various geomechanical applications. It employs Random Forest (RF) prediction models enhanced by Runge Kutta Optimization (RKO) and Beluga Whale Optimization (BWO) algorithms for improved accuracy and efficiency. Validation using UCS samples from diverse soil types yields three distinct models: RFRK (RF + RKO), RFBW (RF + BWO), and an individual RF model, each contributing valuable insights. The RFBW model particularly stands out with high R² values (0.994) and a favorable RMSE (73.93), indicating superior predictive and generalization capabilities. This method represents a significant advancement in UCS prediction, offering efficiency and time-saving benefits across geomechanical fields.