| Abstrakt: |
The present study focuses on the estimation of uniaxial compressive strength (UCS) in geotechnical projects. UCS is a crucial parameter in designing such projects, and traditional methods of measurement outlined by standards are often time-consuming and cumbersome. Additionally, obtaining standard core samples is always a challenging task in fractured or discontinuous rock masses. To overcome these limitations, alternative techniques such as simple regression (SR), multivariable regression (MVR), and artificial intelligence (AI) have been utilized to estimate UCS based on easily derived geotechnical parameters. While SR and MVR provide simple equations that can be used without the need for complex calculations, AI techniques offer the potential for higher prediction accuracy. However, AI models often require more complex solutions due to their intricate algorithms and computational processes. This makes AI less practical for ongoing geotechnical projects in which simple and quick estimations are often preferred. In this study, both the SR and the MVR models were employed to estimate the UCS using the Schmidt hammer rebound number (RN), density (ρ), porosity (ϕ), and point load index (Is50) as input parameters. Various SR and MVR models were tested, and the best-performing models were selected based on performance indices such as the normalized root mean square error (NRMSE), relative root mean square error (RRMSE), variance accounted for (VAF), efficiency (E), and correlation coefficient (R2). Furthermore, this study proposes a new performance index (PImod) to evaluate the predictive capabilities of the models. This index likely incorporates additional criteria or metrics beyond the aforementioned traditional performance indices. Overall, the equations derived from the regression models in this study offer a simpler and more practical approach for geotechnical practitioners to quickly assess UCS in any given area, providing a useful tool for their work. [ABSTRACT FROM AUTHOR] |
