Özet:

Uniaxial compressive strength (UCS) is one of the most critical design parameters of engineering projects, which is directly used parameter in many projects and classification systems. Considering the difficulties in obtaining the UCS, it is remarkable that it is estimated using machine learning-based approaches. In this study, two different models were constructed using laboratory results of the 137 core samples. Apparent porosity (n), p wave velocity (Vp), and unit weight (n) values are the input parameters in these models; the UCS was tried to estimated by machine learning-based methods. For this purpose, three different machine learning methods, such as Gaussian Process Regression (GSR), Support Vector Machine (SVM), and Ensembles of Tree (ET) were employed. R2 (Coefficient of Determination), RMSE (Root Mean Square Error), MSE (Mean Square Error), and MAE (Mean Absolute Error) performance indices were used to evaluate the performances of three different machine learning methods for five different data sets of two different models. According to these assessments, it was determined that all three different machine learning methods were successful for estimating UCS in general; however, the ET method generally had higher prediction performance.

Anahtar Kelimeler: