Application of differential evolution algorithm and comparing its performance with literature to predict rock brittleness for excavatability
Saffet Yağız, Aitolkyn Yazitova, Halil Karahan
Abstract
The aim of this study was to estimate brittleness of intact rock by applying differential evolution (DE) algorithm and then to compare the results obtained from the optimum model with literature. For this aim, several models including linear and nonlinear were developed for predicting the brittleness via DE algorithm using the dataset obtained from 48 tunnel cases around the world. Each model were developed using 80% of the dataset as training and 20% of the dataset as testing in random. After that, developed models are compared according to the coefficient of correlations (r2), computer process unit (CPU), mean-squared error (MSE) and number of function evaluation (NFE) values to choose the best accurate one among them. It is found that the values r2, MSE, NFE and CPU ranged between 0.9385–0.9501, 8.2616–9.938, 7217–11,176 and 4.91–36.22, respectively, with the quadratic model (QM) indicating the best performance. It is concluded that the DE algorithm is itself very powerful tool for estimating the brittleness; however, the QM is superior especially for simulations in which computational time and optimisation is a critical.