A new approach for estimation of rock brittleness based on non-destructive tests

This paper aims to propose predictive equations for estimation of rock brittleness as a function of intact rock properties including rock density (r), Schmidt hammer (Rn) and wave velocity (Vp) using two optimization techniques, artificial neural network (ANN) and FA-ANN (Firefly Algorithm and ANN). Using ANN and FA-ANN techniques, 10 different models were developed and compared to find the optimum one implementing some performance indices such as coefficient of determination (R2) and root mean square error (RMSE). In addition, a ranking system was performed to select the best models. It was found that in developing ANN models, the Model number 1 is superior to other 4 models (models 2-5). Likewise, in developing hybrid FA-ANN technique, model number 9 was better than other 4 models (models 6-10). Further, the best models obtained with these two intelligent techniques were compared to show that hybrid model is better than a simple ANN model. It was found that R2, RMSE, and total ranking are obtained as 0.826, 0.1481, and 19 for ANN while those are 0.896, 0.0812 and 36 for FA-ANN, respectively. It was also concluded that the model 9 of FA-ANN technique indicates the best performance among all developed hybrid models.