Prediction of the blast-induced ground vibration in tunnel blasting using ANN, moth-flame optimized ANN, and gene expression programming

摘要

The blast-induced ground vibration (BIGV) is a severe environmental impact of blasting as it can affect the integrity of the structures and cause civil unrest. In this study, the BIGV of Daejeon tunnel was predicted taking into consideration parameters such as hole length, the charge per delay, number of holes, total charge, distance from the measuring station to the blasting point and the rock mass rating as the input parameters, while the peak particle velocity (PPV) was the targeted output parameter. An artificial neural network (ANN) model was first simulated. The optimum ANN structure obtained was optimized using a novel moth-flame optimization algorithm (MFO). The gene expression program (GEP) was also used to develop another new model. The proposed models were compared with the multilinear regression (MLR) model and the selected empirical models for the PPV predictions. The performance of the proposed model was evaluated using statistical indices such as adjusted coefficient of determination (adj R2), mean square error (MSE), mean absolute error (MAE), and the variance accounted for (VAF). The proposed MFO-ANN outperformed other models with the adj R2 of 0.9702 and 0.9577, VAF of 97.0472 and 95.9832, MSE of 0.0009 and 0.0008, and MAE of 0.0233 and 0.0216 for the respective training and testing phases. The sensitivity analysis was conducted using the weight partitioning method (WPM), and the charge per delay has the highest influence on the predicted PPV. This study indicates the suitability of the proposed models for the prediction of PPV.