A Model Adapted to Predict Blast Vibration Velocity at Complex Sites: An Artificial Neural Network Improved by the Grasshopper Optimization Algorithm

Many factors complicate the blasting vibration velocity at complex sites because of their nonlinear relationships. Traditional empirical formulas often yield unsatisfactory prediction results. To improve the prediction accuracy of the peak particle velocity (PPV), this paper combines the ability of an artificial neural network (ANN) to solve complex nonlinear function approximations and the global optimization ability of 10 metaheuristic optimization algorithms and establishes an improved ANN prediction model. On the basis of the blasting vibration data monitored during blasting excavation of the left abutment groove of the Baihetan hydropower station, the maximum charge per delay, distance from the blast face, height difference, and acoustic wave velocity were selected as the input parameters. Through a comprehensive evaluation of the running time results, the root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>), a new algorithm, the grasshopper optimization algorithm (GOA), which is suitable for optimizing an ANN to predict PPV, is obtained. In comparison, the GOA-ANN model has good generalizability, with an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.978, an RMSE of 0.240, and an MAE of 0.198. When the main factors affecting blasting vibration at complex sites change, the prediction results of the GOA-ANN model better match the actual monitoring values. This research provides a reference for accurate PPV prediction at complex sites.