A hybrid artificial bee colony algorithm-artificial neural network for forecasting the blast-produced ground vibration

• Published in: Engineering with computers Vol. 33; no. 3; pp. 689 - 700
• Main Authors: Taheri, Khalil, Hasanipanah, Mahdi, Golzar, Saeid Bagheri, Majid, Muhd Zaimi Abd
• Format: Journal Article
• Language: English
• Published: London Springer London 01.07.2017; Springer Nature B.V
• Subjects: Artificial neural networks; Blasting; CAE) and Design; Calculus of Variations and Optimal Control; Optimization; Classical Mechanics; Computer Science; Computer-Aided Engineering (CAD; Control; Ground motion; Math. Applications in Chemistry; Mathematical and Computational Engineering; Mathematical models; Neural networks; Open pit mining; Optimization; Original Article; Prediction models; Search algorithms; Swarm intelligence; Systems Theory; Vibration measurement; Vibration monitoring; Weight; ANN; ABC-ANN; Multiple regression; Ground vibration
• ISSN: 0177-0667; 1435-5663
• DOI: 10.1007/s00366-016-0497-3

Drilling and blasting is an inseparable part of the rock fragmentation process in open-pit mines. Prediction of blast-produced ground vibration is considered as an important issue in blasting works. The aim of this study is to propose a hybrid model for predicting blast-produced ground vibration in the Miduk copper mine, Iran, using combination of the artificial neural network (ANN) combined with artificial bee colony (ABC) (codename ABC-ANN). Here, ABC was used as an optimization algorithm to adjust weights and biases of the ANN. The predicted values of ground vibration by ANN and ABC-ANN models were also compared with several empirical models. In this regard, 89 blasting events were monitored and values of two influential factors on ground vibration, i.e., maximum charge weight used per delay (MC) and distance between monitoring station and blasting-point (DI) together with their peak particle velocity values (as an index of ground vibration) were carefully measured. The results of the predictive models have been compared with the data at hand using mean absolute percentage error, root mean squared error and coefficient of correlation ( R 2 ) criteria. Eventually, it was indicated that the constructed ABC-ANN model outperforms the other models in terms of the prediction accuracy and the generalization capability.