Prediction of blast-induced ground vibration using empirical models and artificial neural network (Bakhtiari Dam access tunnel, as a case study)

Abstract

Blasting operation is among the most common methods of rock excavation in the civil engineering and mining operations. Ground vibration is the most unfavorable effect of blasting operation such that failure to accurately control this problem causes damage to adjacent structures. In this regard, geotechnical engineers face the challenge of accurately predicting blast-induced ground vibrations. Geographical location of Bakhtiari Dam (located in the southwest of Iran) is needed to construct an access road to its nearest city through the rough topography. To establish the access road in the plan, blasting operation methods have been used. In this study, blast-induced ground vibrations in the study area are evaluated using five common functional forms of the empirical model and their corrected regression coefficient for the area. Then, the ground vibrations generated in the study area were predicted by designing an artificial neural network model. For this purpose, the maximum charge per delay, the distance between the blast point and monitoring stations, and the ground vibration values were surveyed for 80 blast events, and their necessary parameters were determined. A total of 64 datasets were used to obtain the coefficients of the empirical models and to create the artificial neural network model. In addition, 16 datasets were used to estimate the performance and accuracy of each model. To measure the accuracy of the constructed models, some statistical parameters were also used. The results show that in the study area, the artificial neural network model presents the most accurate and appropriate model for predicting blast-induced ground vibrations. The neural network proposed in this research is suggested for areas with geological features resembling those of the present study.