Author ORCID Identifier

Year of Publication


Degree Name

Master of Science in Mining Engineering (MSMIE)

Document Type

Master's Thesis




Mining Engineering

First Advisor

Dr. Jhon Silva

Second Advisor

Dr. Steven Schafrik


Ground vibrations are a critical factor in the rock blasting process. The instantaneous load application exerted by the gas pressure during the detonation process acts on the blasthole walls creating dynamic stresses in the adjacent rock. This triggers different sorts of stress waves, mainly divided into two categories: body and surface waves. The first comprises the P and the S waves, while the second comprises Rayleigh waves. These waves spread concentrically starting at the blast location and move along the ground surface and its interior, being attenuated as they reach further distances.

In most cases, and accepting the hypothesis that the attenuation of the vibrational waves is proportional to the distance and inverse to the energy released during the blast, the vibration from a large blast can be perceived from far away. In any case, the ground vibrations can affect pit slopes’ stability, and they can also damage man-made structures. Therefore, ground vibrations need to be predicted, monitored, and controlled to minimize the vibration-caused disturbance to nearby or far elements.

The assessment of vibrations produced by blasting has traditionally relied on maximum charge weight per delay scaling laws. These two-parameter or three-parameter models depend on a curve fit to measured data. In this approach (scaled laws), the ground vibration waveforms are not used in the vibration level estimation, neither are other blast design parameters, such as burden, spacing, hole diameter, explosive density, uniaxial compressive strength of the rock, Young’s modulus, subdrilling, stemming, and charge length, to name a few. To provide a more comprehensive approach to ground vibration modeling, including the aforementioned variables, artificial neural networks (ANN) have been employed in several studies worldwide with promising results.

The present thesis uses ANN applied to ground vibration modeling, considering the blasting parameters in the input, unlike the empirical approaches, using data from an open-pit gold mine in La Libertad region, Peru. The results from this study are then compared against the traditional scaled distance approach. Two datasets were used, the first was comprised of 178 shots and the second, 80 shots. The first dataset was collected at the La Arena community, and the second was collected at the La Ramada community. Both of these communities are the most populated in the direct area of influence of the mine. When comparing the measured and predicted PPV values using the scale-distance method in the La Arena community, the coefficient of determination () found was 0.1166, while the found when comparing the measured and predicted PPV values using the optimum trained artificial network was 0.5915. Following the same comparison, the value found in the La Ramada community was 0.1035 using the scaled distance method, and the found using the optimum trained artificial network was 0.5139.

Digital Object Identifier (DOI)