Mining Goes Digital – Mueller et al. (Eds)
© 2019 Taylor & Francis Group, London, ISBN 978-0-367-33604-2
671
Enhancement of explosive energy distribution using UAVs and
machine learning
J. Valencia, R. Battulwar, M. Zare Naghadehi & J. Sattarvand
Department of Mining and Metallurgical Engineering, University of Nevada, Reno, USA
ABSTRACT: Blasting engineers need right information at the right time to achieve safe
and productive blasting. This paper explains how to use Image Processing (IP) and Machine
Learning (ML) to compensate the drilling inaccuracy through adjustment of the drillhole
explosives in order to enhance the Explosive Energy Distribution (EED). Accordingly, images
of the drilled blasting site are collected by a specially programmed Unmanned Aerial Vehicle
(UAV) and processed by a photogrammetry software to provide orthomosaic map and Digital Elevation Model (DEM) of the site. Subsequently, the models are analyzed to identify
as-built locations of the drillholes, and finally, an optimization model is run to manipulate
the amount and configuration of the charges in each drillhole in order to reach the most uniform explosive energy distribution in the rock mass. Current paper’s focus is on the process
of models generation and identification of the exact locations of the drillholes.
1 INTRODUCTION
The final position of blast holes still is a task that is not completely controlled in mine sites.
The best control for this purpose would be surveying, but usually the surveyors in the mines
are dedicated to many other activities and do not exclusively control drillholes. Although
majority of new drill rigs have accurate GPS devices, but comparison of the designed patterns with the results of surveyed drillholes in some case studies surprisingly shows a considerable difference (Valencia López et al, 2015). These errors can be attributed to a number
of sources such as the loss of signal, improper coverage in the site, malfunctioning hardware,
etc. Therefore, blasters may not reach the perceived goals without knowing of the correct
as-built drillhole locations and all the analysis, calculations, modeling and simulation would
therefore be based on incorrect information. However, the time-consuming task of surveying of all drillholes is a challenging procedure which cannot be easily accommodated in an
open pit mining operation. Present study proposes a solution to control the final drillhole
locations using Drone Imaging, Image Processing and Machine Learning. This is a running
research with some uncompleted components and the paper will be focused on the process
of data collection (imaging) and generating the blast site models as well as the automated
drillhole location identification.
Previous modelling results have indicated that changes in post-blast fragment strength
significantly influences the performance of the SAG mill, producing up to a 20% increase in
throughput (Michaux & Djordjevic, 2005). However, the effective explosive energy represents
only around 26% of the total available energy (Sanchidrián et al, 2007). In order to obtain an
optimal EED, it is crucial to keep a good control of the drilling accuracy.
General procedure of the developed technology is depicted in Figure 1. The process starts
with taking high resolution images from the drilled site. A conventional off-the-shelf drone
such as DJI products can be used for this purpose, however, in order to take high-resolution
pictures (few mm/pixel) the drone needs to get close to the ground and take images at 20 to
30 meters above the ground level. A special ground control application is programmed to
accomplish this step. Next, a commercial photogrammetry software is used to process the
© 2019 Taylor & Francis Group, London, ISBN 978-0-367-33604-2
671
Enhancement of explosive energy distribution using UAVs and
machine learning
J. Valencia, R. Battulwar, M. Zare Naghadehi & J. Sattarvand
Department of Mining and Metallurgical Engineering, University of Nevada, Reno, USA
ABSTRACT: Blasting engineers need right information at the right time to achieve safe
and productive blasting. This paper explains how to use Image Processing (IP) and Machine
Learning (ML) to compensate the drilling inaccuracy through adjustment of the drillhole
explosives in order to enhance the Explosive Energy Distribution (EED). Accordingly, images
of the drilled blasting site are collected by a specially programmed Unmanned Aerial Vehicle
(UAV) and processed by a photogrammetry software to provide orthomosaic map and Digital Elevation Model (DEM) of the site. Subsequently, the models are analyzed to identify
as-built locations of the drillholes, and finally, an optimization model is run to manipulate
the amount and configuration of the charges in each drillhole in order to reach the most uniform explosive energy distribution in the rock mass. Current paper’s focus is on the process
of models generation and identification of the exact locations of the drillholes.
1 INTRODUCTION
The final position of blast holes still is a task that is not completely controlled in mine sites.
The best control for this purpose would be surveying, but usually the surveyors in the mines
are dedicated to many other activities and do not exclusively control drillholes. Although
majority of new drill rigs have accurate GPS devices, but comparison of the designed patterns with the results of surveyed drillholes in some case studies surprisingly shows a considerable difference (Valencia López et al, 2015). These errors can be attributed to a number
of sources such as the loss of signal, improper coverage in the site, malfunctioning hardware,
etc. Therefore, blasters may not reach the perceived goals without knowing of the correct
as-built drillhole locations and all the analysis, calculations, modeling and simulation would
therefore be based on incorrect information. However, the time-consuming task of surveying of all drillholes is a challenging procedure which cannot be easily accommodated in an
open pit mining operation. Present study proposes a solution to control the final drillhole
locations using Drone Imaging, Image Processing and Machine Learning. This is a running
research with some uncompleted components and the paper will be focused on the process
of data collection (imaging) and generating the blast site models as well as the automated
drillhole location identification.
Previous modelling results have indicated that changes in post-blast fragment strength
significantly influences the performance of the SAG mill, producing up to a 20% increase in
throughput (Michaux & Djordjevic, 2005). However, the effective explosive energy represents
only around 26% of the total available energy (Sanchidrián et al, 2007). In order to obtain an
optimal EED, it is crucial to keep a good control of the drilling accuracy.
General procedure of the developed technology is depicted in Figure 1. The process starts
with taking high resolution images from the drilled site. A conventional off-the-shelf drone
such as DJI products can be used for this purpose, however, in order to take high-resolution
pictures (few mm/pixel) the drone needs to get close to the ground and take images at 20 to
30 meters above the ground level. A special ground control application is programmed to
accomplish this step. Next, a commercial photogrammetry software is used to process the