Mining Goes Digital – Mueller et al. (Eds)
© 2019 Taylor & Francis Group, London, ISBN 978-0-367-33604-2
557
Artificial intelligence using real-time data
L.-P. Campeau & M. Dubois
Newtrax Technologies, Montreal, Canada
ABSTRACT: Great changes have been taking place in the underground mines of the world
for several years now. More and more data is being collected at each stage of the extraction
process, at ever-increasing frequency and accuracy. In this context, the use of this data poses
new challenges in analysis and prediction. This presentation will discuss the various Machine
Learning techniques used by Newtrax to extract value out of this data and examples of concrete applications of these techniques.
1 INTRODUCTION
The digitalization of underground mines has recently opened a whole new area of possibilities in safety, automation and optimization of operations. One of these applications is in
the field of predictive maintenance, where the connectivity of equipment can be leveraged
to greatly improve maintenance policies. This article will present the recent progress and
applications made by Newtrax in the field of predictive maintenance of underground mobile
equipment using real-time data and Artificial Intelligence (AI). First, an introduction to the
research context and definitions will be presented, followed by a description of the models
used, and their application’s results.
2 CONTEXT AND DEFINITIONS
Given that the audience of this paper comes from many different backgrounds and that it
brings together notions that might not be familiar to all of them, the first part of this article
will be dedicated to defining the context of this research and the important concepts.
2.1 Newtrax system
Newtrax provides Internet-of-Things (IOT) solutions for real-time data collection in underground hard rock mines. It supplies the hardware required to collect and transmit the data as
well as the software solutions for safety and productivity applications. The big data collected
through the system come from a large variety of sources. The positioning data of personnel is
collected through tracking-enabled cap lamps (personal safety devices), environmental data
is collected through geotechnical and ventilation instruments and vehicle data like positioning and sensor telemetry is collected through a large number of sensors that are adapted to all
equipment brands and models. The resulting contextualized data raises interest for Machine
Learning applications for many reasons, for example: the goal of the operation is very simple
(extract ore), the system is relatively isolated underground, the granularity of data is high
(samples every second), years of historical labels exist, and more. The Newtrax system has
been collecting data now for many years across all continents at more than 75 mine sites.
Figure 1 illustrates the personnel tracking solution.
© 2019 Taylor & Francis Group, London, ISBN 978-0-367-33604-2
557
Artificial intelligence using real-time data
L.-P. Campeau & M. Dubois
Newtrax Technologies, Montreal, Canada
ABSTRACT: Great changes have been taking place in the underground mines of the world
for several years now. More and more data is being collected at each stage of the extraction
process, at ever-increasing frequency and accuracy. In this context, the use of this data poses
new challenges in analysis and prediction. This presentation will discuss the various Machine
Learning techniques used by Newtrax to extract value out of this data and examples of concrete applications of these techniques.
1 INTRODUCTION
The digitalization of underground mines has recently opened a whole new area of possibilities in safety, automation and optimization of operations. One of these applications is in
the field of predictive maintenance, where the connectivity of equipment can be leveraged
to greatly improve maintenance policies. This article will present the recent progress and
applications made by Newtrax in the field of predictive maintenance of underground mobile
equipment using real-time data and Artificial Intelligence (AI). First, an introduction to the
research context and definitions will be presented, followed by a description of the models
used, and their application’s results.
2 CONTEXT AND DEFINITIONS
Given that the audience of this paper comes from many different backgrounds and that it
brings together notions that might not be familiar to all of them, the first part of this article
will be dedicated to defining the context of this research and the important concepts.
2.1 Newtrax system
Newtrax provides Internet-of-Things (IOT) solutions for real-time data collection in underground hard rock mines. It supplies the hardware required to collect and transmit the data as
well as the software solutions for safety and productivity applications. The big data collected
through the system come from a large variety of sources. The positioning data of personnel is
collected through tracking-enabled cap lamps (personal safety devices), environmental data
is collected through geotechnical and ventilation instruments and vehicle data like positioning and sensor telemetry is collected through a large number of sensors that are adapted to all
equipment brands and models. The resulting contextualized data raises interest for Machine
Learning applications for many reasons, for example: the goal of the operation is very simple
(extract ore), the system is relatively isolated underground, the granularity of data is high
(samples every second), years of historical labels exist, and more. The Newtrax system has
been collecting data now for many years across all continents at more than 75 mine sites.
Figure 1 illustrates the personnel tracking solution.