3
done with the impact energy of a piston inside a rock drill.
Therefore, the relevant MWD parameters also varies for
different drilling methods.
For core drilling feed force and rotational speed are
the fundamental parameters for the rock excavation and
are consequently important to measure with high quality.
However, the response parameters such as penetration rate
and rotational torque are important to monitor to control
and optimize both the drilling process and the wearing of
the drill tools. The general objectives are therefore:
• To continuously monitor the relevant drill param-
eters at predefined intervals of the drilling process.
• To distinguish and characterize transition zones in
the rock mass, using MWD data.
• To distinguish and characterize weakness zones in
the rock mass, using MWD data.
• Characterize the bit wearing process in order to pre-
dict the remaining useful life.
In this project this is done by a Cyber-Physical-System
(CPS) developed for core drilling. The work will result in
improved understanding of how MWD parameters are
responding to different features in the rock mass and how
wear influence the behavior of the drilling systems.
Objective—Predictive Maintenance
Equipment maintenance costs in the mining industry are
estimated to range from 40% to 70% of the overall setup
costs, and 20% to 60% of the operational expenditure [3].
Maintenance is not seen as an operational necessity any-
more, but as a strategic aspect, since the maintenance of the
equipment is crucial for the overall production efficiency
[4]. In the mining industry maintenance has the potential
to be a prominent controllable cost. It has not largely taken
advantage of the emerging technologies to incorporate pre-
dictive capabilities and therefore, achieve downtime reduc-
tion and to enhanced operational performance, comparing
to other industries [5].
Last but not least, the adoption of a predictive mainte-
nance strategy has been considered to increase availability of
the production equipment and to avoid unplanned down-
times through the use of condition monitoring. Predictive
Maintenance is the maintenance strategy that takes advan-
tage of the large amounts of real-time and historical data in
order to detect anomalies in equipment behaviors as early
as possible, to predict the future health state of the equip-
ment, and potential future failure modes. Proactive mainte-
nance plans can be formulated with the aim to eliminate or
mitigate the impact of the predicted failures [6].
In the context of Mine.io, the Predictive Maintenance
module covers the whole data analytics lifecycle, i.e.,
descriptive (“what happened”), predictive (“what will
happen”), and prescriptive (“what should I do”) analyt-
ics. Therefore, it is able to detect anomalies, predict the
future health state, and provide recommendations about
maintenance actions for the mining equipment (e.g., drill-
ing machines). To do this, the Predictive Maintenance
technological solution is implemented in an Autonomous
Analytics as a Service platform (AAaaS) that simplifies the
development of data-driven applications by automating the
design, configuration, execution, and deployment of (deep)
Machine Learning (ML) pipelines [7]. It enables users to
define custom analyses using an interoperable interface, to
store and manage algorithms and models, to process diverse
data sources, and to provide analytical results.
The automation of ML pipelines is enabled by
Automated Machine Learning (AutoML) approaches,
that automate the selection and training of the models
and increase the efficiency of their overall development
by reducing the time effort and time needed. Therefore,
AutoML automates the configuration of ML pipelines
which makes advanced data analytics methods more acces-
sible to non-ML experts and less labor-intensive for data
scientists. It has the potential to significantly enhance pre-
dictive maintenance applications by optimizing predictive
models and thus, increasing reliability of results [8].
Measurement System—Sensors
A Sandvik DE110 is used for the transformation from an
analogue drill rig to an intelligent drill rig with informa-
tion feedback for the machine operator. No sensors are
installed on the original drill rig to support the operator.
Consequently, the operator remains uninformed about
the directional orientation of the drilling process and
the extent to which the drill remains within the target
rock. Additionally, the operator has to manually adjust
the advance rate in accordance with the anticipated rock
strength. The wear of the drill bit is also only possible by
regular visual inspection. While special drill heads with
measuring capabilities are used for large and deep bore-
holes like in oil and gas industry, such a drill head is far too
big and too expensive for a small scale of simple boreholes
without any special requirements, which are needed pri-
marily for production in mining.
For digitization of the drill rig, the entire measurement
technology should be located exclusively in or on the actual
drilling machine. One requirement is that no structural
changes on the rig are necessary for integration. In addi-
tion to the mechanical requirements for the sensors, there
done with the impact energy of a piston inside a rock drill.
Therefore, the relevant MWD parameters also varies for
different drilling methods.
For core drilling feed force and rotational speed are
the fundamental parameters for the rock excavation and
are consequently important to measure with high quality.
However, the response parameters such as penetration rate
and rotational torque are important to monitor to control
and optimize both the drilling process and the wearing of
the drill tools. The general objectives are therefore:
• To continuously monitor the relevant drill param-
eters at predefined intervals of the drilling process.
• To distinguish and characterize transition zones in
the rock mass, using MWD data.
• To distinguish and characterize weakness zones in
the rock mass, using MWD data.
• Characterize the bit wearing process in order to pre-
dict the remaining useful life.
In this project this is done by a Cyber-Physical-System
(CPS) developed for core drilling. The work will result in
improved understanding of how MWD parameters are
responding to different features in the rock mass and how
wear influence the behavior of the drilling systems.
Objective—Predictive Maintenance
Equipment maintenance costs in the mining industry are
estimated to range from 40% to 70% of the overall setup
costs, and 20% to 60% of the operational expenditure [3].
Maintenance is not seen as an operational necessity any-
more, but as a strategic aspect, since the maintenance of the
equipment is crucial for the overall production efficiency
[4]. In the mining industry maintenance has the potential
to be a prominent controllable cost. It has not largely taken
advantage of the emerging technologies to incorporate pre-
dictive capabilities and therefore, achieve downtime reduc-
tion and to enhanced operational performance, comparing
to other industries [5].
Last but not least, the adoption of a predictive mainte-
nance strategy has been considered to increase availability of
the production equipment and to avoid unplanned down-
times through the use of condition monitoring. Predictive
Maintenance is the maintenance strategy that takes advan-
tage of the large amounts of real-time and historical data in
order to detect anomalies in equipment behaviors as early
as possible, to predict the future health state of the equip-
ment, and potential future failure modes. Proactive mainte-
nance plans can be formulated with the aim to eliminate or
mitigate the impact of the predicted failures [6].
In the context of Mine.io, the Predictive Maintenance
module covers the whole data analytics lifecycle, i.e.,
descriptive (“what happened”), predictive (“what will
happen”), and prescriptive (“what should I do”) analyt-
ics. Therefore, it is able to detect anomalies, predict the
future health state, and provide recommendations about
maintenance actions for the mining equipment (e.g., drill-
ing machines). To do this, the Predictive Maintenance
technological solution is implemented in an Autonomous
Analytics as a Service platform (AAaaS) that simplifies the
development of data-driven applications by automating the
design, configuration, execution, and deployment of (deep)
Machine Learning (ML) pipelines [7]. It enables users to
define custom analyses using an interoperable interface, to
store and manage algorithms and models, to process diverse
data sources, and to provide analytical results.
The automation of ML pipelines is enabled by
Automated Machine Learning (AutoML) approaches,
that automate the selection and training of the models
and increase the efficiency of their overall development
by reducing the time effort and time needed. Therefore,
AutoML automates the configuration of ML pipelines
which makes advanced data analytics methods more acces-
sible to non-ML experts and less labor-intensive for data
scientists. It has the potential to significantly enhance pre-
dictive maintenance applications by optimizing predictive
models and thus, increasing reliability of results [8].
Measurement System—Sensors
A Sandvik DE110 is used for the transformation from an
analogue drill rig to an intelligent drill rig with informa-
tion feedback for the machine operator. No sensors are
installed on the original drill rig to support the operator.
Consequently, the operator remains uninformed about
the directional orientation of the drilling process and
the extent to which the drill remains within the target
rock. Additionally, the operator has to manually adjust
the advance rate in accordance with the anticipated rock
strength. The wear of the drill bit is also only possible by
regular visual inspection. While special drill heads with
measuring capabilities are used for large and deep bore-
holes like in oil and gas industry, such a drill head is far too
big and too expensive for a small scale of simple boreholes
without any special requirements, which are needed pri-
marily for production in mining.
For digitization of the drill rig, the entire measurement
technology should be located exclusively in or on the actual
drilling machine. One requirement is that no structural
changes on the rig are necessary for integration. In addi-
tion to the mechanical requirements for the sensors, there