Mining is one of the oldest activities of humanity, as the extraction of stones, ceramics and metals proved to be essential to develop tools and weapons and to drive forward human civilisation. Possibly the oldest mine – the “Lion Cave” – dates back to 41 000 BC. Located in Swaziland, its pre-historic operators mined haematite to make red-pigment ochre. The mine was likely in operation until 23 000 BC and at least 1200 tons of soft haematite had been removed in this timespan. As time progressed, mining diversified and production methods improved. The ancient Egyptians, Greeks and Romans mined different minerals, such as malachite, copper and gold. Philipp II, the father of Alexander the Great, is believed of having conquered gold mines in Thrace, which provided him with 1000 talents (26 tons) of gold per year. Needless to say that Alexander’s conquests would have not been possible without these extensive mining operations. Over the ages, mining activities continued to intensify. Today, a tier-one open-pit copper mine like Chuquicamata in Chuquicamata, Chile, with a depth of 900 m, provides for a production of 443,000 tons of copper and 20,000 tons of molybdenum p.a. Naturally such levels of production come with a price tag. Thousands of workers, numerous heavy machines and investments that go into the millions and billions are required to set up a mine and to maintain its operation. At the same time large amounts of waste – the so-called tailings – are generated, often posing a significant environmental risk. The fact that ore yields have dramatically decreased over time has worsened the situation; today, the extraction of 1 ton of metal ore requires vast amounts of energy and can easily generate hundreds of tons of waste.iv Were it not for a significant technological progress in the extraction, transport and processing of the ores, today’s mining operations could not be sustained. Despite all these technological advances, the mining industry is at a decision point. The conventional trend of the last hundred years of counteracting shrinking ore yields by making the mining machinery faster and bigger is at its limits. Today’s ore haulers weigh as much as 600 tons and require a net engine power of 2722 kW v to sustain operation. At the same time waste heaps have grown larger and larger – operations are clearly at their physical limits. Time is running out for enhancements and improvements, if mining is to continue, a drastic paradigm shift seems to be the only solution. This paradigm shift will require humanity to mine more efficiently and intelligently, by aiming to extract only these rocks that contain the ore and doing so in a manner, which results in the smallest possible ecological footprint. This is where OffWorld’s Swarm Robotic Mining Architecture comes into play. The overarching purpose of OffWorld is to enable the human settlement of space by developing a new generation of small, smart, learning industrial robots. This robotic workforce has numerous things to do: build landing pads, excavate underground habitats, extract water ice and materials, make drinkable water, breathable air and rocket propellant, manufacture basic structures and solar cells, produce electricity, etc. OffWorld’s overall vision is to operate thousands of robots that can mine, manufacture and build on the Moon, the as-teroids and Mars. These robots need to be small and robust, extremely adaptable, modular and reconfigurable, autonomous and fast learning – they are lightyears ahead of the 2 million industrial robots that currently work in factories and warehouses. Space is a tough place. The environment is harsh, resources are limited and the room for errors is close to zero. If a robot can succeed in space than it can surely excel in the terrestrial industry as well. This and the fact that OffWorld builds a swarm approach that relies on a small form factor, intelligence and surgical precision, has the potential to reduce the total cost of operations, can shorten the life of mine or industrial operation and can be easily scaled up and down in size. With all these benefits in mind, OffWorld is looking into a reduction in the total cost of operations of at least an order of magnitude within any industrial sector. This paper will introduce the design philosophy behind OffWorld’s robotic work-force and will present the masterplan for developing space-bound systems by first maturing them in large scale deployments in terrestrial industries.

The SOLSA project aims to develop an innovative on-line-on-mine-real-time expert system, combining sonic drilling, mineralogical and chemical characterization and data treatment. Ideally, this combination, highly demanded by mining and metallurgical companies, will speed up exploration, mining and processing. In order to evaluate the instrumental parameters for the SOLSA expert system, portable and laboratory analyses have been performed on four samples with contrasting lithologies: siliceous breccia, serpentinized harzburgite, sandstone and granite. More precisely, we evaluated the influence of the surface state of the sample on the signals obtained by portable X-Ray Fluorescence (pXRF) for chemistry and portable Infra-Red spectroscopy (pIR) for mineralogy. In addition, laboratory Raman spectroscopy, X-Ray Diffraction (XRD), XRF and ICP-OES laboratory analyses were performed to compare surface bulk mineralogical and chemical analyses. This presentation highlights (1) the importance of coupling chemical and mineralogical analytical technologies to obtain most complete information on samples, (2) the effect of the sample surface state on the XRF and IR signals from portable instruments. The last point is crucial for combined instrumental on-line sensor design and the calibration of the different instruments, especially in the case of pXRF.

For quite some time, there has been extensive research into different technologies for indoor positioning systems. Of these systems only a handful are suitable for employ in an underground mining environment. Especially as GPS is not available in underground environments, alternative systems need to be employed. Many of the currently available technologies lack the necessary precision and robustness needed to enable automation of mobile equipment. Modern approaches now look into combining different technologies to harness the best features of each candidate compensating for deficits of the other systems. In the Horizon 2020 funded Real-Time Mining research project, the Institute for Advanced Mining Technologies of RWTH Aachen University together with the Netherlands Organisation for applied scientific research (in Dutch: TNO) are also conducting research in this field. The goal is to develop an underground positioning system based on the combination of inertial measurement units (IMU), ultra-wideband radio technology (UWB) and geometrical sensors. While the partner TNO is developing a new IMU system based on the TNO DriftLess technology, RWTH Aachen University is focussing on the UWB part and laser-scanners. In the end, through shrewd sensor fusion the different technologies will be combined to enable precise localisation of mobile equipment in underground environments. Taking a closer look at the UWB technology, next to hardware and software developments, different measurement campaigns were undertaken during the time of this research project. It was found that the precision and accuracy as well as the robustness of the ultra-wideband radio technology is sufficient for the mining context. Hence, in this contribution, we will present our findings during the development of an underground localisation system for the ultra-wideband radio technology.

The main target of the UNEXMIN H2020 project (www.unexmin.eu) is to develop a fully autonomous submersible robot (UX-1) which can map flooded mine workings, and collect information about potential resources remaining in them. The most recent information about these abandoned mines could be more than 100 years old; some of them still could hold significant reserves of resources. To identify the ores/minerals in these mines many technological challenges have to be overcome: limited space and weigh for instrumentation, the UX-1 is continuously moving without contacting the mine walls and limited energy consumption because the whole robot is running only on its own battery pack. Multispectral imaging was selected as a feasible and promising method to characterize minerals. The often more than one metre of water severely limits the useful electromagnetic wave-lengths available for sensing, so the multispectral unit is designed to work between 400 to 850 nm where the water has acceptable transparency. The use of classical spectrometers is limited to single point measurements, the maximum that they can be used for is for line-scan, but this requires a powerful light source with high energy consumption. Even with the development of the 2D multispectral CCDs, there is no camera on the market which has the required channel number together with the required resolution. With the availability of high power, energy efficient monochromatic light sources (LEDs) which can be switched on and off with millisecond accuracy, the “reverse spectrometry” seems a good solution. This is where a sensitive, high resolution greyscale camera is used to record the different wavelengths in a sequence synchronized with the triggering of different wavelength light sources. The spectra of the individual points are built/ merged by the combination of the sequential images during post-processing and referred to every xyz-point. Because the mine waters can have very high dissolved ion content it can have very intense colour which can have strong effect on the measured mineral colour. Thus a reference path will be in the multispectral imaging unit continuously measuring the water transmittance to allow correction of colour effects. The wavelength selective absorption effect of the water will also be corrected with the measured distance of the multispectral imaging unit and the actual measured point. The surface roughness and inclination will also effect the actual measured intensity of a point, which can be corrected only to a certain degree, thus detected points with high inclination (higher than ca. 15–20°) will be omitted from post processing and offline interpretation. To have the best possible identification of the minerals, a database will be built, starting with the most common minerals from the test sites of the UX-1. This database will be populated with information acquired by the same multispectral imaging unit to minimize the instrumental differences of the spectra. The software control, data storage and post processing of the data is under development with Research Computing International Ltd in the UNEXMIN project. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 690008.

Motivation Mine digitization is a consequent approach to establish industry 4.0 / IoT related mine operation models based on various dimensions: flexibility, coverage, real-time capability and analytics. Networking technology, wired and wireless, can be easily deployed large scale. Miniaturized sensors, thus can be placed anywhere. Laser technology has been successfully used for more than a decade in the manufacturing industry. However, due to restrictions found in challenging heavy industry environments, such as dust, fog, rain or snow, laser technology can only rarely be found in mining applications. At the same time, technology-supported geometrical environmental scanning is essential for the control of mining machines. GPS in open pit mining is the state of the art technology for machine allocation and dispatch, whereby an underground equivalent is still missing. Because of this technology gap, many machines are frequently operated beyond their original design boundaries, and not according to the production planning which may result in significant safety impacts and collisions. Recent breakthroughs in radar technology both in 2D/3D passive scanning as well as /3D Active localization is bound to trigger a revolution in mining. In close collaboration with major universities, radar technology has been developed to mature and ruggedized industrial sensors by indurad. The public funded project “UPNS4D+” which stands for “Underground 4D+ Positioning, Navigation and Mapping System“, funded by BMBF (FKZ: 033R126), focuses on fully autonomous operated vehicles, including navigation, orientation, collision avoidance by driving autonomously around obstacles whether detected with the radar-tag system or by environmental Radar-scan. Asset and Personnel Localization Radar-tags are suitable to detect any tagged object or person. Vehicle based Radar-radios are used to measure distances and angles to radar-tags, relative to the vehicle. Any other machinery, person helmets, equipment can be tagged and thus can be localized. Based on this information, collision avoidance systems can be realized, by informing the vehicles operator or as break assistance system. Next to important localizations “geotags”, e.g. at crushers, the system can be used to exactly position vehicles, like LHDs to perfectly dump the moved material. Virtual fences can be realized to stop machinery if anyone enters a secured area. This enables fast operation e.g. at drill rigs, where manual work is required, when drill pipes have to be added. In room and pillar environments road crossings can be secured, by detecting exactly the own position at the crossing and observation other vehicles. Environmental Face and Rib Mapping Radar-scan Mapping is further, very advanced radar based technology to measure 2D planes or even the complete 3D environment around vehicles. As well infrastructure based usage might be considered, e.g. at crossings or crushers. Autonomous mapping radar scans algorithms are developed to reconstruct the surrounding and to detect the own driven trajectory including 3D translation, rotation.

‘It is always dark ahead of the pick!’ This centuries-old miners’ expression still reveals the uncertainty about the upcoming rock properties during exploration and extraction processes. It is still tough to predict what a drill rig or a cutting machine will experience during operation. However, in terms of safety, energy consumption and the performance of the whole machine it would be beneficial to be able to monitor such an extraction process. Hence, different sensors or sensor combinations are tested during cutting and drilling processes within RealTime Mining project. First aim is to depict the machine performance of the machine at any time. In a second step sensor information is also used to conclude on mechanical rock properties during the process. Measuring the machine performance for cutting and drilling is quite similar and has been condensed under the terms Monitoring-While-Cutting (MWC) respectively Monitoring-While-Drilling (MWD). Both monitoring systems contain a bundle of sensors to depict the whole process. As an example, the energy demand of such a machine can be determined by measuring the power consumption of the engines constantly. Furthermore, the process parameters like advance rates and drilling or cutting speed have to be evaluated as well to be able to depict the whole extraction machine. To conclude on mechanical rock properties several other sensor solutions have been tested and finally integrated into those monitoring systems. One of the most important rock properties for drilling and cutting is the rock strength. Increasing rock strength during an extraction process leads to increasing forces that are needed to break a certain amount of rock. Hence, e.g. measuring the torque of a drill string or the cutting forces can be an indicator on rock resistance or rock strength. Not minor important, is the characteristic rock breakage behavior which can be classified by the use of ‘acoustic’ sensors. Dependent on the rock properties that currently is drilled or cut through a characteristic fracture occurs in front of the tool. This results in audible and also inaudible characteristic acoustic waves that propagate through the machine body and can be gathered on the machine by piezo-electric sensors. The interpretation of these signals could lead to a material classification already during the extraction process. Several tests of these sensor technologies have been conducted in laboratory environment as well in field tests. The most promising results are going to be presented.

The SOLSA project aims to construct an analytical expert system for on-line-on-mine-real-time mineralogical and geochemical analyses on sonic drilled cores. A profilometer is indispensable to obtain reliable and quantitative data from RGB and hyperspectral cameras, and to get 3D definition of close-to-surface objects such as rheology (grain shape, grain size, fractures and vein systems), material hardness and porosities. Optical properties of minerals can be analyzed by focusing on the reflectance. Preliminary analyses were performed with the commercial scan control profilometer MI-CRO-EPSILON equipped with a blue 405 nm laser on a conveyor belt (depth resolution: 10 μm; surface resolution: 30x30 μm2 (maximum resolution; 1m drill core/4 min). Drill core parts and rocks with 4 different surface roughness states: (1) sonic drilled, (2) diamond saw-cut, polished at (3) 6 mm and (4) 0.25 μm were measured (see also abstract Duée et al. this volume). The ΜICRO- EPSILON scanning does not detect such small differences of surface roughness states. Profilometer data can also be used to access rough mineralogical identification of some mineral groups like Fe-Mg silicates, quartz and feldspars). Drill core parts from a siliceous mineralized breccia and laterite with high and deep porosity and fractures were analyzed. The determination of holes’ convexity and fractures) is limited by the surface/depth ratio. Depending on end-user’s needs, parameters such as fracture densities and mineral content should be combined, and depth and surface resolutions should be optimized, to speed up “on-line-on-mine-real- time” mineral and chemical analyses in order to reach the target of about 80 m/day of drilled core.

The main target of the UNEXMIN project is to develop a fully autonomous submersible robot (UX-1) which can map flooded underground mines, and also deliver information about the potential raw materials of the mines. There are ca. 30 000 abandoned mines in Europe, from which many of them still could hold significant reserves of raw materials. Many of these mines are nowadays flooded and the latest information about them could be more than 100 years old. Although it is giving limited information, magnetic measurement methods, which detecting the local distortions of the Earth’s magnetic field can be very useful to identify raw materials in the mines. The source of the magnetic field which is independent of any human events comes from the Earths own magnetic field. The strength of this field depends by the magnetic materials in the near environment of the investigated point. The ferromagnetic materials have powerful effect to influence the magnetic field. In the nature, iron containing minerals, magnetite and hematite have the most powerful effect usually. The magnetic measurement methods are rapid and affordable techniques in geophysical engineering practice. For magnetic field strength and direction measurement FGM-1 sensors (manufactured by Speake & Co Llanfapley) were selected for the UX-1 robot. The sensor heads overall dimension are very small and their energy consumption is negligible. The FGM-1 sensor was placed and aligned in a plastic cylinder to ensure that the magnetic-axis aligned with the mechanical axis of the tube for more accurate measurement. There are 3 pairs of FGM-1 sensors needed for the proper determination of the current magnetic field (strength and direction). The position of sensor pairs need to be perpendicular compared to each other. The 3 pairs of FGM-1 sensors generate an arbitrary position Cartesian coordinate system. We further developed / had installed temperature sensors to all FGM-1 probes, to compensate the temperature dependency even though it has small effect. The UX-1 robot also contains the electronic block, which controls the three FGM-1 magnetic field sensor pairs, and store the measured data. The block contains the power module, the sensor interface modules with temperature compensation, the microcontroller module and the RS485 communication module also. The output data is a temperature compensated frequency value for each sensor pair. The measured magnetic signal from the local XYZ coordinate system (local for the UX-1) should be converted to a universal coordinate system during post processing of the data. The exact position, facing and inclination of the robot must be known in the whole dive time to be able to do the above conversion. The measured magnetic signal will be placed into the measured mine map, reconstructed from the delivered 3D point cloud, thus the exact location of the magnetic anomalies can be identified. Not much magnetic source is estimated in the operating environment of the robot, but its own generated magnetic noise can be significant. There will be many cooling fans, micro-controllers and multiple thrusters inside the pressure-hull of the UX-1, which generate magnetic field. The constant magnetic noise coming from the cooling fans can be compensated, but the varying fields caused by eg. the different thrusters’s speed is problematic. We design a calibration method, where the effect of the main thrusters (even with changing speed) and the effect of the constant cooling fans could be compensated. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 690008.

Mining operations rely on effective extraction policies, which base on concerted management and technical arrangements. In addition to commodities, mining of data is the increasingly matter of subject in mining engineering. The Horizon 2020 project – Real-Time-Mining supports the ongoing paradigm shift of pushing mining activities from discontinuous to continuous operation. In this respect, the partners TU Bergakademie Freiberg (TU BAF) and IBeWa Consulting tackle the issue of physical and logical data acquisition in underground mining. The first aspect of the project addresses the ‘logical’ provision of data. Mining technology is increasingly interacting among each other and integrated into globally distributed systems. At the same time, the integration of current mining devices and machineries into superordinated systems is still complex and costly. This means only a few number of mining operators is capable to integrate their operation technology into a Supervisory Control and Data Acquisition (SCADA) system. TU BAF presents the middleware OPC Unified Architecture, which is a platform independent middleware for data exchange and technology interconnection among distributed systems. By installing a SCADA demonstrator at the research and education mine Reiche Zeche, TU BAF intends to present the technical feasibility of a SCADA system basing on OPC UA even for SME mining operations. The second aspect of the project addresses the ‘physical’ provision of data via wireless transmission. The targeted use cases are mobile machineries and the surveillance of remote mine sites. Mobile machineries in underground mining are increasingly equipped with data management and autonomous operation systems. Correspondent data exchange to superordinated systems is mostly realized via Wireless Local Area Network (WLAN). A comprehensive WLAN signal coverage, however, is generally not maintained in under-ground mines due to lacking technical and economic feasibility. With the intention to in-crease the coverage/expense ratio at underground WLAN installations, TU BAF and IBe-Wa Consulting installed a WLAN test loop at Reiche Zeche mine basing on leaky feeder cables. Simultaneously, IBeWa Consulting pushes forward the surveilability of remote and/or hardly accessible mining sites by Through The Earth (TTE) data transmission. Current test performances present an enhanced stability for data transmission at ore / gneiss formations beyond 200m, primarily basing on a better alignment of the system to the isotropic characteristics of the bedrock.

This paper presents the development and quantification of a catalogue of Sustainable Performance Indicators (SPIs) for the assessment of the benefits real-time mining can offer in small and complex mechanised underground mining operations. The SPIs investigated in detail include: ‒ grade accuracy and error of the resource model, ‒ high/low grade ore classification accuracy and error, ‒ additional high grade ore identified per unit volume, ‒ profit expected per unit volume, ‒ ore classification accuracy per unit volume assigned to the stockpiles. A case study utilising the Red Lake gold mine located in Northwestern Ontario, Canada, which is owned by Goldcorp Inc., was designed with the aim to assess the effect of real time sensory data acquisition and resource model update on the SPIs. The methodology broadly comprises of three steps. Firstly, the provided dataset was used to develop a virtual asset model (VAM) representing the true 3D grade distribution in order to simulate the ‘sublevel cave and fill’ mining method and the associated grade data acquisition from the development drillholes and face monitoring, the development and production muck pile, LHD/scooptram and conveyor belt transport, taking into account the sensor parameters. Next, the acquired data was assimilated into the models developed for the purpose of detailed statistical assessment of the SPIs, thereby enabling optimised decision-making during the production of ore in order to meet the grade requirements. Finally, an evaluation of the sensor performance was carried out using three additional levels of sensor error and interpretation bias (10, 20 and 30%). The three models used for the quantification of the SPIs include: ‒ resource block model (RBM): which represents the 3D grade distribution in the ore body; ‒ grade control model: which enables selective stope production (drilling, charging and blasting) based on the underlying requirements pertaining to e.g. cut-off grade, time and economic constraints; and ‒ logistics model: which classifies the ore grades for conveyance and stockpiling, in order to eventually facilitate for the mixing of run of mine ore to meet the grade requirements before milling at the processing plant. The improvement of the SPIs when real time monitored data is used in the update of the models has been verified. It is also shown that the noise in the acquired data, which directly reflects both the accuracy and precision of the sensors, has a measurable effect on the values obtained for the SPIs. However, 10 to 20% noise does not appear to reduce significantly the improvements achieved, while 30% noise has a more profound effect on the SPIs and the quality of improvements achieved through real time data assimilation in the models. The work carried out demonstrates that there is a need for robust sensor technologies that allow for minimum bias in grade estimation and maximum classification accuracy. It is also expected that sensor performance is likely to vary from site to site and possibly within the same ore deposit mined due to local geological conditions (heterogeneity), variations in the underground environment were sensors are installed (affecting sensor performance), the mining method used (affecting the access and availability of real time monitored data) as well as the specifics of the sensor technologies used. Thus, it is suggested that sensor performance needs to be evaluated and quantified for the mine and area considered for sensor installation given the local geological, operational and mining method related characteristics and opportunities for monitoring.