Search Documents

By N. Duda, W. Koperska, S. Anufriiev, P. Stefaniak

In mineral processing plants ore extracted from mines is processed to obtain a concentrate with the highest possible concentration of this valuable product. It is a complicated multi-stage process. Usually, the first step is sieving the ore, which is typically done using vibrating mining screens. There is often only a limited number of these machines in the technological chain, which is why their continuous operation, and proper maintenance is crucial for the whole process. Predictive maintenance is the best strategy in this case because it allows for minimising the number and duration of unexpected stoppages. It consists of the prediction of future failures based on available real-time data, and predictive models built upon this data using, for example, machine learning techniques. In this paper we concentrate on feature generation for future predictive maintenance model development for a vibrating screen, based on an example of a KGHM ore processing facility in Polkowice-Sieroszowice. We focus on two types of measurements: bearings temperature and vibrational data. Based on the raw data, the process of more advanced feature engineering is described. By transforming the temperature data we calculate characteristics of bearing cooling and heating dynamics, while based on vibrational data, the aggregated characteristics of the vibrations are derived. Such features are calculated on a long-term basis, and then they are compared to the data related to vibrating screen maintenance to estimate their significance for the prediction of particular failures of the machine.

Sep 1, 2021

By F. Navarro, F. Soto, G. Díaz, E. Jélvez, G. Morales, N. Morales

The increasing availability of tools on the web provides a great opportunity for the research community to boost investigation by accessing computational resources from all over the world. For example, data analytics libraries implemented in languages like Python can be accessed through many open and commercial platforms, therefore providing researchers with a common set of tools or shared datasets. Unfortunately, that is not the case with research in resource assessment and mine planning, where the most standardised tools are commercial applications designed to work on the desktop, but there are not many standard platforms for research. As a result, researchers from different groups often need to start from scratch in order to develop their work, and rarely share implementations details. Similarly, apart from a few publicly available examples, there is a lack of common datasets that are validated and can be used to compare results between different approaches. In this paper, we introduce a platform that provides tools for resource assessment and mine planning, which can be accessed remotely utilising Jupyter Notebooks. In its current version, the platform provides (i) definition and implementation of basic structures for managing samples and models (loading, saving, adding or removing columns, etc.), (ii) definition of experimental and modelled variograms for spatial continuity analysis, (iii) definition of standard interpolation methods in resource evaluation (inverse distance and kriging), (iv) definition of standard problems in mine planning (ultimate pit and precedence constrained production scheduling), (v) generation of 2D standard graphs like grade distribution, conditional means, tonnage-grade, and production plans (vi) a set of datasets, which can be used for case studies, and (vii) a set of notebooks with examples and documentation. In this paper, we introduce the platform and show some examples that illustrate some of the possibilities that it has for research and education in the area of resource assessment and mine planning.

Sep 1, 2021

By M. S. Shishvan, S. Klerkx, J. Sattarvand

The inaccurately drilled blastholes from the planned locations cause changes in the spatial distribution of explosive energy; some areas of the blast may receive a surplus of explosive energy, while others see a reduction. This results in less uniform fragmentation, reduces the efficiency of mining operations and increases costs. Additionally, it can lead to slope stability issues, flyrock, and uneven benches. To mitigate these problems, this study attempts to reduce differences in the explosive energy distribution (EED) through an optimization approach that adjusts the height of explosive columns in blastholes. Two optimization algorithms, the Genetic Algorithm (GA) and Tabu Search (TS), are tested. The latter is preferred for its local search neighbourhood strategy that gives superior computation times. Applications on different blast patterns result in improvements of 0.53–1.54%, or 2.14–3.94% when only the affected blocks are considered.

Sep 1, 2021

By S. Blancher, A. Perroton, H Smith, T. Nzima, T. Wallmach, V. Dieudonné

Chromite beneficiation is generally performed by crushing and density separation using spirals. At the Dwarsrivier Chrome mine spiral efficiencies seam to vary, the reason for which is not evident. This sometimes results in the production of low-grade products, which are sold at relatively low prices. This paper represents a Qemscan® study that investigated mineralogical reasons for these variations. Variations in liberation of chromite particles as well as variations in chromite compositions were investigated. The results showed that the so-called low grade chromite product can be improved to become a high grade chromite product by means of a refinement in mineral processing. A detailed Qemscan® database (SIP file) was developed to account for variations in chromite chemistries. It can be shown that chemical variations with respect to Cr, Al, Mg and Fe in chromite particles reporting to the scavenger tails are larger than in the high grade product. A combination of these chemical variations together with a decrease in liberation of chromite particles in the scavenger tails can potentially help to recover more chromite particles from this specific process stream.

Sep 1, 2021

By S. F. Böde, K. Böde

This paper focuses on developing a comprehensive and pragmatic information technology (IT) strategy for low-margin construction activities. This forms part of the realignment of the general business strategy, and refocusing the operational model. The main elements of the IT strategy formation are analysis of the current IT environment in terms of software and hardware, carrying out a gap analysis, and developing individual functional digitalisation strategies in construction execution based on Building Information Modelling (BIM), business administration processes, communication, business intelligence (BI) and reporting, and data security. Further, substantial emphasis is given to the sustainable strategy implementation in terms of hardware and software readjustments, their integration, user training, and defining the way forward.

Sep 1, 2021

By M. Kumral, S. Li, M. Balci, Z. Shao

Main haulage roads play an important role in the performance of underground mines because of the cost of transportation of the extracted ore, equipment, and workers. During mine development, these roads are designed through manual tools or CAD-based software. However, the optimisation of these roads is still a complex problem due to some constraints such as minimum turning radius and maximum gradient. By utilising Dubins path theory, three elevation cases; namely low, medium and high are defined, considering gradient constraint as well as the minimum turning radius in a three-dimensional space. The shortest path between two points can then be found by the three-dimensional Dubins path algorithm with elevation constraint. As a further step, a three-dimensional Dubins path algorithm with elevation constraint for multiple points is also developed for multi-point cases. The effectiveness of these algorithms is tested through case studies in order to determine the optimal path between different points.

Sep 1, 2021

By P. Stefanek, N. Duda, B. Jachni, P. Stefaniak, B. Bursa

The Zelazny Most Tailings Storage Facility (TSF) is the largest reservoir of post–flotation tailings in Europe. The total amount of tailings from the mines in Lubin, Rudna, and Polkowice are stored each year at the disposal facility, which reaches approximately 30 million tons. It is crucial to accurately monitor and maintain such a huge storage facility, for the safety and economics of the entire region. In general, the monitoring network installed at the Zelazny Most TSF includes four groups of components: (a) geotechnical network (VW pore pressure transducers and inclinometers), (b) hydrological network (open piezometers, water level gauges, drainage discharge measurements, (c) geodetic survey network (e.g. surface and in-depth benchmarks, laser measurements, GPS measurements within GEOMOS network and (d) seismic network (e.g. accelerometer stations, cone penetration tests.) Apart from the above, chemical water analyses, tests of soil samples, monitoring of spigotting processes, and monitoring of relief wells’ discharge are also conducted. In total, approximately 2900 monitoring instruments are installed within the area of the Zelazny Most TSF. Such large amounts of data from the various information sources makes manual data analysis by the engineers almost impossible. In this paper we discuss a machine learning driven approach to improve the quality of the monitoring and maintenance of such facilities. We focus on the seismic cone penetration test (SCPT) data analysis. First, we carefully describe the collected data, its quality and availability. Then we discuss possible events that can occur in the TSF that need to be predicted using SCPT. Based on the information obtained in these two steps, the applicability of several machine learning methods is estimated. Some of them are then tested on available historical data. The algorithms described in this article will be further developed in the IlluMINEation project (H2020).

Sep 1, 2021

By T. J. Otto, I. D. Ungerer, G. C. Lindeque, T. J. van Heerden

Anglo American Kumba Iron Ore (Kumba) continues to implement technologies focussed on improving safety and productivity; this forms part of the groups FutureSmart Mining initiative. The paper provides insights into the implementation of three technology solutions that improves the productivity of heavy mining equipment (HME). Enhancements to the short term mine planning (STMP) process, the effective use of a fleet management system (FMS) and the deployment of drones or Remotely Piloted Aircraft Systems (RPAS) are discussed. The implementation of these and other technologies contributed to a 13% and 22% increase in the overall equipment efficiency (OEE) of the primary HME at Kumba’s Sishen and Kolomela mines respectively. Four key success factors (KSF) are highlighted namely that Kumba’s technology implementation drive is a key enabler to the company’s strategy, has a strong focus on value-add, is underpinned by a holistic and integrated implementation approach and is people-centric.

Sep 1, 2021

By K. M. Seiler

Automation in mining requires accurate maps of road networks on site. Because roads on open-cut mines are dynamic in nature and continuously changing, manually updating road maps is tedious and error-prone. This paper investigates the possibility of automatically deriving an accurate representation of the road network using GPS data available from haul trucks operating on site. We present an overview of approaches proposed in literature and test the performance of publicly available methods on GPS data collected from trucks operating on site. Based on shortcomings seen in all tested algorithms, a post-processing step is developed which geometrically analyses the created road map for artefacts typical of free-drive areas on mine sites and significantly improves the quality of the final road network graph.

Sep 1, 2021

By M. E. Rossi, C. Badenhorst

Conditional bias occurs when the expected value of the true grade conditional on the estimated grade is not equal to the estimated grade. It is a quantification of how the average of estimates differs from the true value. Conditional bias is present in all resource estimates due to the unavoidable properties of kriging, and indeed all linear estimation (regression) methods. In a mining resource estimation context, the degree of conditional bias is dependent on the spacing of the conditioning data; more data leads to the expectation of less conditional bias This paper discusses conditional bias from two perspectives: 1. 2. Estimates seeking to eliminate, as much as possible, conditional bias; and Estimates that accept a larger amount of conditional bias in resource models, in order to achieve a specific outcome. The problem of conditional bias in estimation can only be fully appreciated if the objective or purpose of a resource model is understood. There are instances where conditional bias should be avoided because it is not reasonable to make decisions using estimates that are known to be wrong in expected value; for example, when making a final decision in grade control with respect to the destination of the material. In most instances in the mining resource estimation context, however, the estimated values will not be “final estimates”, but “interim estimates”, meaning that in the future more data will be obtained, and the ”interim estimate” will then be updated. In this context, the most useful resource model is an interim model that is able to predict the final model estimates, despite not having all the data available that will be collected in the future. This is the objective, for example, of a recoverable “long-term” model developed prior to operating a mine, for example at the feasibility stage. It attempts to predict the ore that a processing plant will receive, well into the future. In such cases, it can be unequivocally demonstrated that conditional bias is not only unavoidable, it is a necessity.

Sep 1, 2021

By M. Z. Naghadehi, E. Emami, J. Sattarvand, R. Battulwar

In this paper, a methodology for a computerised recognition of joint sets on 3D point cloud models of rock masses is presented. The process starts with classifying joints on a 3D rock mass surface through training a deep network architecture and validated using manually labeled datasets. Then, individual joint surfaces are identified using the Density-Based Scan with Noise (DBSCAN) clustering algorithm. Subsequently, the orientations of the identified joint surfaces are computed by fitting least-square planes using the Random Sample Consensus (RANSAC). Finally, the joint planes are classified into different joint sets, and the dip direction and dip angle for each set are calculated. The performance of the proposed methodology has been evaluated using a case study. The results show that the presented procedure is fast, accurate, and robust.

Sep 1, 2021

By T. Bohmer, L. Auret, N. Pillay, Z. Horn, P. Rambalee, A. Haasbroek

As part of a broader digital technology-enabled transformation programme, a major mining company is systematically implementing Advanced Process Control (APC) solutions and associated instrumentation throughout their operations. This paper highlights the approach and benefits of this holistic approach for a specific instrumentation installation, and APC installation at a South African coal processing plant. For the coal processing plant under consideration, a systematic approach was followed. Process sections/units are considered from upstream to downstream processing. APC solutions are designed for each process section and each process section is interrogated in terms of availability of sufficient process measurements to ensure stable, capable, and ultimately, optimised performance. Where insufficient process measurements are available, instrumentation upgrades are implemented. Finally, APC solutions are implemented, and evaluated in terms of productivity, safety, health, and environmental benefits. For the flotation section, an opportunity was identified to install froth sensors. The installed froth sensors provide measurements on froth velocity, bubble size, height, colour, and stability. The improved visibility that the froth sensors provided on the flotation process resulted in quick gains in terms of productivity: for example, instability in the feed to the flotation section could be quantified. This resulted in targeted process troubleshooting in an upstream thickener, and subsequent improvement of process stability, and reduced feed variance. The measurements from the froth sensors were further integrated in an APC solution. Proportional mass pull contributions per cell and per line can be inferred from the froth sensors, serving as controlled variables, while pulp level set points serve as the manipulated variables. On-off tests indicated that the advanced process control solution provided increases in yield and product quality.

Sep 1, 2021

By Z. Saw, C. Poveda, N. L. Mai

Optimising the throughput of a mining network comprising all equipment from shovels to semi-autogenous grinding (SAG) mills is a complicated task, yet the potential benefits can be significant. Key modelling challenges stem from the sheer number of possible routes which material can travel, different equipment operating under different constraints, and the rate of SAG mills being calculated through a nonlinear function of the blended materials’ rock attributes. Nevertheless, this is an ideal onapplication area of mathematical programming, a branch of operations research. This paper describes an application of mixed integer linear programming (MILP) to optimise the mining equipment network of one of the world’s largest copper mines.The model takes as input a year-long monthly mine schedule of shovels in different pits, pushbacks, sequences and faces which produce material to be moved through a network of crushers, conveyors, silos, distributors, stockpiles and SAG mills, ultimately outputting a daily operational plan for all equipment that maximises the throughput of the network. All operational constraints and business rules are considered, including three quality attributes of copper ore. Comparing model results with actual data shows an improvement of 4.47% in network throughput tonnage. Additionally, although the model is solved using a free, open-source solver, optimal solutions are still achieved within a practical timeframe.

Sep 1, 2021

By F. C. Harris, S. J. Gaab, J. Sattarvand, S. Duzgun

Underground mines are inherently dangerous due to potential hazards such as fires, explosions, and caving of rock. Although these events are rare, once they occur a speedy and safe egress of mineworkers is critical to survival. Static evacuation systems use exit signs or directional lines but can get obstructed or be hardly visible in smoky or dusty conditions. This paper presents the investigation of the effectiveness of a smart evacuation technology in underground mining, where a real-time guidance system leads individuals on an optimised path to the egress while avoiding danger zones, compared to conventional practices. Participants simulated an evacuation from a virtual reality mine, using the conventional system and the smart evacuation system with real-time guidance. The highest reduction in evacuation time using the smart method amounted to nearly 40% on average. Moreover, 83% of participants preferred the smart guidance system, and agreed that it could enhance safety.

Sep 1, 2021

By R. Minnitt, C. Morgan, D. Fourie

The problems associated with outliers and their adverse effects on resource estimation in deep level South African gold mines are presented, and solutions suggested in available literature are reviewed. Traditional Witwatersrand type outlier capping uses the 99th percentile grade as the upper limit, a somewhat blunt approach. The occurrence of high and low grade areas in mining operations and the estimation of resources using geostatistical techniques require that the assumption of stationarity be upheld. Honouring the stationarity of data requires that mining areas be subdivided into appropriately sized and spaced domains. In order to overcome high-grade conditionally biased block estimates, the overly strong influence of extreme grades is mitigated by using naturally occurring grade sub-domains, by identifying outliers within the grade sub-domains, allocating relevant capping limits, and weighting them in a post-estimation process to. Rather than applying a capping value across a domain it applies a maximum cap based on the extreme value and its range of influence; the method also ensures that relevant capping occurs even in lower grade sub-domains. The proposed method for capping outliers recognises that metal contained in high grade chip samples should not be over-capped, and based on the grade of surrounding samples. The post-processing method of capping outliers does not interfere with the kriging equations. The method is simple, elegant, and can be easily applied. An advantage is that the output from the kriging process is not overwritten, allowing for comparison between original kriging results and post-processing results.

Sep 1, 2021

By A. Melkumyan, K. Silversides

Chemical assays collected from blast holes have the potential to greatly improve the local accuracy of grade models. Assays must be assigned to the correct domain or they will introduce errors. Boundary surfaces created from the widely spaced exploration holes (~50m) do not provide sufficient resolution to correctly delineate domain boundaries between closely spaced (3-5m) blast holes. This paper demonstrates a probabilistic method for classifying production blast holes near domain boundaries. Gaussian processes models are applied to identify the assays and assay ratios most suited to distinguishing particular domains. These models are combined with the exploration based surfaces to classify the blast holes that can be confidently assigned to a particular domain. A reiterative method is used, where information from the blast holes closest to the exploration holes is used to label blast holes further away. This is tested on two domain boundaries in a banded iron formation-hosted iron ore mine.

Sep 1, 2021

By N. Madani, S. Abulkhair

In multivariate geostatistics, it is common to have different types of complexities between variables of interest. In this context, an inequality constraint is an example of complex bivariate relationships. Unfortunately, traditional co-kriging and co-simulation algorithms cannot reproduce this type of bivariate complexity, leading to the overestimation of disturbing elements. This paper proposes a new algorithm based on a hierarchical sequential Gaussian co-simulation framework, integrated with inverse transform sampling, to model inequality constraints between variables. First, the proposed methodology's validity was evaluated by applying it to a real case study from an iron deposit, with an inequality constraint between iron and aluminum oxide. Then the simulated results were compared with a conventional hierarchical co-simulation algorithm to investigate the effect of inverse transform sampling on the quality of the co-simulation. The results showed that the proposed algorithm can reproduce an inequality constraint between variables.

Sep 1, 2021

By S. Andika, O. A. Putri

Maintaining fleet productivity, which consists of balancing loader and truck performance, has always been a challenging task. An automatic dispatching system (ADS) is implemented to help monitoring fleet condition and assign trucks to a designated loader. Integrated data from real time Global Positioning System, Vehicle Health Monitoring System, and other production related data are acquired to support the ADS. However, the abnormal truck cycle has been the main culprit which resulted in unnecessary loader hanging time and trucks queuing time. Hanging and queuing time accounted for 2.34 and 2.07 minutes respectively per loading cycle or equal to 2,861 loose cubic metres per fleet per day production losses. Using integrated dispatching system data, the authors conducted research to study a truck’s hauling cycle that affects fleet productivity. The data is modelled to obtain the arrival rate pattern in a front loading area by adopting the queue theory algorithm. This paper aims to understand the potential pattern of a truck’s cycle abnormality and to resolve its occurrence.

Sep 1, 2021

By M. Stachowiak, M. Nowicki, K. Walas, A. Skoczylas, P. Stefaniak

In this paper the authors present a walking robot dedicated to the inspection of a belt conveyor’s transportation system in an underground mine. The developed robot can independently diagnose all of the key rotating elements of the belt conveyor (pulleys, bearings, drums, gears, engines) based on vibration and acoustic signals as well as typical RGB (red, green, blue) and thermal images. The article presents key diagnostic procedures applied to real industrial data obtained during robot testing.The work was developed as part of the ‘THING–subTerranean Haptic INvestiGator’ project. The authors presented the main assumptions of the research methodology developed for the needs of technical diagnostics of conveyor elements typical of inspection work performed by mining staff. Based on industrial data, the use of algorithms for elements such as rollers, drive stations, and the conveyor belt were shown. The main assumption of the research is to develop a robot adapted to the conditions of a deep mine, capable of carrying out an autonomous inspection of mining infrastructure. The conducted tests and the developed fault detection algorithms were aimed at preparing the robot for underground tests in the operating conditions of the mine.

Sep 1, 2021

By I. Kaabachi, F. Meunier, N. Morales

In open pit mining, ore and waste are hauled outside of the excavation area by trucks that move along a system of ramps. The design of these ramps must be carefully addressed because they have to comply with geomechanical constraints to ensure the stability of the pit walls, provide access to the different levels and phases, but also maximise the value of the extracted ore and reduce the extraction of waste. Because an optimisation approach able to deal with the design of the full ramp system seems to be out of reach at the moment, this paper addresses the problem of designing one ramp providing access to a single pushback, which has been determined beforehand. The optimisation objective is therefore to find a ramp that gives access to one given potential volume, while minimising its impact on the economic value of the volume. Two versions of the problem are formalised. One version allows the extraction of ore and waste only if this extraction is necessary for the ramp system. This version somehow minimises the intrinsic value of the ramp system. In the second version, it is possible to extract more than is needed by the ramp system. For instance, this takes into account the possibility of attaining ore on the wall opposite to the ramp as soon as the ramp reaches its level. In the setting mentioned above, the contribution of this work is three-fold. First, the computational complexity of the two different versions of the problem is presented and discussed; in particular, they are shown to be NP-hard. Second, compact integer programs modelling these two versions are proposed. Third, an efficient GRASP algorithm able to cope with the large instances of the problem is proposed. To assess the proposed approaches, the models and algorithms are applied to real instances with an off-the-shelf solver, illustrating and discussing their potential and relevance for the two versions of the problem. They are also compared with other approaches from the literature in terms of value and resulting designs.

Sep 1, 2021